Unsaturated PDF

UNSATURATED SOILS
PROCEEDINGS OF THE FIFTH INTERNATIONAL CONFERENCE ON UNSATURATED SOILS, BARCELONA, SPAIN, 68 SEPTEMBER 2010
Unsaturated Soils
Editors
Eduardo Alonso & Antonio Gens
Technical University of Catalonia, Barcelona, Spain
VOLUME 1
CRC Press/Balkema is an imprint of the Taylor & Francis Group, an informa business 2011 Taylor & Francis Group, London, UK Typeset by Vikatan Publishing Solutions (P) Ltd., Chennai, India Printed and bound in Great Britain by Antony Rowe (a CPI Group company), Chippenham, Wiltshire All rights reserved. No part of this publication or the information contained herein may be reproduced, stored in a retrieval system, or transmitted in any form or by any means, electronic, mechanical, by photocopying, recording or otherwise, without written prior permission from the publisher. Although all care is taken to ensure integrity and the quality of this publication and the information herein, no responsibility is assumed by the publishers nor the author for any damage to the property or persons as a result of operation or use of this publication and/or the information contained herein. Published by: CRC Press/Balkema P.O. Box 447, 2300 AK Leiden, The Netherlands e-mail: [email protected] www.crcpress.com www.taylorandfrancis.co.uk www.balkema.nl ISBN: 978-0-415-60428-4 (set of 2 volumes Hbk + CD-rom) ISBN: 978-0-415-60429-1 (Vol 1) ISBN: 978-0-415-60430-7 (Vol 2) ISBN: 978-0-203-83401-5 (eBook)
Unsaturated Soils Alonso & Gens (eds) 2011 Taylor & Francis Group, London, ISBN 978-0-415-60428-4
Table of contents
Preface Organization
xvii xix
VOLUME 1 Keynote lectures

Design of stiffened slabs-on-grade on shrink-swell soils J.-L. Briaud, R. Abdelmalak & X. Zhang Geotechnical and environmental models involving unsaturated soils and rocks S. Olivella Energy geotechnology: Implications of mixed fluid conditions J.C. Santamarina & J. Jang 3 15 33
General reports
Geotechnical problems and case histories from the perspective of unsaturated soil mechanics C.W.W. Ng & A.K. Leung Constitutive modelling of unsaturated soils: Discussion of fundamental principles D. Sheng Unsaturated soils: Compacted versus reconstituted states A. Tarantino 53 91 113
Unsaturated soil behaviour

Microstructure
Development of microstructure in compacted earthen building materials C.T.S. Beckett & C.E. Augarde Evolution of meso-structure of intact loess during triaxial shear test X.-W. Fang, C.-N. Shen, Z.-H. Chen & Z.-B. Zhou A model for capillary pressure in unsaturated granular soils within the pendular saturation regime J.L. Justo & M. Vzquez An insight into role of clay-fluid molecular interactions on the microstructure and macroscale properties of swelling clays D.R. Katti, K.S. Katti, P.M. Amarasinghe & S.M. Pradhan Volume changes of swelling clayey soils at microscopic scale level T. Maison, F. Laouafa & J.-M. Fleureau Influence of the microstructure on the hydro-mechanical behaviour of a natural silty clay M.M.A. Mascarenha, M.P. Cordo-Neto & E. Romero 139 145
151
157 161 167
Stress-strain and strength characteristics

Hydro-mechanical response of an expansive silt-bentonite mixture M. Ajdari, H. Nowamooz, F. Masrouri, G. Habibagahi & A. Ghahramani Deformation behavior in unsaturated soils R. Baltodano-Goulding Collapse potential of a compacted weathered serpentinite from Skiros Island, Greece M. Bardanis & S. Grifiza Collapsibility study for tropical soil profile using oedometric tests with controlled suction J.C.B. Benatti, M.G. Miguel, R.A. Rodrigues & O.M. Vilar Stress path effects on shear stiffness, damping ratio and volumetric behaviour of normally consolidated unsaturated compacted clay M. Biglari, A. dOnofrio, C. Mancuso, M.K. Jafari & A. Shafiee Shrinkage during wetting of fine-pored materials: Does this accord with the principle of effective stress? G.E. Blight Shear strength of an unsaturated silty sand F. Casini, P. Minder & S.M. Springman Dilatancy and shear strength of an aged compacted fill A.C.F. Chiu & X. Zhao An experimental study on the cyclic behavior of natural loess Y.-J. Cui, A.-M. Tang, J.P. Karam & J.M. Terpereau Inundation collapse strain and creep measurements for two Campanian pyroclastic soils P. De Simone Simple shear tests on unsaturated undisturbed specimens containing a weak layer J.L. Deng, H. Kameya, Y. Tsutsumi, J. Koseki & J. Kuwano Volume change behaviour due to moisture variation in vertisoil from a semiarid region in Pernambuco, Brazil S.R.M. Ferreira Constant water content triaxial compression tests with a compacted soil G.B. Georgetti & O.M. Vilar The relationship between the critical state shear strength of unsaturated soils and the soil-water characteristic curve J.A. Infante Sedano & S.K. Vanapalli Unsaturated shear strength parameters of tropical volcanic residual soils R. Karlinasari & P.P. Rahardjo Direct shear test under constant volume state for unsaturated soil S. Kato, B.S. Kim & S. Shibuya Effects of matric suction on elastic shear modulus for three unsaturated soils S. Kawajiri, S. Shibuya, S. Kato & T. Kawaguchi Mechanical behaviors and water retention characteristics of micro-porous ceramic particles produced by burning sludge J. Kawazu, K. Kawai, S. Kanazawa & A. Iizuka Effect of suction hysteresis on the shear strength of unsaturated soil interfaces C.N. Khoury & G.A. Miller Critical state behaviors of unsaturated soil in the direct shear test under constant pressure B.S. Kim, S. Kato & S. Shibuya 175 181 187
193
199
205 211 217 223
229 235
241 247
253 259 265 271
277 283 289
vi
The decrease of the strength of unsaturated silty sand and the Bishops parameter J.C. Leal, G. Gallegos & E. Rojas Study on gas phase in unsaturated soils Y. Liu, C.G. Zhao & Q. Fang A new unsaturated direct shear apparatus for measuring shear strength of unsaturated soils Part 2: Results and comparison between plane strain and triaxial loading conditions A. Mirzaii, S.S. Yasrebi & E. Hefzi A constitutive relationship for collapsible soils in terms of stress state variables J.D. Nelson, C.N. Hatton & K.C. Chao Direct shear strength of compacted bentonite under different suctions T. Nishimura, H. Rahardjo & J. Koseki Investigation of the behavior of an unsaturated sand using a cyclic direct shear device T. Nishimura, I. Shahrour & H.B. Bian Pore pressure under undrained condition for compacted unsaturated silty soil having a low matric suction T. Nishimura & J. Koseki The relationship between the elastic and shear modulus of unsaturated soils W.T. Oh & S.K. Vanapalli Shear strength of a pyroclastic unsaturated soil from suction-controlled direct shear tests R. Papa & M.V. Nicotera Experimental study of the collapsible behavior of a tropical unsaturated soil R.A. Rodrigues & O.M. Vilar Swelling pressure evolution in compacted bentonite: Experiments and modelling M. Snchez, M.V. Villar, R. Gmez-Espina, A. Lloret & A. Gens Constant volume triaxial compression tests on unsaturated soil prepared from slurry M. Shimizu & J. Terakata Assessment of the cv coefficient for unsaturated Warsaw Clay Z. Skutnik, M. Lendo-Siwicka & K. Garbulewski Laboratory investigations on static liquefaction potential of pyroclastic soils involved in rainfall-induced landslides of the flow-type G. Sorbino, G. Migliaro & V. Foresta Desorption and consolidation behaviour of initially saturated clays S. Tripathy, R. Bag & H.R. Thomas Compressive strength of unsaturated fissured clays L.E. Vallejo Effect of suction and plasticity on the shear strength of sand/silt mixtures M.V. Villar, P.L. Martn, R. Gmez-Espina & A. Lloret Experimental study on K0 coefficient of Ningming unsaturated expansive soil H.P. Yang, J. Xiao, G.F. Zhang & R. Zhang Experimental investigation of stress increment direction dependency of plastic flow in unsaturated soils S. Zargarbashi & N. Khalili
297 303
309 317 323 329
335 341 347 353 359 365 371
375 381 387 391 397
403
Water retention and hydraulic behaviour

Laboratory determination of Soil-Water Characteristic Curves for cracked soil M.M. Abbaszadeh, S. Houston, C. Zapata, W. Houston, B. Welfert & K. Walsh The void ratio dependency of the retention behaviour for a compacted clay C. Air Farulla, A. Battiato & A. Ferrari 409 417
vii
Comparison between the in situ and laboratory water retention curves for a silty sand A. Askarinejad, F. Casini, P. Kienzler & S.M. Springman On the inverse problem solution for laboratory determination of unsaturated water hydraulic conductivities K.V. Bicalho, J.-M. Fleureau, Y.-J. Cui & C. Romanel Water retention curve and shrinkage of a waste from the paper industry J.C.O. Bizarreta & T.M.P. de Campos Soil-water characteristic curves obtained through the wetting paths for a tropical soil profile B.H. Bonder & M.G. Miguel An investigation on the effect of initial water content and dry density on drying soil-water characteristic curve of a cohesionless soil M. Chetia & S. Sreedeep A transfer function of a soil water characteristic curve model for sands E. Imre, K. Rajkai, R. Genovese & C. Jommi Prediction of soil-water characteristic curve using neural network A. Johari & A.A. Javadi Hysteresis between wetting and drying diffusivity parameters D. Mabirizi & R. Bulut A simple method to consider water density changes in the calculation of the degree of saturation of swelling clays D. Marcial Hydromechanical behaviour of a natural sand using different suction controlled techniques H. Nowamooz, M.I. Arsenie, C. Chazallon & F. Masrouri Investigations on water retention behaviour of deformable soils S. Salager, A. Ferrari, M. Nuth & L. Laloui Study on the soil-water characteristic curve including stresses of an unsaturated soil C.-N. Shen, X.-W. Fang, Z.-H. Chen & Z.-B. Zhou Unsaturated hydraulic characteristics of soil with significant oversize particles S. Somasundaram, T. Shenthan, C. Benson & S. Nannapaneni Unsaturated hydraulic properties of fine-grained soil from the blanket layer of old railway lines in France V.N. Trinh, A.-M. Tang, Y.-J. Cui, J.C. Dupla, J. Canou, N. Calon, A. Robinet, L. Lambert & O. Schoen Soil-Water Characteristic Curve (SWCC) and volumetric deformation law for a plastic clay under high suction M. Vzquez, P. Durand & J.L. Justo Atmospheric drying and laboratory wetting of stockpiled product coal D.J. Williams A study of the water retention characteristics and volume change of lime-treated London Clay X. Zhang, M. Mavroulidou, M.J. Gunn & Z. Cabarkapa Comparison of four methods for determining the soil water retention curve M. Zielinski, P. Sentenac, A. Atique, M. Snchez & E. Romero
423
429 433 441
447 453 461 467
473 479 485 491 495
501
509 513
519 525
Thermal and chemical effects

Hydro-mechanical behavior of pre-heated bentonite Y.F. Arifin & T. Schanz Lime stabilisation of pyroclastic soils M. Cecconi, V. Pane, F. Marmottini, G. Russo, P. Croce & S. dal Vecchio 533 537
viii
Effect of temperature on collapse potential of kaolin clay A. Haghighi, G. Medero, P. Woodward & L. Laloui Influence of freeze-thaw action on mechanical behavior of unsaturated crushable volcanic soil T. Ishikawa, S. Miura & K. Itou Influence of temperature on the water retention capacity of soils analysis using solution thermodynamics A.C. Jacinto, A. Ledesma, M.V. Villar & R. Gmez-Espina Influence of temperature rise under undrained condition on the shear behavior of unsaturated sandy soil considering solubility of gaseous components Y. Katsuno, M. Iwabuchi, K. Ozawa, R. Uzuoka & M. Kazama Thermal pulse effects on the stiffness degradation of unsaturated clayey materials A. Lima, J.A. Pineda & E. Romero Effects of a microbiological compound for the stabilisation of compacted soils on their microstructure and hydro-mechanical behaviour L. Morales, E. Garzn, E. Romero & C. Jommi Influence analysis of the contaminants fluid in a compacted soil collapse E.Q. Motta & S.R.M. Ferreira Thermal damage in unsaturated geomaterials M.M. Kharazi, C. Arson & B. Gatmiri Thermo-hydro-mechanical behaviour of an expansive bentonite/silt mixture using different suction controlled techniques H. Nowamooz, R. Gmez-Espina, M.V. Villar & F. Masrouri Direct measurement of thermal expansion in unsaturated soils X. Pintado & A. Lloret Determining the thermal conductivity as a function of water content for a soil with different percentages of gravels C.M. Rubio, R. Josa, D.R. Cobos, C.S. Campbell & F. Ferrer Temperature effects on soil-water retention properties of densely compacted GMZ01 bentonite W.M. Ye, M. Wan, B. Chen, Y.-J. Cui & J. Wang
543 549
555
561 567
573 579 585
589 593
597 601
Testing techniques: Laboratory, in situ and centrifuge

Evaluation of filter paper calibrations for indirect determination of soil suctions of unsaturated soils K.V. Bicalho, F.A.M. Marinho, J.-M. Fleureau & A. Gomes Correia Useful laboratory guidelines for unsaturated soils research J.L. Boyd & V. Sivakumar Centrifuge modeling applied to swelling soils B. Caicedo & C. Segura Centrifuge modeling of soil atmosphere interaction B. Caicedo, J. Tristancho & L. Thorel Physical modelling on unsaturated soils using centrifuge B. Caicedo, J. Tristancho & L. Thorel Commentaries on the consolidation-swell test K.C. Chao, J.D. Nelson, D.D. Overton & E.J. Nelson Detecting water content changes in sand samples by means of Electrical Resistivity Tomography R.M. Cosentini, G. Della Vecchia, S. Foti & G. Musso Calibration of a TDR probe in an expansive soil S. Fityus & K. Rajkai 609 615 621 627 633 641
647 653
ix
Effective thermal conductivity of partially saturated soils S.S. Hamuda, M. Rouainia & B.G. Clarke A new unsaturated direct shear apparatus for measuring shear strength of unsaturated soils Part 1: Apparatus features and test procedures E. Hefzi, S.S. Yasrebi & A. Mirzaii Constitutive modeling of unsaturated soil behavior using a refined suction-controlled true triaxial cell: Preliminary observations L.R. Hoyos, D.D. Prez-Ruiz & A.J. Puppala Modeling unsaturated soil response under large deformations using a novel suction-controlled ring shear device L.R. Hoyos, C.L. Velosa & A.J. Puppala Small-strain stiffness of compacted silty sand using a proximitor-based suction-controlled resonant column device L.R. Hoyos, E.A. Suescn, J.A. Pineda & A.J. Puppala Efficient measurement of unsaturated hydraulic conductivity using a wetting front advancing method X. Li, L.M. Zhang & D.G. Fredlund Advances in tensiometer-based suction control systems S.D.N. Loureno, D. Gallipoli, D.G. Toll, C.E. Augarde & F.D. Evans A new permeameter for the determination of the unsaturated coefficient of permeability M.P.H. Moncada & T.M.P. de Campos Suction measurements on a natural unsaturated soil: A reappraisal of the filter paper method J.A. Muoz-Castelblanco, J.-M. Pereira, P. Delage & Y.-J. Cui The use of filter paper method for field suction investigation M.G.G. Nacinovic, C.F. Mahler, R.L.S. Izzo, P.F.S. Brandt & A.S. Avelar The challenges of performing laboratory controlled cone penetration tests in unsaturated soils M. Pournaghiazar, A.R. Russell & N. Khalili Characterization of unsaturated zone susceptible to landslides in tropical volcanic residual soil with in-situ tests P.P. Rahardjo & R. Karlinasari Analysis of a drawdown test displaying the use of transparent soil in unsaturated flow applications G. Siemens, S. Peters & W.A. Take Calibration of Smart Irrigation Sensor (SIS-UMS) for the blanket layer soil from old railway lines V.N. Trinh, A.-M. Tang, Y.-J. Cui, J.C. Dupla, J. Canou, N. Calon, A. Robinet, L. Lambert & O. Schoen Wave velocity analysis of state parameter changes in multi-phase granular materials F. Wuttke, M. Asslan & T. Schanz A novel non-contacting laser displacement transducer technique for volume change measurements of unsaturated soils X. Zhang, M. Mavroulidou, M.J. Gunn, Z. Cabarkapa & J. Sutton Author index
659
665
671
677
683
689 695
701
707 713
721
727
733
739
745
751
757
VOLUME 2 Modelling
Fundamentals and constitutive modelling
Mechanical model for partially saturated sandy soils: Application to liquefaction H.B. Bian, T. Nishimura & I. Shahrour Saturation induced instability on bonded geomaterials G. Buscarnera & R. Nova On the definition of constitutive models for compacted soft rocks R. Cardoso & E.E. Alonso Effect of anisotropy on the prediction of unsaturated soil response under triaxial and oedometric conditions F. DOnza, D. Gallipoli & S. Wheeler Modelling criteria for a microbiologically stabilised compacted soil in the framework of elastoplasticity G. Della Vecchia, L. Morales, E. Garzn, C. Jommi & E. Romero Some remarks on the hydro-mechanical constitutive modelling of natural and compacted Boom clay G. Della Vecchia, C. Jommi, A. Lima & E. Romero Soil water characteristic curves and adsorption potential S. Frydman & R. Baker Numerical simulations of the thermomechanical behavior of clays on volumetric and deviatoric paths M.S. Ghembaza, Z. Bellia & H. Trouzine Parameter calibration for hydro-mechanical modelling using numerical simulations of test results R. Gmez, E. Romero, A. Lloret, J. Suriol & C. Jommi Evaluation of a constitutive model for unsaturated soils: Stress variables and numerical implementation N.A. Gonzlez & A. Gens Prediction of the behaviour of unsaturated soils using evolutionary polynomial regression: An incremental approach A.A. Javadi, A. Johari, A. Ahangar-Asr, A. Faramarzi & D.G. Toll A fully coupled hysteretic finite element model for hydro-mechanical analysis of unsaturated soils N. Khalili, S. Zargarbashi & M.A. Habte A simple elasto-plastic model for unsaturated soils and interpretations of collapse and compaction behaviours M. Kikumoto, H. Kyokawa, T. Nakai & H.M. Shahin An introduction of jumped kinematic hardening rule to elastoplastic model for unsaturated geo-materials Y. Kohgo An effective stress based model for the dependency of a water retention curve on void ratio D. Man Rotational multiple yield surface framework and the prediction of stress-strain response for saturated and unsaturated soils M.J. Md Noor & I.B. Mohamed Jais 763 771 779
787
795
803 811
817
823
829
837
843
849
857 863
869
xi
Laboratory modelling: Settlement due to groundwater fluctuation in partially saturated soil M.J. Md Noor, I.B. Mohamed Jais & J.D. Nyuin Hydraulic hysteresis coupled to volume change effects in unsaturated soil behavior M. Morvan, H. Wong & D. Branque Mapping soil-water profile utilizing non-linear neural network based model M. Mukhlisin, A. El-Shafie & M. Raihan Taha A fundamental interpretation of the critical state strength of unsaturated soils E.J. Murray & V. Sivakumar Determination of the effective stress parameter: A pore network study E. Nikooee, G. Habibagahi & A. Ghahramani Assessment of volumetric collapse using probabilistic approach I. Otlvaro Calle, M.P. Cordo-Neto & G.M. Medero Is the degree of saturation a good candidate for Bishops parameter? J.-M. Pereira, O. Coussy, E.E. Alonso, J. Vaunat & S. Olivella On the time-dependent behaviour of unsaturated geomaterials J.-M. Pereira & V. De Gennaro A porous model based on porosimetry to simulate retention curves E. Rojas, G. Gallegos & J. Leal Modeling the tensile strength of soils E. Rojas, D. Hurtado, M.L. Prez-Rea & A. Zepeda Modelling the unsaturated behaviour of structured soils M. Rouainia & J.-M. Pereira Fractals, double porosity and the soil-water characteristic curve A.R. Russell Constitutive model developments for compacted unsaturated fine grained soils based on porosimetry R. Salgueiro, S. Olivella & J. Suriol Hydro-mechanical coupling for unsaturated soils D. Sheng & A.N. Zhou Simulation of static compaction with unsaturated soil/water coupled F.E. analysis M. Shibata, K. Kawai, S. Kanazawa, A. Iizuka, S. Tachibana, S. Ohno & M. Honda Elastoplastic modelling of hydraulic and mechanical behaviour of expansive soils D.A. Sun & W.J. Sun A constitutive model for cemented unsaturated soils and weak rocks R. Tamagnini, M. Mavroulidou & M.J. Gunn Constitutive model of unsaturated structured soils under cyclic loading C. Yang, M.S. Huang & Y.-J. Cui Hydraulic conductivity function for unsaturated cracked soil L.M. Zhang, J.H. Li & D.G. Fredlund Parameter identification in BBM using a parallel asynchronous differential evolution algorithm Y. Zhang, X.-T. Feng, D. Gallipoli & C. Augarde Use of results from undrained tests for the constitutive modeling of unsaturated soils X. Zhang
875 881 887 895 901 907 913 921 927 933 939 945
951 957 965 973 979 987 993
1001
1009
xii
Analytical and numerical modelling

A driver for the integration of coupled hydro-mechanical constitutive laws for unsaturated soils F. Cattaneo, G. Della Vecchia & C. Jommi Numerical simulation of volume change of a swelling soil in the northeast of Brazil L.M. Costa, S.R.M. Ferreira, L. do N. Guimares & I.D.S. Pontes Filho A new analytical solution for one dimensional transient flow in unsaturated soils H. Farshbab Aghjani & A. Soroush Effect of mesh resolution on long-term water balance calculations M.D. Fredlund, R. Thode & R.G. Nelson Fully coupled analysis of unsaturated porous media using a meshfree method A. Khoshghalb & N. Khalili Two-dimensional fundamental solution for dynamic behaviour analysis of unsaturated soils P. Maghoul, B. Gatmiri & D. Duhamel Drying shrinkage and cracking initiation in clays: Main modeling concepts and application to the ventilation of a tunnel H. Peron, J. Eichenberger & L. Laloui Behaviour of an unsaturated soil column under loading and watering F. Salehnia & B. Gatmiri Comparison of the nonlinear elastic and elastoplastic models in THM response of unsaturated soils F. Salehnia & B. Gatmiri Explicit stress integration with reduced drift for Barcelona Basic Model W.T. Soowski, D. Sheng & S.W. Sloan Active earth force in cohesionless unsaturated soils using bound theorems of plasticity S.A. Stanier & A. Tarantino Implicit integration of a constitutive model for partly saturated structured porous materials R. Tamagnini, M. Mavroulidou & M.J. Gunn Numerical simulation of an unsaturated slope failure during an earthquake R. Uzuoka, K. Maruyama, T. Mori, M. Kazama, T. Unno & N. Sento 1017 1025 1029 1035 1041
1049
1055 1061
1067 1075 1081 1087 1093
Engineering applications and case histories

Soil-atmosphere interaction
Influences of the soil-atmosphere interface interactions on the soil water content and temperature profiles K.V. Bicalho, G.P.D. Vivacqua & Y.-J. Cui An experimental root barrier to protect the foundations of a building from the desiccating effects of tree roots G.E. Blight Soil heat and evaporation from soil G.E. Blight A prototype soil column to calibrate numerical models accounting for soilatmosphere interaction M. Caruso & C. Jommi Use of satellite observations for estimating the degree of saturation of superficial soils L. Montrasio, R. Valentino & C. Quintavalla 1103
1109 1115
1121 1127
xiii
Study of stratification resistant to salt damage S. Nomura, K. Kawai, S. Kanazawa, A. Iizuka, S. Ohno, P. Thirapong & S. Tachibana An environmental chamber for studying the soil-atmosphere interaction A.-N. Ta, A.-M. Tang, Y.-J. Cui & J. Thiriat
1133 1141
Foundations
Settlement prediction of cyclically loaded footings resting on partially saturated sand by means of a kinematic hardening model A. Becker Performance and design of foundations on unsaturated expansive soil J.T. Bryant & M.A. Haque Identification, characterization and analysis of deep foundation in collapsible soilRegional Prison of Eunpolis, Bahia/Brazil R.Q. Coutinho, B.P.F. Castro & K.C.A. Dourado Pile horizontal bearing capacity in tropical soil under distinct moisture contents M.A. de Almeida, M.G. Miguel & S.H.C. Teixeira The effect of ripping clay soil on swell behaviour S. Fityus, O. Buzzi, M. Holt & T. Gunther Comparison of bearing capacity of unsaturated sand using Cone Penetration Tests (CPT) and Plate Load Tests (PLT) F.M.O. Mohamed, S.K. Vanapalli & M. Saatcioglu Water content profiles for design of foundations on expansive soils D.D. Overton, K.C. Chao & J.D. Nelson Bearing capacity of a strip footing on an unsaturated sand T. Schanz, Y. Lins & S.K. Vanapalli A simple technique for estimating the 1-D heave in expansive soils S.K. Vanapalli, L. Lu & W.T. Oh A technique for estimating the shaft resistance of test piles in unsaturated soils S.K. Vanapalli, K.D. Eigenbrod, Z.N. Taylan, C. Catana, W.T. Oh & E. Garven Bearing capacity of an unsaturated sand from model footing tests S.K. Vanapalli, R. Sun & X. Li Interpretation of the bearing capacity of unsaturated soils extending the effective and the total stress approaches S.K. Vanapalli & W.T. Oh Three dimensional numerical simulation of a residential building on expansive soil subject to a leaking underground X. Zhang 1149 1155
1161 1169 1175
1183 1189 1195 1201 1209
1217
1223
1229
Slopes
Stochastic modelling of unsaturated slope stability P. Arnold & M.A. Hicks Influence of seasonable water pressure variations on unsaturated slopes stability F.A.J. Franch & M.M. Futai Shear and soil-water retention behaviour of a variably saturated residual soil and its implication on slope stability A. Jotisankasa & J. Tapparnich 1237
1243
1249
xiv
Back-analysis of infiltration characteristic of a saprolitic hillslope by considering permeability heterogeneity A.K. Leung & C.W.W. Ng The significance of curved shear strength envelope in modelling rainfall induced shallow failure of slopes M.J. Md Noor & B.A. Hadi The influence on the slope stability of the initial degree of saturation of a fill ground T. Mori, R. Uzuoka & M. Kazama Test site experience on mechanisms triggering mudflows in unsaturated pyroclastic soils in southern Italy M. Pirone, R. Papa & M.V. Nicotera Performance of an instrumented slope under a capillary barrier system H. Rahardjo, C.J. Hua, E.C. Leong & V.A. Santoso Matric suction in unsaturated pyroclastic slopes A. Scotto di Santolo & A. Evangelista Analysis of influencing factors in strengthening expansive soil slope with the coverage of geo-membrane J.P. Yuan, J.H. Wu & Z.Z. Yin
1255
1261
1267
1273 1279 1285
1291
Embankments
Investigation of the water flow in an embankment built of organic soil: Experimental results and large scale field test E. Birle, M. Boso & D. Heyer Suction changes during the construction of an embankment from A10 Motorway in Portugal R. Cardoso, E. Maranha das Neves & P. Almeida Santos Influences of spatially varying porosity on unsaturated flow through earth structures T.M.H. Le, D. Gallipoli, S. Wheeler & M. Snchez Improvement and controlling deformation of the expansive soil ground L. Miao, F. Wang & Y. Cui Failure of peat dykes by drought D. Nterekas, J.P. Oostveen, F.A.J.M. Mathijssen, R.B.J. Brinkgreve, M.A. Hicks & F. Molenkamp Numerical analysis of the behaviour of embankments constructed with expansive soils J.I. Ortega, A. Lloret, S. Olivella & E. Romero Banking simulation with a constitutive model for unsaturated soil M. Tanaka, K. Kawai, S. Kanazawa, A. Iizuka, S. Tachibana, S. Ohno & T. Takeyama 1299 1305 1313 1321 1325 1331 1337
Pavements
Water migration in pavement subgrade E.J. Nelson & D.D. Overton 1345
Geoenvironmental applications
Gas migration in sand/bentonite mixtures through preferential paths D. Arnedo, E.E. Alonso, S. Olivella & E. Romero Geotechnical issues in the design and construction of PMOBs A.R. Cabral, M. Ltourneau, J.S. McCartney, J. Parks, E. Yanful & Q. Song Coupled solute mass transport and evaporation in saline soil A.S. Dunmola & P.H. Simms 1353 1361 1369
xv
Non-isothermal infiltration in plastic bentonite B. Franois & L. Laloui Numerical modelling of the resaturation of swelling clay with gas injection P. Gerard, J.-P. Radu, J. Talandier, R. de La Vaissire, R. Charlier & F. Collin THMC analysis of saturation and heating processes of an expansive clay barrier in radioactive waste isolation L. do N. Guimares, A. Gens, M. Snchez & S. Olivella Laboratory studies of the addition of bentonite to landfill cover soil R.L.S. Izzo, C.F. Mahler & K. Huse Thermal-hydro-mechanical behavior of unsaturated bentonite blocks: Experimental approaches J.O. Lee, W.J. Cho, S.K. Kwon & J.W. Choi Water infiltration and methane emission through three different cover layers of an experimental Municipal Waste Landfill at Muribeca, Recife, Pernambuco, Brazil R.L. Lopes, M.C.M. Alves & J.F.T. Juc Evaluation of landfill gas emission in experimental cover layers in Brazil F.J. Maciel, R.L. Lopes & J.F.T. Juc Behaviour of wet fill using residual soil F.A.M. Marinho & F. Massad Unsaturated behaviour of mine tailings in low precipitation areas L.A. Oldecop, L. Garino, J.J. Muoz, R. Rodriguez & C. Garca Evaluation of a capillary barrier at the Rocky Mountain Arsenal L.O. Williams, D.L. Hoyt, G.A. Hargreaves, S.F. Dwyer & J.G. Zornberg Numerical study on rainfall infiltration into landfill earthen final cover T.L.T. Zhan, G.W. Jia, Y.M. Chen & D.G. Fredlund
1375 1383
1389 1395
1401
1407 1413 1419 1425 1431 1437
New areas
LNAPL retention in partially saturated silty sand M. Alferi, E. Romero, A. Dominijanni & M. Manassero Water retention curve for evolving marl under suction cycles R. Cardoso, G. Della Vecchia, C. Jommi & E. Romero Effects of crude oil pollution on the shear strength behaviour of an unsaturated porous soil medium M. Mohammadi, S.S. Yasrebi & A. Mirzaii Effects of hydraulic cycling on the stiffness response of a rigid clay J.A. Pineda, G. Mitaritonna, E. Romero & M. Arroyo Effect of drying and wetting cycles on the shear strength of argillaceous rocks R.D. Tovar & J.E. Colmenares Pore geometry induced suction and the slaking of shales L.E. Vallejo & A. Stewart-Murphy Influence of suction on the mechanical behaviour of rock joints under shear loading M.T. Zandarin, E. Alonso & S. Olivella Author index 1445 1451
1459 1465 1471 1477 1483
1489
xvi
Preface
The development of Unsaturated Soil Mechanics in recent decades has been remarkable and it has resulted in momentous advances in fundamental knowledge, testing methods, computational procedures, prediction methodologies and geotechnical practice. The advances have spanned the full spectrum of theory and practice from formulating new conceptual bases for understanding through the introduction of new or improved testing and monitoring techniques to the development of more rational approaches for solving engineering problems. In addition, unsaturated materials exhibiting complex behaviour such as residual soils, swelling soils, compacted soils, collapsing soils, tropical soils and solid wastes have been integrated in a common understanding of shared behaviour features. It is also noteworthy that unsaturated soil mechanics has proved surprisingly fruitful in expanding to other neighbouring areas such as swelling rocks, rockfill mechanics, CO2 sequestration, methane hydrates and freezing soils. Equally important, unsaturated soil mechanics uses concepts and methods fully compatible with those used in saturated soils thus providing a unified approach capable of encompassing the full range of soil mechanics. As a consequence, geotechnical engineering involving unsaturated soils can be now approached from a more rational and systematic perspective leading towards an improved and more effective practice. A series of International Conferences on Unsaturated Soils, held at approximately regular intervals, have provided essential venues to report and discuss advances in the theory and practice in this area of soil mechanics and geotechnical engineering. The first International Conference took place in Paris in 1995, followed by the second in Beijing, China, in 1998, the third in Recife, Brazil, in 2002 and the fourth in Carefree, Arizona (US) in 2006. In turn, they were preceded by a successful series of Conferences on expansive soils started at Texas A&M University in 1965. More recently, Regional Conferences (European, Asian-Pacific) have also been organized attesting to the continuing growth of the field. The present Proceedings contain the papers submitted to the 5th International Conference, held in Barcelona, that have been accepted for publication after peer review. They contain significant advances in the areas of Unsaturated soil behaviour, Testing techniques, Constitutive and Numerical Modelling and Applications. The areas of application include soil-atmosphere interaction, foundations, slopes, embankments, pavements, geoenviromental problems and emerging topics. They are complemented by three Keynote Lectures and three General Reports, written by eminent researchers in the field, covering the general issues of modelling, testing and applications. This collection of papers should constitute an essential reference for academics, researchers and practitioners. The Conference has been organized under the auspices of the International Society for Soil Mechanics and Geotechnical Engineering through Technical Committee TC-106 (formerly TC6). We thank the Spanish Society for Soil Mechanics and Geotechnical Engineering for their sponsorship. We would also like to acknowledge the contribution of the International Advisory Committee to the success of the Conference. The Conference was organized by the Department of Geotechnical Engineering and Geosciences of the Technical University of Catalonia in Barcelona. Their support throughout the preparation of the Conference is gratefully recognized. Finally, thanks are due to all the authors who have submitted papers to the Conference, to the reviewers who undertook an essential but anonymous task and to the authors of Keynote Lectures and General Reports for their special contributions to the Conference. In spite of the impressive developments in the field of Unsaturated Soil Mechanics in recent decades, the subject is advancing at a very fast rate resulting not only in significant improvements in basic understanding, modelling and testing capabilities, and geotechnical practice but also in a vigorous expansion into other related areas. These two volumes should be considered as a comprehensive record of the current state-of-the art of Unsaturated Soil Mechanics and a sound basis for further progress in the future. Eduardo Alonso Antonio Gens
xvii
Organization
Conference Chairmen E. Alonso (UPC) A. Gens (UPC) Secretary M. Arroyo (UPC) Organizing Committee (UPC) A. Lloret A. Ledesma S. Olivella E. Romero J. Vaunat A. Josa J. Gili P. Prat J. Suriol International Advisory Committee and TC106 (*) Members S.L. Barbour (Canada) P. Bil (Portugal) H. Bilsel (Turkey) J. Blatz (Canada) G.F. Blight (S. Africa) T. Campos (Brazil) A. Cara (Albania) A. Carrillo (Peru) R. Charlier (Belgium) J. Colmenares (Colombia) Y.-J. Cui (France) E. Dapena (Spain) V. de Gennaro (France) J.L. de Justo (Spain) P. Delage (France) S.G. Fityus (Australia) D. Fredlund (Canada) S. Frydman (Israel) D. Gallipoli (UK) K. Garbulewski (Poland) B. Gatmiri (Iran) L. Guimares (Brazil) G. Habibagahi (Iran) A. Harraz (Egypt) S. Houston (USA)
xix

L. Hoyos (USA) E.A. Isahanov (Kazakhstan) E. Imre (Hungary) A. Izuka (Japan) C. Jommi (Italy) N. Khalili (Australia) Y. Kohgo (Japan) L. Laloui (Switzerland) S. Leroueil (Canada) R. Lytton (USA) S.M. Mamoune (Algeria) C. Mancuso (Italy) E. Maranha das Neves (Portugal) F. Marinho (Brazil) D. Masin (Czech & Slovakia) F. Masrouri (France) G.A. Miller (USA) F. Molenkamp (Netherlands) C. Ng (Hong Kong) L. Oldecop (Argentina) S.W. Park (Korea) H. Rahardjo (Singapore) S. Rao (India) E. Rojas (Mexico) M. Snchez (UK) T. Schanz (Germany) D. Sheng (Australia) D. Smith (Australia) A. Soriano (Spain) S. Springman (Switzerland) A. Tarantino (Italy) D. Toll (UK) H. Toyota (Japan) S.K. Vanapalli (Canada) O. Vilar (Brazil) S. Wheeler (UK) V. Whenham (Belgium) L. Zhan (China) J. Zornberg (USA) D. Znidarcic (Croatia)
(*) TC106 COMMITTEE ON UNSATURATED SOILS The TC106 Committee of the ISSMGE plays a leading role in the promotion of the series of international UNSAT Conferences. It participates actively in the technical aspects of the conferences and looks after the quality and general interest of the Conference. The Committee promotes cooperation and exchange of knowledge in unsaturated soils including, compacted, expansive, collapsible, arid, and residual soils.
xx
Keynote lectures
Design of stiffened slabs-on-grade on shrink-swell soils

J.-L. Briaud
Zachry Department of Civil Engineering, Texas A&M University, Texas, USA
R. Abdelmalak X. Zhang
Geotechnical & Heavy Civil Engineering Department, Dar Al-Handasah., Mohandessin, Giza, Cairo, Egypt Department of Civil and Environment Engineering, University of Alaska, Fairbanks, Alaska, USA
ABSTRACT: Stiffened slabs-on-grade are one of the most efficient and inexpensive foundation solutions for light structures on shrink-swell soils. After removing the top soil, the stiffening reinforced concrete beams, say 1 m deep and 0.3 m wide, are formed in the natural soil and placed every 3 m in both directions. The slab is typically 0.1 m thick. Such stiffened slabs are often called waffle slabs because of the geometric analogy with a waffle. They are commonly used in the USA for the foundation of houses or any light weight structure on shrink-swell soils and cost about 100$/m2 (2010). This paper presents a simple design procedure and associated charts for calculating the depth of the beams required to limit the differential movement of the foundation due to bending to an acceptable amount. This bending of the foundation is due to the shrinking and swelling of the soil during the dry and wet seasons under the edges of the structure. The procedure consists of using the change in suction or the change in water content selected for design purposes in the region, the anticipate depth of influence of these seasonal changes, the soil properties, and the beam depth, to calculate the distortion of the slab under load. If the distortion is too large for the type of structure considered, the beam depth is increased until the acceptable value is reached.
INTRODUCTION
Stiffened slabs-on-grade (Fig. 1) are one of the most efficient and inexpensive foundation solutions for light structures on shrink-swell soils. After removing the top soil, the stiffening reinforced concrete beams, say 1 m deep and 0.3 m wide, are formed in the natural soil and placed every 3 m in both directions. The slab is typically 0.1 m thick. Such stiffened slabs are often called waffle slabs because of the geometric analogy with a waffle. They are commonly used in the USA for the foundation of houses or any light weight structure on shrink-swell soils and cost about 100$/m2 (2010). This paper presents a simple design procedure and associated charts for calculating the depth of the beams required to limit the differential movement of the foundation due to bending to an acceptable amount. Some of the other foundation solutions used for buildings on shrink-swell soils are thin post tension slab on grade, slab on grade and on piles, and elevated structural slab (Fig. 1). The thin post tension slab on grade (say 0.1 m thick) is a very good solution for flat surfaces which are not carrying any structures such as tennis courts. The thin post tension slab on grade is not a good solution for the foundation of light
and relatively rigid structures because, while post tensioning increases the maximum bending moment that the slab can resist compared to the same reinforced concrete slab and provides a minor improvement in stiffness, it is generally too flexible and leads to differential movements of the foundation incompatible with what a rigid structure can tolerate. This can lead to cracking in the walls of the structure. Of course, thick post-tensioned slabs can work well and post tensioning a reinforced concrete stiffened slab
Stiffened Slab on Grade Stiffened Slab on Grade and on Piers
Elevated Structural Slab on Piers

air gap
Thin Post Tensioned Slab
Figure 1. Some foundations used for light buildings on shrink-swell soils.
on grade increases the maximum bending moment that the slab can resist. The benefit associated with the added cost of post-tensioning a reinforced concrete stiffened slab is not clear. The slab on grade and on piles is not a good solution either as it anchors the slab on grade while the soil may want to swell. In this case the soil pushes up on the slab and the piles hold it down at the connections between the pile and the slab. Under these conditions, the slab is likely to break if swelling is excessive. If the soil shrinks a gap can appear below the slab on grade which loses its support. Since such slabs are not designed to take the load in free span they can fail under these conditions as well. The elevated structural slab on piles is a very good solution but can be expensive (about 200$/m2 in 2010), and represents an unnecessary expense for light structures. It is the solution of choice for more expensive structures. In the case of the elevated structural slab on piles, the soil can move up and down without impacting the structure since a gap of sufficient magnitude (sometimes 0.3 m or more) exists under the beams and the slab. The load is taken up by the piles which must be designed for the tension generated by the swelling of the soil which is typically more severe than the shrinking design case. Of course no matter how well designed and constructed the foundation is for light structures on shrink-swell soils, some poor practices by the owner can be very detrimental to the structure including trees too close to the structure and poor drainage. Proper owner maintenance is important and requires some education of the owner. 2 DEVELOPMENT OF THE DESIGN APPROACH
2. Obtain a realistic shape of the mound that would be formed if the soil were to be partially covered by a perfectly flexible cover and were to shrink or swell in the free field. 3. Simulate the effect of the weather on the soil in 5 cities over a 20 year period to obtain reasonable estimates of the maximum amplitude of variation of the suction over a 20 year period. 4. Place the foundation on top of that mound, solve the problem of soil-structure interaction and find out the maximum bending moment and deflection of the foundation slab. Perform an extensive parametric simulation study to find out what parameters are most important for the bending moment and the deflection. 5. Develop simple design charts based on the results of tasks 1 through 4 to obtain the beam depth and the beam spacing without having to use a computer. Task 1 consisted of performing numerical simulations to develop the percent diffusion vs. time factor curves for the free shrink test much like the percent consolidation vs. time factor curves for the consolidation theory. These curves were used in conjunction with the free shrink test to obtain the coefficient of diffusivity of the intact soil. Then other numerical simulations were used to develop a correction factor for including the influence of cracks on the intact value of the diffusion coefficient. Task 2 consisted of using an advanced form of Mitchells theory (1979) to develop a more realistic solution for the equation describing the shape of the mound created when a flexible membrane covers partially a shrinking or swelling soil. Task 3 required special numerical simulations using the FAO 56-PM method with weather input and soil input to obtain the maximum change in suction for several cities in the USA. These cities were selected because they have the most significant problems with shrink swell soils: San Antonio, Austin, Dallas, Houston, and Denver. College Station was added because it is the home of Texas A&M University. The weather input over the 20 year period considered included the rain fall, the air temperature, the wind speed, the relative humidity, and the solar radiation. An average vegetation cover was selected and the soil diffusion was varied over a reasonable range. Task 4 consisted of running simulations with the foundation and its loading on top of the deformed mound as many have done before. The maximum bending moment in the stiffened slab and the deflections were obtain in a series of cases where many parameters were varied. This thorough sensitivity analysis gave a very clear indication of what parameters were most important and
The primary design issues for stiffened slabs on grade are the sizing of the beams that stiffened the slab and the spacing of the beams. The beams are typically from 0.6 to 1.2 m deep, 0.15 to 0.3 m wide, and the spacing varies from 3 to 6 m. The slab itself is usually set at a thickness of 0.1 m. Any quality design for such a foundation must include the parameters related to the soil, the weather, the foundation, and the super-structure. The goal of such a design is to ensure that the stiffness of the foundation is such that the movements due to swelling and shrinking of the soil under a given weather condition do not distort the super-structure excessively. The development of the design approach to size the beams and their spacing consisted of a number of research tasks which took place over the last 10 years. These tasks were. 1. Develop a simple way to obtain the coefficient of unsaturated diffusivity of the intact soil (lab) and the cracked soil (field).
what parameters could be neglected within the precision sought. Task 5 took advantage of all the results obtained to select the major factors and organize design charts which gave the beam depth for a given weather-soil index. This paper covers Task 4 and Task 5 of the research effort and details the steps of the proposed design procedure with the charts. The results of the other tasks can be found in the work of Abdelmalak (2007) with further background found in Zhang (2004). Many design methods have been established for the design of stiffened slabs on shrink-swell soils; they include the Building Research Advisory Board method (BRAB 1968), the Wire Reinforcement Institute method (WRI 1981), the Australian Standards method (AS 2870 1990, 1996), and the Post-Tensioning Institute method (PTI 1996, 2004). These design methods handle this problem by implementing different moisture diffusion, soil-weather interaction, and soil-structure interaction models. They all have advantages and drawbacks which lead the authors to develop this new method. 3 NUMERICAL SIMULATION OF THE SOIL-SLAB INTERACTION
Figure 2. Geometry and boundary conditions for an edge lift case.
Finite element simulations for plates resting on a semi-infinite elastic continuum have been used to solve the problem of a slab resting on a soil mound since the 1970s. The starting point for all these simulations is to assume a certain soil mound shape. The mound shape equation developed in Task 2 was used to obtain the initial mound shape. Then, a 2D finite element simulation was carried out for a flat foundation slab centered on the mound. The output of this simulation was the deflected shape of the slab, the bending moment diagram across the slab and the shear force diagram across the slab. Abdelmalak (2007) used the finite element package, ABAQUS/STANDARD, to simulate the problem. The mesh is shown in Figure 2 for the case where the soil is swelling around the edges of the slab (edge lift) as would happen in the winter time. Figure 3 shows the mesh in the case where the soil is shrinking around the edges of the slab (edge drop) as would happen in the summer time. A two dimensional plane strain condition was simulated and boundary conditions are shown in Figures 2 and 3. If L is the width of the slab, the mesh was 3 L wide and 1.5 L deep and half the problem was simulated because of symmetry. The soil and the concrete were simulated as elastic materials and a load was imposed on the slab in addition to its weight. As a simulation example, consider a plane strain problem for a foundation slab on an edge
Figure 3. Geometry and boundary conditions for an edge drop case.
drop mound: The foundation slab has a width of L 16 m, an equivalent thickness of 0.38 m, a modulus Econc 20000 MPa, and it is loaded with 7.5 kPa. The foundation rests on an edge drop mound with a shape obtained from the equation derived in Task 2. The input parameters for the equation were a depth of active zone equal to 3.5 m, a free field imposed suction change of 1.82 pF at the boundary (or a suction change such that the ratio between the suction after and before is 66), and a soil with a field coefficient of diffusivity field 0.02 m2/day. The modulus of the soil for the finite element simulation was taken as Esoil 60 MPa.
0.09 0.08 0.07
Soil mound and foundation elevations
0.06 0.05 0.04 0.03 0.02 0.01 0.00 -0.01 0 1 2 3 4 5 6 7 8
Initial Mound Elev. Final Mound Elev. Final Found Elev
x-coordinate (m)
Figure 4. Initial and final soil mound profiles and final foundation slab profile.
Bending moments and shearing forces
140 120
Shearing Force, V Bending Moment, M
100 80 60 40 20 0 -20 -40 -60 0 1 2 3 4 5
x-coordinate (m)
Figure 5. Bending moments and shearing forces results.
The initial and final soil mound profiles and the final slab profile for the example cited above are presented in Figure 4. Due to the slab weight and the load, the surface of the soil mound settles. Because the slab is stiff, a gap exists near the edge between the soil surface and the underside of the slab. The length of unsupported slab or gap length Lgap is about 3.5 m in this example. This overhanging of the slab creates a significant bending moment and associated deflection. The bending moment and shear force diagrams across the slab are shown in Figure 5. As expected, the maximum bending moment coincides with the point of zeroshear. Note that the distance from the edge of the slab to the point of separation between soil and slab, Lgap, is smaller than the distance between the edge of the slab and the point of the maximum bending moment. In fact at the separation point the bending moment is only about half the maximum value. The reason is that the overhanging slab does not behave as a pure cantilever. 4 INFLUENCE FACTORS AND RANGE OF VALUES SIMULATED
Many factors are involved in the design of a slab on grade of shrink-swell soils. The weather factors
include the change of soil suction at the ground surface in the free field, U0, and the change of soil suction at the ground surface under the edge of the foundation, Uedge. The soil factors include the soil modulus, Esoil, the shrink-swell index, Iss (difference between the swell limit and the shrink limit), and the field coefficient of unsaturated diffusivity, field. The slab factors include the slab length, L, the slab bending stiffness EI, and the loading q. Obviously, in order to cover all the combinations of parameters, a very large number of simulation cases would have to be performed to develop a design procedure that addresses all these parameter effects. As a first step, a sensitivity study was undertaken to examine the influence of these parameters on the design of the foundation slab. This was done in hope to reduce the required number of simulations. The range of the parameters included in the sensitivity study was chosen to realistically cover the anticipated variations. The change of soil suction at the ground surface in the free field, U0, recommended by AS 2870 (1996) varies from 1.2 to 1.5 pF; this recommendation is based on databases of field measurements in Australia. To be safe the chosen range of U0 was expanded to be from 1 to 1.6 pF and five values were examined within that range 1, 1.15, 1.3, 1.45, 1.6 pF. The change of soil suction at the ground surface under the edge of the foundation, Uedge was shown to be 0.5 times U0 (Abdelmalak 2007). Five depths of active moisture zones, H, were examined: 1.5, 2.5, 3.5, 4.5, and 5.5 m. The periodic time for the cyclic surface suction change due to weather variations was assumed to be 365 days. Three soil moduli, Esoil were examined for the edge drop case (mounds resulting from soil shrinkage): 20, 60, and 100 MPa. These soil moduli are relatively high but were so chosen because drier soils tend to be stronger. For the edge lift cases (mounds resulting from soil swelling) the modulus values chosen for the sensitivity study were 6, 10, 15, 20 MPa. These moduli are smaller than the ones for the edge drop case because wet soils tend to be softer. The mound shape elevations for the edge lift case were scaled down by one half as will be explained later. Five soils were chosen with the following shrink-swell indices, Iss: 75, 60, 45, 30, 15%, and the corresponding coefficients of saturated permeability were assumed to be 5 E-10, 7.5 E-10, 1 E-09, 2.5 E-09, and 5 E-09 m/sec respectively. This range of coefficients of saturated permeability was chosen to match the suggested values in Casagrande chart (Holtz & Kovacs 1981). The slope of the soil water characteristic curve (SWCC expressed as gravimetric water content versus suction), Cw was calculated using the shrink-swell index Iss and the empirical relationship Cw 0.5 Iss (Abdelmalak 2007). The field coefficient of
V (kN) or M (kN . m)
y-coordinate (m)
Leqv Leqv
Figure 6. Tributary area used to calculate the line load on the beam.
unsaturated diffusivity, field, were chosen as 0.00724, 0.01244, 0.02042, 0.07110, and 0.26544 m2/day. The slab parameters were varied as follows: slab length, L, 4, 6, 8, 10, 12 m, beam depths 0.3, 0.6, 0.9, 1.2, and 1.5 m, beam width 0.3 m, and beam spacing 4 m. By equating the stiffness of these beam-stiffened slabs to the stiffness of a flat slab, a stiffnessbased equivalent thickness can be calculated. The equivalent thicknesses of these beam stiffened flat slabs were, 0.127, 0.253, 0.3795, 0.506, and 0.633 m respectively. The load on the slab is given in terms of how much line load the beam has to sustain. This line load applied to the beam is obtained by multiplying the distributed pressure applied on the tributary area of the slab by the beam spacing (Fig. 6). The line loads imposed on the beams including dead weight were 2, 2.75, 3.5, 4.25, and 5 kN/m. The slab modulus of elasticity was 20000 MPa (concrete). 5 RESULTS OF THE SENSITIVITY STUDY
Figure 7. A sketch of a foundation slab on grade on a curved mound. Table 1. Parameters for the reference case.
wimposed Parameter Iss(%) H(m) U0(pF) D(m) L(m) (kPa) Reference case 45 3.5 1.3 0.9 8 3.5
1.4
Influence of soil shrink-swell potential on Leqv
Normalized Leqv (Leqv/Leqv (reference case))
1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5
Edge drop case Edge lift case

2
Normalized Lss (Lss/Lss (reference case))
Before presenting the results, it is helpful to give the notation and definitions of all parameter (Fig. 7). Note that the parameters and their definitions are also listed in the nomenclature at the end of the paper. The reference case which was chosen corresponded to the median values of all the parameter (Table 1). The influence of each parameter variation will be compared to the reference case. The equivalent cantilever length, Leqv, was chosen as the most important parameter in the design as it directly impacts the maximum bending moment and the maximum deflection. The influence of each parameter on Leqv will be discussed next. 5.1 Influence of the soil shrink-swell potential
Figure 8. Influence of the soil shrink-swell potential on the equivalent cantilever length.
An increase in the shrink-swell index creates a non-linear monotonic increase in the equivalent cantilever length. The reason is that increasing the shrink-swell index increases the ym values which in turn increases the soil mound distortion and consequently increases the foundation slab distortion. The average slope of the normalized equivalent cantilever length vs. the normalized shrink-swell index curve (Fig. 8) was 0.342 for the edge drop case and 0.369 for the edge lift case. 5.2 Influence of the depth of active moisture zone
Figure 8 shows the relationship between the soil shrink-swell potential, represented by the shrinkswell index, and the equivalent cantilever length.
Figure 9 shows the relationship between the depth of the active moisture zone and the equivalent cantilever length. An increase in the depth
1.4 1.2
Influence of depth of active moisture zone on Leqv

1.2 1 0.8 0.6 0.4 0.2 0 2 0
Influence of soil surface suction change on Leqv
1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5

0.2 0.4 0.6 0.8 1 1.2 1.4
Normalized H (H /H (reference case))
Normalized U0 (U0 /U0 (reference case))
Figure 9. Influence of the depth of active moisture zone on the equivalent cantilever length.
Figure 10. Influence of soil surface suction change on the equivalent cantilever length.
Influence of slab beam depth on Leqv
of the active moisture zone creates a non-linear monotonic increase in the equivalent cantilever length. The reason is that increasing the depth of the active moisture zone increases the ym value which in turn increases the soil mound distortion and consequently the slab distortion. However, the slope of the normalized equivalent cantilever length vs. the normalized depth of the active moisture zone decreases at depth of the active moisture zone increases. The average slope of the normalized equivalent cantilever length vs. the normalized depth of the active moisture zone curve was 0.493 for edge drop case and 0.483 edge lift case. This indicate that the equivalent cantilever length seems to be somewhat more sensitive to the depth of the active moisture zone than to the soil shrink-swell potential; however the depth of active moisture zone is also a function of soil shrink-swell potential. 5.3 Influence of the suction change at the ground surface
1.6 1.4
1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2
Normalized D (D /D (reference case))
Figure 11. Influence of slab stiffness on the equivalent cantilever length.
Figure 10 shows the relationship between the suction change at the ground surface vs. the resulting equivalent cantilever length. An increase in the soil surface suction change increases, almost linearly, the equivalent cantilever length. The average slope of the normalized equivalent cantilever length vs. the normalized soil surface suction change curve is 0.372 for edge drop case and 0.481 edge lift case. However, the range of the soil surface suction change was smaller than that of either the normalized shrink-swell index or the normalized of depth of active moisture zone. 5.4 Influence of the slab stiffness
relationship is close to linear up to the average beam depth in this sensitivity study and then the curve becomes non linear. Note that the equivalent cantilever length cannot be larger than the half length of the slab which represents an upper bound for Leqv. The average slope of the normalized equivalent cantilever length vs. the normalized beam depth curve was 0.628 for the edge drop case and 0.792 for the edge lift case. Considering the linear portion only, the slopes were 0.904 for edge drop case and 0.98 for the edge lift case. 5.5 Influence of the slab length
Figure 11 shows the relationship between the slab stiffness, represented by the depth of the stiffening beams, and the resulting equivalent cantilever length. An increase in beam depth significantly increases the equivalent cantilever length. The
Figure 12 shows the relationship between the slab length and the resulting equivalent cantilever length. The equivalent cantilever length increases almost linearly with the slab length until it reaches a maximum value and then a constant value. The slab length affects two phenomena, moisture diffusion and slab curvature. For small 0.5 L/H ratios, the difference in suction between the center of the slab and the edge of the slab is small and the curvature of the soil mound decreases. Consequently, the slab curvature decreases, and so does the equivalent cantilever length. On the other hand, increasing the slab length decreases the slab curvature and
therefore the equivalent cantilever length. These two counteracting effects come to a balancing point at which the maximum equivalent cantilever length is reached. Hence, the influence of the slab length on the equivalent cantilever length can be addressed by introducing a reduction factor, Fsl, as shown on Figure 13 which can be viewed as an idealized form of Figure 12. 5.6 Influence of the slab load
The increase in slab load slightly decreases the equivalent cantilever length with a linear trend. Increasing the slab load compresses the soil mound which reduces its curvature thereby decreasing the resulting equivalent cantilever length. The average slopes of the normalized equivalent cantilever length and normalized slab beam depth curve were 0.141 for the edge drop case and 0.232 for the edge lift case. 5.7 Influence of soil modulus of elasticity
Figure 14 shows the relationship between the slab load and the resulting equivalent cantilever length.
Influence of slab length on Leqv
1.2 1 0.8 0.6 0.4 0.2 0 0 0.5
Edge drop case Edge lift case 1 1.5 2 2.5 3 3.5 4
Normalized L (L /2H (reference case))
Figure 12. Influence of slab length on the equivalent cantilever length.

slab length factor
1.2 1 0.8
0.6 0.4 0.2 0 0 0.5 1 1.5 2 2.5 3 3.5 4 Edge drop case Edge lift case
For the edge drop case, the soil is shrinking and the increase in suction leads to a stiff soil. On the other hand in the edge lift case, the soil swells and the decrease in suction leads to a softer soil. As reasonable estimates, values of 20, 60, and 100 MPa were used for the soil modulus, Es, of the stiff soil in the edge drop case and values of 5, 10, and 20 MPa for the softer soil of the edge lift case. Furthermore in the case of the edge lift case it was recognized that the pressure imparted on the soil by the slab an its load would lead to a mound shape exhibiting less curvature than the case of a free swelling soil. Figure 15 shows a typical influence of the vertical pressure on the percent swell of a soil. Note that, in Figure 15, the percent swell of the soil under very low pressure is about 10% and the percent swell under 30 kPa of pressure is 5%. An arbitrary but seemingly reasonable reduction factor of 0.5 was applied on the free swell mound elevations before placing the slab and running the simulation of the edge lift case. Figure 16 shows the relationship between the soil modulus of elasticity and the resulting maximum bending moment. Figure 16 shows that the Mmax increases when the soil becomes stiffer. 5.8 Conclusion of the sensitivity study
Fs1
L /2H
Figure 13. Slab length factor for reducing the equivalent cantilever length.
Influence of slab imposed loads on Leqv
1.2
The sensitivity study showed that the following factors influence the design of slabs on grade on shrink-swell soils. These factors are cited in order of significance starting with the most significant
Pressure-Swelling curve for a remoulded clay (PI = 48%)
350 300
Normalized Leqv (Leqv /Leqv (reference case))
1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 Edge drop case Edge lift case
Pressure (kPa)
250 200 150 100 50 0 0 2 4 6 8 10 12
Normalized wimposed (wimposed/wimposed (reference case))
Soil swell (%)
Figure 14. Influence of slab load on the equivalent cantilever length.
Figure 15. Typical pressure-swelling characteristic of clay (after Mitchell 1979).
Maximum Moment (KN . m/m')
240
Influence of soil modulus of elasticity on Mmax

Edge drop
210
Edge Lift_full mound Edge Lift_ half mound
180
150
120
90 0 20 40 60 80 100 120
E (MPa)
Figure 16. Mmax.
Influence of soil modulus of elasticity on
the slab stiffness deq. The design parameters were the maximum bending moment (or equivalent cantilever length), the shear force, and the slab deflection. It was decided that the charts would have the design parameter on the vertical axis and a combined soil weather index on the horizontal axis. The chart would contain several curve each one referring to a given slab stiffness. So the next step was to develop a combined soil weather index ISW. Since the sensitivity study showed that the equivalent cantilever length increased with ISS, H, and U0, it was decided to define the soil-weather index as ISW ISS H U0 (4)
factor: slab stiffness, depth of active moisture zone, shrink-swell potential, soil surface suction change, slab length, imposed loads, soil stiffness. 6 6.1 NEW DESIGN CHARTS Design parameters
Figure 17 shows the relationship between the soil-weather index and the equivalent cantilever length from the runs performed in the sensitivity study (median slab stiffness) (Abdelmalak 2007). A hyperbolic function, in the form of Equation 5 where L0, a, and b are constants, gave the best fit with a coefficient of determination equal to 0.944. Leqv L0 aI S W , R 2 0.944 1 bI S W (5)
The design parameters necessary to size the beams and their spacing are the maximum bending moment Mmax, the maximum shear force Vmax, and the maximum deflection max of the slab. In the proposed design procedure, these quantities are presented by using the equations for a cantilever beam modified with a factor which makes them applicable to a slab on grade. These modification factors are based on all the numerical simulations. For a true cantilever beam, the equivalent cantilever length Leqv would be the length of the cantilever beam, the maximum shear factor FV would be one, and the maximum deflection factor Fmax would be 8. For the stiffened slab on grade, these factors were obtained from the numerical simulations. Leqv (2Mmax/q)0.5 Fv Vmax qLeqv (1) (2) qL4 eqv (3)
6.3 Parametric analysis Based on the success found in using the combined soil weather index (Fig. 17), a parametric analysis was undertaken to develop the design charts. Seven representative mounds were chosen including mounds with the minimum and maximum soilweather indices to cover the whole possible range and five intermediate mounds. For each mound, five slab stiffnesses were chosen and numerically simulated for both the edge lift case and the edge drop case. Abdelmalak (2007) summarized the input parameters for the numerical simulations that were carried out to construct the design charts. The slab stiffness was represented by the slab equivalent depth, deq. The slab equivalent depth deq represents the thickness of a flat slab which would
F max
max EI
Note that Leqv should be checked against 2H and if Leqv is less than approximately 2H (Fig. 13), Leqv must be reduced according to Figure 13. 6.2 Soil-weather index
The sensitivity study showed that the main parameters to be included in the design procedure should be the shrink-swell index ISS, the depth of the active zone H, the change of soil surface suction U0, and
Figure 17. Relationship between the soil-weather index and the equivalent cantilever length.
10
F max
have the same moment of inertia as the moment of inertia of a stiffened slab with a beam depth equal to D, a beam width equal to b and a beam spacing equal to S. The slab equivalent depth can be calculated by Equation 6: S deq b D
3 3
4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50
F max design chart (Edge drop)

max qL4 eqv F max EI
deq=0.63 m deq=0.51 m deq=0.38 m deq=0.25 m deq=0.13 m
(6)
6.4
Suction based design charts
0.00 0 0.2 0.4 0.6 0.8 1 1.2
Iss H Uedge (m)
The large number of cases mentioned above were numerically simulated using ABAQUS/STANDARD (Abdelmalak 2007) and the output design parameters were plotted for the edge drop cases and for the edge lift cases, versus the soil-weather index. Note that in the design charts the soil weather index is used in terms of the suction change at the edge of the slab Uedge rather than the suction change in the free field U0. The numerical simulations indicated that Uedge is equal to one half of U0. For the edge drop case, Figure 18, Figure 19, Figure 20, and Figure 21 show the design charts for Leqv, Lgap, Fmax, and FV respectively. For the edge lift case, Figure 22, Figure 23, and Figure 24 show the design charts for Leqv, Fmax, and FV respectively.
Figure 20. Maximum deflection factor suction based design chart for edge drop case.
FV design chart (Edge drop)
0.90 0.80 0.70 0.60 0.50
deq=0.63 m deq=0.51 m deq=0.38 m
FV
0.40 0.30 0.20 0.10 0.00 0 0.2 0.4 0.6 0.8 1
Vmax = Fv qLeqv
deq=0.25 m deq=0.13 m
I ss H Uedge (m)
1.2
Figure 21. Maximum shear factor suction based design chart for edge drop case.
7.00 6.00 5.00
Leqv design chart (Edge drop)
Mmax
qL2 eqv 2
Leqv design chart (Edge lift)

8.00 7.00 6.00 5.00
M max
qL 2 eqv 2
Leqv (m)
4.00 3.00 2.00 1.00 0.00 0 0.2 0.4 0.6 0.8 1
deq=0.63 m deq=0.51 m
Leqv (m)
deq=0.38 m deq=0.25 m deq=0.13 m
4.00 3.00 2.00 1.00
1.2
Iss H Uedge (m)
0.00 0 0.2 0.4 0.6 0.8 1 1.2
I ss H Uedge (m)
Figure 18. Equivalent cantilever length suction based design chart for edge drop case.
Figure 22. Equivalent cantilever length suction based design chart for edge lift case.
F max design chart (Edge lift)
max qL 4 eqv F max EI
5.00 4.50 4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50 0.00 0 0.2
Lgap design chart (Edge drop)

4.50 4.00
3.50 3.00
Lgap (m)
F max
2.50 2.00 1.50 1.00 0.50 0.00
0.4
0.6
0.8
1.2
0.2
0.4
0.6
0.8
1.2
Iss H Uedge (m)
Iss H Uedge (m)
Figure 19. Unsupported length suction based design chart for edge drop case.
Figure 23. Maximum deflection factor suction based design chart for edge lift case.
11
1.20 1.00 0.80 0.60 0.40 0.20 0.00 0 0.2
FV design chart (Edge lift)

7.00 6.00
Leqv design chart (Edge drop)

M max qL 2 eqv 2
L eqv (m)
Vmax Fv qL eqv
deq=0.63 m deq=0.51 m deq=0.38 m deq=0.25 m deq=0.13 m 0.4 0.6 0.8 1 1.2
5.00 4.00 3.00 2.00 1.00 0.00 0 0.1 0.2 0.3 0.4 0.5
FV
0.6
Iss H Uedge (m)
H w edge (m)
Figure 24. Maximum shear factor suction based design chart for edge lift case
Figure 25. Equivalent cantilever length water content based design chart for edge drop case.
Lgap design chart (Edge drop)
5.00 4.50 4.00 3.50 3.00 deq=0.63 m deq=0.51 m deq=0.38 m deq=0.25 m deq=0.13 m
6.5
Water content based design charts
The change in surface suction plays an important role in the mound shape equation; consequently it appears as an intrinsic component in the soilweather index. However, measuring suction is more complicated than measuring water content, practitioners are more familiar with water content than with suction, and there is a much larger database of water content data than suction data. Therefore, it is convenient to have design charts in terms of water content in addition to having design chart in terms of suction. The soil water characteristic curve indicates that the change in suction U0 (pF) and the change in water content w0 are linked by the specific water capacity Cw. w0 Cw U0 2 wedge (7)
L gap (m)
2.50 2.00 1.50 1.00 0.50 0.00 0
0.1
0.2
0.3
0.4
0.5
0.6
H wedge (m)
Figure 26. Unsupported length water content based design chart for edge drop case.
F max design chart (Edge drop)
4.50 4.00 3.50 3.00
Indeed the change in water content at the edge of the slab wedge is one half of the change in water content in the free field w0 since Equation 7 is linear and since U0 2 Uedge as mentioned before. Furthermore Abdelmalak (2007) showed that there is a simple empirical relationship between the specific water capacity Cw and the shrink-swell index Iss: Cw Iss/2 Therefore, IS-W 2 H w0 H wedge (9) (8)
F max
2.50 2.00 1.50 1.00 0.50 0.00 0 0.1 0.2 0.3 0.4 0.5
H wedge (m)
0.6
Figure 27. Maximum deflection factor water content based design chart for edge drop case.
FV design chart (Edge drop)
0.90 0.80 0.70 0.60 0.50
With Equation 9 it was easy to transform the suction based design charts into water content based design charts by simply replacing the definition of the soil weather index on the horizontal axis. An advantage of the water content base design charts is that consultants may have sufficient water content data in their files to estimate the depth of the active zone H and the change in water content w0. As an example Briaud et al. (2003) organized a water content database which gave average values of w0 for three cities in Texas. Note that the design charts are
Vmax Fv qLeqv
FV
0.40 0.30 0.20 0.10 0.00 0 0.1 0.2 0.3 0.4 0.5
0.6
H wedge (m)
Figure 28. Maximum shear factor water content based design chart for edge drop case.
12
Leqv design chart (Edge lift)

8.00 7.00 6.00
M max
qL 2 eqv 2
Leqv (m)
5.00 4.00 3.00 2.00 1.00 0.00 0 0.1 0.2 0.3 0.4 0.5
criterion. If the deflection criterion is not met, a larger beam depth is assumed. The input data for this example consists of the soil weather data and the slab data: Soil and weather data: Depth of movement zone, H 3.0 m Soil surface water content change w0 20% Slab data: Slab dimensions 20 20 m Beam spacing, s 3.0 m (for both directions) Beam depth, h 1.2 m Beam width, b 0.3 m Slab load, w 10 kPa The calculations then proceed with the SoilWeather Index Isw calculations
0.6
H wedge (m)
Figure 29. Equivalent cantilever length water content based design chart for edge lift case.
F max design chart (Edge lift)
4.50 4.00 3.50 3.00
2.50 2.00 1.50 1.00 0.50 0.00 0 0.1 0.2 0.3 0.4 0.5
wedge 0.5 w0 0.5 0.2 0.1 or 10% Is-w wedge H 0.1 3 0.3 m and then the slab bending stiffness EI E bh3/12 2 107 0.3 1.23 / 12 8.64 105 kN.m2 which leads to the equivalent slab thickness
3 /12 b h3/12 s deq 1/3 deq h (b/s) 1.2 (0.3/3)1/3 0.56 m
F max
0.6
H wedge (m)
Figure 30. Maximum deflection factor water content based design chart for edge lift case.
FV design chart (Edge lift)
1.20 1.00 0.80
The values of the design parameters are read on the water content charts for the edge drop case
Vmax Fv qLeqv
0.60 0.40 0.20 0.00
Leq 5.3 m for maximum moment Lgap 3.6 m for information Fmax 2.9 for maximum deflection Fv 0.8 for maximum shear The maximum bending moment is calculated as q 10 3 30 kN/m line load on the beam
2 Mmax 0.5 q Leq 0.5 30 5.32 421.3 kN m
FV
0.1
0.2
H wedge (m)
0.3
0.4
0.5
0.6
Figure 31. Maximum shear factor water content based design chart for edge lift case.
The maximum deflection is calculated as

4 /Fmax EI max q Leq 30 5.34/2.9 8.64 105 max 9.5 103 m
given in terms of wedge which is equal to one half of w0. For the edge drop case, Figure 25, Figure 26, Figure 27, and Figure 28 show the design charts for Leqv, Lgap, Fmax, and FV respectively. For the edge lift case, Figure 29, Figure 30, and Figure 31 show the design charts for Leqv, Fmax, and FV respectively. 7 DESIGN EXAMPLE
The maximum shear force is calculated as Vmax Fv q Leq 0.8 30 5.3 127.2 kN This results in a distortion of 0.5 L/max 10/9.5 103 1050 Leq/max 5.3/9.5 103 558 Ratios less than 450 are typically acceptable according to the American Concrete Institute. This ratio
Note that the design process advances by trial and error in the sense that the beam dimension and spacing are assumed and then the resulting deflection is calculated and checked against the distortion
13
however depends on the type of structure and the local code should be used for this step. Since L is much larger than 2H there is no need for a reduction of Leq. This complete the design for the edge drop case. The edge lift case would proceed similarly. This example shows how simple the proposed method is. Note that this example is an extreme case as a w0 of 20% corresponds to extreme weather conditions. This is why the beam depth is significant. NOMENCLATURE The following list gives the parameters associated with this paper and their definitions. Figure 7 illustrates their definition: b D deq em beam width total beam depth thickness of a flat slab with the same moment of inertia as the stiffened slab edge moisture distance. (Distance from the edge of the slab to the point where the water will penetrate horizontally below a weightless perfectly flexible cover) bending stiffness of the slab product of the modulus of elasticity of the slab material and the moment of inertia of the slab qL4 eqv maximum deflection factor, F max max EI . This factor is 8 for a cantilever beam but will be different for the slab on grade since it is not exactly a cantilever situation Vmax maximum shear factor, Fv qL . This eqv number would be 1 for a cantilever beam but will be different for the slab on grade since it is not exactly a cantilever situation depth of active zone. (Depth to which the variation of water content or suction will create movement of the soil) shrink-swell index. (Range of water content between the shrinkage limit obtained in a free shrink test and the swell limit obtained in a free swell test. The shrink swell index is a very good indicator of a soil shrinkswell potential) total line load applied to the beam by the slab over the tributary area in Figure 6 (including dead weight of slab and imposed loads). q s w equivalent cantilever length. (Length of slab which gives the maximum bending moment in the beam when using the cantilever beam 2 formula Mmax qLeqv /2) unsupported length. (Length of slab without soil support underneath it) maximum bending moment in the slab beam center to center spacing maximum shear force in the slab
w ym max Uedge U0 wedge w0
average pressure generated by the slab weight and the load on the slab vertical movement. (Difference in elevation due to swelling or shrinking between the two extremities of the em distance) difference in elevation between the center of the slab and the edge of the slab change in suction at the ground surface at the edge of the slab change in suction. (Change in suction in pF (log units) in the free field at the ground surface) change in water content at the ground surface at the edge of the slab change in water content. (Change in water content in the free field at the ground surface)
REFERENCES
Abdelmalak, R. 2007. Soil structure interaction for shrinkswell soils -A new design procedure for foundation slabs on shrink-swell soils, Ph.D. dissertation, Department of Civil Engineering, Texas A&M University, College Station, TX. Australian Standard (AS2870). 1990. Residential Slabs and Footings, Part 2: Guide to Design by Engineering Principles, AS 2870.2 Standard House, Sydney NSW, Australia. Australian Standard (AS2870). 1996. Residential Slabs and Footings, AS 2870 Standard House, Sydney NSW, Australia. Briaud, J.-L., Zhang, X., and Moon, S. 2003. Shrink Test- Water Content Method for Shrink and Swell Predictions. ASCE Journal of Geotechnical and Geoenvironmental Engineering, 129(7): 590600. Building Research Advisory Board (BRAB). 1968. National Research Council Criteria for Selection and Design of Residential Slabs-on-Ground, U.S. National Academy of Sciences Publication 1571. Holtz, R.D., and Kovacs, W.D. 1981. An Introduction to Geotechnical Engineering, Prentice-Hall, Englewood Cliffs, N.J., 180. Mitchell, P.W. 1979. The structural analysis of footings on expansive soils. Research Report No. 1, K. W. G. Smith and Assoc. Pty. Ltd., Newton, South Australia. Post-Tensioning Institute (PTI). 1996. Design and Construction of Post-Tensioned Slabs-on-Ground, 2nd Ed., Phoenix, AZ, USA. Post-Tensioning Institute (PTI). 2004. Design and Construction of Post-Tensioned Slabs-on-Ground, 3rd Ed., Phoenix, AZ, USA. Wire Reinforcement Institute (WRI). 1981. Design of Slab-on- Ground Foundations, Findlay, Ohio, USA, August. Zhang, X. 2004. Consolidation theories for saturatedunsaturated soils and numerical simulation of residential buildings on expansive soils, Ph.D. dissertation, Department of Civil Engineering, Texas A&M University, College Station, TX.
EI Fmax
FV
H Iss
Leq
Lgap Mmax s Vmax
14
Geotechnical and environmental models involving unsaturated soils and rocks

Sebasti Olivella
Department of Geotechnical Engineering and Geosciences, Escola Tecnica Superior dEnginyers de Camins, Canals i Ports, Universidad Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: The general formulation for unsaturated soil analysis is described. This includes mass, momentum and energy balance. The mechanical models for unsaturated soils are defined in terms of stress, suction and temperature and permit to obtain deformations caused by variations in these variables. Geotechnical problems in which the mechanical part plays a major role are described. Problems involving swelling and collapse deformations are first described and secondly problems involving large permeability variations. Finally, another section is devoted to present applications related to unsaturated soils which include chemical processes.
INTRODUCTION
At present, a general approach for nonisothermal multiphase flow and deformation in geological porous media is applied to a variety of environmental and geotechnical problems. The formulation is solved usually by numerical methods as it includes several nonlinear functions. Existing programs have been developed by combination of existing ones or newly developed. Numerical solutions using finite elements are quite standard. Finite element method was introduced in unsaturated soil geotechnics mainly because the mechanical part of the problem (Lloret and Alonso, 1980), while in unsaturated flow and multiphase flow finite volumes (Pruess, 1987) was most commonly used. In order to model failure problems and specially when softening of the materials takes place, other methods are incorporated like mesh free methods, extended finite elements, mixed finite elements. However, little investigations have been performed in these fields using unsaturated flow and deformation approaches. Li et al (2005) developed a comprehensive formulation for slope stability analysis which is based on mixed finite elements and includes THM calculations. A field of research which is receiving significant attention corresponds to the development of formulations including double structure which are required in some cases for unsaturated soils and especially for expansive materials (Alonso et al 1999). In this context, Sanchez et al (2005, 2008) developed a formulation that permits to analyze problems involving soils which can be idealized as having two pore volumes undergoing different processes.
Another field of research concerns the unsaturated behavior rockfill materials which are used for earth dam construction. Oldecop and Alonso (2001, 2003) have developed a model that represents well the response of rockfill under relative humidity changes. The problem of discontinuities and interfaces is receiving attention for unsaturated soils recently. This is necessary in problems of stability of geotechnical structures (such as earth dams and embankments) when natural discontinuities exist and unsaturated conditions prevail. It is also necessary to model soil-structure contacts. In case of environmental problems such us storage of nuclear waste interfaces between clay barriers elements and host rock require special treatment. Zandarin et al (2009) have developed a special element which incorporates an elastoplastic mechanical response under stress, temperature and suction effects. In addition, transport properties (permeability, diffusivity, thermal conductivity) specialize depending on the normal and longitudinal directions using aperture variations and step functions. In the case of environmental problems it is also necessary to incorporate geochemical aspects as processes of clay deformations are influenced greatly by the presence of solutes (Barbour, S.L. and Yang, N., 1993; Di Maio, C., 1996; Hueckel T., 1997), but little investigations have been done considering unsaturated conditions. Saaltink et al (2004) coupled a geochemical formulation with a multiphase flow problem and solved problems related to evaporation in mine tailings (Acero et al 2008). The work by Guimaraes et al (2007) couple reactive transport and multiphase flow in deformable materials and
15
was applied to swelling clay barriers subjected to heating and hydration. Actually, swelling and collapse processes show significant influence of the presence of the solutes, and therefore the mechanical model requires the incorporation of the effect of solutes. Liu et al (2005) have developed a constitutive model for the chemo-hydro-mechanical response of unsaturated clays. As concrete is also viewed as porous unsaturated medium some common features with unsaturated geological media are of interest. Chemo-thermo-hygro-mechanical response has been investigated and of particular interest is the work by Gawin et al (2003) which incorporates the thermochemical and mechanical degradation of the porous material into a coupled THMC model. An emerging related field of research is the sequestration of carbon dioxide in geological formations. Vilarrasa et al (2010) have investigated the mechanical response of the rocks when injection at high pressures of carbon dioxide is carried out. The conditions of high confinement, two phase flow and chemical interactions require also specific characterization of properties of the cap rock which is the element that is foreseen to confine the injected gas. 2 GENERAL FORMULATION FOR UNSATURATED SOIL ANALYSIS
where s is the mass of solid per unit volume of solid and js is the flux of solid. From this equation, an expression for porosity variation can be obtained using the flux of solid that has been written as the velocity of the solid multiplied by the volumetric fraction occupied by the solid phase and the density: Ds (1 ) Dss du (1 ) Dt Dt dt s Ds () du where () Dt t dt
(2)
Equation (2) expresses the variation of porosity caused by volumetric deformation and solid density variations. Two fluid phases are considered: gas and liquid. The total mass balance of a component i present in each phase (for instance water that may evaporate) is expressed as:
i ( l l Sl igg Sg ) ( jil jig ) f i t
(3)
A general formulation for multiphase flow in porous media is described briefly in this section. It will be considered that the state variables (unknowns) are: solid displacements, u (three spatial directions) and liquid pressure, Pl. Balance of momentum for the medium as a whole is reduced to the equation of stress equilibrium together with a mechanical constitutive model which relates stresses with strains. Strains are defined in terms of displacements. Small strains and small strain rates for solid deformation are assumed. Advective terms due to solid displacement are neglected once the formulation is written in terms of material derivatives (in fact, material derivatives are approximated as eulerian time derivatives). In this way, volumetric strain is properly considered. The governing equations for non-isothermal multiphase flow of water and gas through porous deformable saline media have been presented by Olivella et al. (1994). A detailed derivation is given there, and a brief description of the formulation for thermo-hydro-mechanical problems is presented here. Mass balance of solid present in the medium is written in the eulerian form as: du (s (1 )) ( js ) 0 js s (1 ) dt t (1)
Where Sl,Sg are the phase degree of saturations; il , ig are the mass fractions of the component i in each phase (mole fraction can be considered as well); l, g are the phase densities, jil , jig are the mass fluxes of the component i in each phase and fi is an external supply of mass of component i. In this formulation the components in most cases considered in this paper are the water and air, but air can be substituted by generated hydrogen in case of corrosion (or when modeling gas injection tests). The mass flux of components is a combination of a non-advective flux (diffusion dispersion) written as iil , iig the advective Darcy flux written as ql, qg and another advective term caused by the solid motion, which is proportional to the solid velocity du/dt:
i ( l l Sl ig g Sg ) t iil iig il l ql ig g q g i du f ( il l Sl ig g Sg ) dt
(4)
Expansion of time derivatives and spatial derivatives and the use of the material derivative with respect to the solid velocities, leads to the following equation which is actually a lagrangian: Ds ( il l Sl ig g Sg ) Dt ( il l Sl ig g Sg ) Ds Dt (5)
du ( il l Sl ig g Sg ) dt ( iil iig il l ql ig g q g ) f i
16
The mass balance of solid (equation, 2) is introduced in the mass balance of components to obtain, after some algebra: Ds ( il l Sl ig g Sg ) ( il l Sl ig g Sg ) du dt (6)
This equation contains three storage terms, related to: I. Storativity due to changes in pore pressure which i and the coupled can be obtained from 1i dm dp terms obtained as:
m l 1 dmi 1 dmi . ; mi dpg mi dT 1 d s deformability since s dx
Dt (1 ) Dss ( il l Sl ig gSg ) Dt s ( iil iig il l ql ig gq g ) f i
The volumetric deformation rate term [dv/ dt (du/dt)] couples the mass balance equations with the deformation of the porous medium. This requires the coupled solution of the mechanical equations. The momentum balance for the porous medium reduces to the equilibrium of stresses if the inertial terms are neglected: b0 (7)
where is the stress tensor and b is the vector of body forces. In addition, the balance of energy is introduced in case of non-isothermal conditions. The equation for internal energy balance for the porous medium is established taking into account the internal energy of each phase (Es, El, Eg):
1 K is the II. Solid x deformability of the soil particles with respect to the variable x, which can be pressure or temperature. This term is not significant (sometimes neglected) for soils but it is essential for rocks as it combines with the bulk deformation to produce the changes in porosity which is very small and suffers relatively large variations. III. Soil skeleton deformability since the divergence of solid velocity can be transformed into d u d (d ) dt ( u ) dtv , and by means of an dt appropriate constitutive model that contains suctionandtemperaturedependenciesthevolumetric d ( , s ,T ) d d d . strain rate is: v dt d v ddt dsv ds dsv dT dt dt The mechanical model may include effective or net stress terms or suction terms and temperature terms.
( Ess (1 ) El l Sl E gg Sg ) t ( ic jEs jEl jEg ) f Q
(8)
A model for unsaturated soils such as the elastoplastic model BBM (Alonso et al, 1990) including the effect of effective or net stresses and the suction is written as: d Dd hds d D 1d D 1hds d v m t d m t D 1d m t D 1hds m t 1 1 1 0 0 0
where ic is energy flux due to conduction through the porous medium, the other fluxes (jEs, jEl, jEg) are advective fluxes of energy caused by mass motions and f Q is an internal/external energy supply. Energy generation by deformations can be considered included in this source term and can be incorporated without loss of generality (for instance to analyze failure problems with high velocities). The use of the material derivative allows obtaining an equation formally similar to the mass balance of components. The simultaneous solution of the coupled equations given above produces the spatial and temporal evolution of displacements, liquid pressure, gas pressure and temperature; which are considered as state variables or unknowns in this THM approach. If the mass of i is put into a single variable mi il l Sl igg Sg the balance equation can be transformed into: 1 Ds mi (1 ) Dss du dt Dt mi Dt s fi 1 i ( iil iig il l ql ig g q g ) i m m
(10)
(9)
Where suction can be defined as s max(pg pl,0) and net stress as (pg, pl)m. This is valid for saturated and unsaturated conditions assuming stresses in compression as negatives. The model parameters are included in D which is the stiffness tensor (6 6) or constitutive matrix for changes in net or effective stress and h which is the constitutive vector for changes in suction. Both are nonlinear functions. The preceding set of constitutive relations was cast in the form proposed by Perzyna (1966) to develop a family of viscoplastic constitutive models for soils which can be used either as viscoplastic or, for very low viscosities, as elastoplastic. Incorporation of time effects in deformation of geological materials can be done in several ways: viscoelasticity, viscoplasticity (the type of model described here) and creep (the type of model that is used for saline materials, for
17
secondary consolidation of soils or for long term response of rockfills). In the case of viscoplasticity, total strain rates ( ) are calculated as a sum of elastic ( e) and viscoplastic components ( vp ): e vp (11)
Table 2. Basic relationships for RM. Rockfill Model (RM) Irreversible volumetric deformation p py d v d v dp p py d v
i d s dp i d v i i
d v
With the the viscoplastic component computed through: vp F G (12)
Compres-sibility
i d s
d d s o s ln
s patm patm
where is a fluidity parameter and (F) is a flow function defined as: ( F ) ( F ) if F 0 ( F ) 0 if F 0 (13)
Hardening law Loading-collapse curve (LC)
dpo
* po * po
d v
i

*
py po s po py po s
i * o
A usual expression for the flow function is (Perzyna, 1966): F F Fo

N
py Shear strength critical state parameter
p p s
y i d
(14)
M s M dry M dry
The shape adopted for F and G and the functions for irreversible deformations are summarized in Tables 1 and 2 for the particular case of
Table 1. Basic relationships for BBM. Barcelona Basic Model (BBM) Irreversible volumetric deformation Compressibility d v
M 10 p M sat sat M dry
s
atm
Tensile strength yield surface
F p, q, s
2
ps(s) ps0 kss q M p ps s po s p

2
s dp
1 e p
Plastic potential
G p, q, s
2
s 0 1 r t r t exp s
dpo
* * 1 e po p d v 0
q M p ps s po s p
2
Creep strain
d v dt

Hardening law
Loading-collapse curve (LC) Shear strength critical state parameter Tensile strength yield surface Plastic potential
p* s po s p o pc
c
0
s patm c 1 ln patm t dd 1 q; d 0.1v 3d dt
v
p;
v
M(s) M ps(s) ps0 kss

F p, q, s q M p ps s po s p
2 2
G p, q, s
2
q M p ps s po s p
2
unsaturated models BBM (Barcelona Basic Model) and RM (Rockfill Model). The use of viscoplasticity for engineering applications has the advantage that the solution is regularized and, to some extent, problems involving failure can be solved. It is also robust to account for tensions which are not significant for the solution of the problems but may appear in practical solution of field problems (for instance near the boundary the slope of an earth dam).
18
The elastic part can be written in the following way (for isothermal conditions):
e d v
dp s ds 1 e p 1 e s 0.1
(15)
Depending on the values of the parameters, very high compressibility can be obtained and in such case the nonlinear volumetric stiffness should be limited with a Kmin. When the irreversible part is formulated as described above (i.e. addition of elastic plus viscoplastic) the elastic deformations can be modified independently. For instance, a linear elastic equation can be combined with a viscoplastic part without loss of generality. The non-isothermal case including a linear thermal expansion of the porous material (and solid phase) is a first approximation for incorporation of thermo - mechanical processes. 3 HYDRO-MECHANICAL ANALYSES IN UNSATURATED SOILS. SWELLING AND COLLAPSE
decrease of suction induces volumetric deformations of collapse by variation of capillary forces (LC curve) volumetric deformation induces decrease of suction by mass conservation (retention curve plays a significant role) decrease of suction by mass conservation induces additional deformations of collapse This description of the process indicates that when suction is controlled directly, the determination of parameters is more efficient. It is very difficult to control suction, via control of the volume of the water. Collapse problems can be encountered in a number of civil engineering situations, one of them being the construction of embankments and the subsequent wetting of the soil. Recently, an embankment has been constructed to support a high velocity train track near Barcelona. Shortly after construction, recharge ponds have been planed nearby in order to increase the underground storage of water for future supply. The geometry is depicted in Figure 1 together with the results and model calibration for two oedometer tests. The investigated soil shows relatively small collapse deformations.
The base approach to model unsaturated soil mechanics is the hydro-mechanical part of the formulation. This approach was first considered a natural extension of the coupled hydro-mechanical approach for saturated soils. Therefore the consolidation theory was extended to take into account unsaturated soils. Consolidation theory, however, does not require that constitutive model parameters depend on pressure. In contrast, in the unsaturated soil formulation, it is essential that plastic compressibility depends on suction, giving rise to the collapse process (in the case of BBM model this is: (s)). The other process to be taken into account is the expansivity and for that, the basic term is an elastic term giving isotropic volumetric expansion as suction decreases (in the case of BBM this is: dv (s/(1 e))ds/(s + patm)). The BBM model uses the concept of net stress defined as ( max(pg, pl)) which incorporates the saturated case. These concepts are sufficient for basic analysis in unsaturated soils. Collapse problems are typical in embankments and slopes. Collapse is associated to open structures of the soil and it depends on the conditions of compaction. One of the main difficulties in modeling collapsible soils is the strong coupling between the deformations induced and the suction changes. The process of collapse can be described in the following way: wetting of a dry soil induces decrease of suction (permeabilityincluding relative permeability plays a major role)
1.56 Experimental 1.54 1.52 Specific volume 1.50 1.48 1.46 1.44 1.42 0.010 Modelo
0.100 Vertical stress (MPa)
1.000
1.54 1.52 1.50 Specific volume 1.48 1.46 1.44 1.42 1.40 0.010 Experimental Modelo 0.100 Vertical stress (MPa) 1.000
Figure 1. Finite element mesh and results of oedometer test for the soil under the embankment.
19
At 0.3 MPa and 0.8 MPa the collapse deformation is of the order of 4%. On the other hand, the worse initial condition for collapse occurs when the soil is NC after construction of the embankment. To obtain that condition, the natural soil deposition and the embankment construction have been simulated. This guarantees the NC state after construction with respect to the suction state. The parameters for BBM model in this case that were calibrated using experimental result (figure 1) are: (0) 0.035, 1.2 MPa1, r 0.15, and M 1.1. The preconsolidation stress is 0.09 MPa. The elastic part has been considered linear with E 50 MPa and Poison ratio of 0.3. Figure 2 shows the vertical deformations induced by the recharge. The heave in the ponds is motivated by the increase in pore pressure to positive values (unloading in terms of effective stress) while the settlement in the embankment is caused by the collapse induced by wetting. The stress path in the q-p diagram shows the construction phase and the wetting phase. When the hydration starts the path changes its direction because the yield function F reduces and drags the stress point. This is a common feature of collapse problems, i.e. the dragging of the stress point as strength is reduced.
(a)
(b)
0.3
Point ( 57.7044, 23.347, 0) evolution Point ( 71.2578, 21.8536, 0) evolution Function F=0 Function F=0 q=Mp'
0.2
0.1
0 -0.1 0 p (MPa) 0.1 0.2 0.3
(c)
Figure 2. Vertical displacements induced at long term conditions of water infiltration in the pond (displacements induced by embankment construction are not included). a) Isotropic permeability (k 3 1014 m2), b) Anisotropic permeability (kh 10k, kv k), c) Stress path during construction followed by water infiltration including the final plastification envelopes.
This strength reduction may be due to cohesion variations, preconsolidation stress variation and, variation of other parameters in the curve. In most problems of unsaturated soil, the classical effect of unsaturated behavior is attributed to changes in cohesion or changes in elastoplastic compressibility (). The changes in () imply changes in preconsolidation mean stress (LC curve). In addition to these parameters, it is possible that the friction parameter changes. This dependency is not considered in unsaturated soil mechanics approaches (Alonso et al 1990) because it is not observed in reality. Actually friction is a process which is considered independent of capillary forces or degree of wetting. The possibility of water lubricating the grain contacts could explain the variation of friction angle with suction but the effect is small. In the case of rockfills, the situation is different as the deformations mechanisms are different (Oldecop and Alonso, 2001) and shear deformation includes particle breakage. In fact, there are experimental evidences of strength loses during triaxial testing of rockfills (Figure 3a, b). The simulation of these tests with the rockfill model developed by Oldecop and Alonso (2001) requires to introduce a critical state slope (friction parameter) that ranges from M 1.75 (dry) to M 1.30 (sat) for the inner rockfill used in the construction of Beliche Dam (Alonso et al, 2005). The outer rockfill also required a variable friction ranging from M 1.9 (dry) to M 1.8 (sat). The implication of the different friction angle for different water content is that the yield surface has different inclination during plastic deformation and the shear deformations will be higher the wetter the rockfill becomes (Figure 3d). Another application of unsaturated soil effects to earth dams has been presented by Pinyol et al (2008) and shows that the retention properties of the rockfill and clay core influence the response of the pressure evolution during drawdown. Failure of geomembranes is another example of the practical problems that collapse may produce. The problems appear as the weight of the water acts on its surface and the soil underneath collapses. Figure 4 shows a picture of a reservoir constructed with a membrane for water supply to an artificial snow production system. A damage on the membrane occurred due to deformation by collapse of the underlying soil, and this can be explained by soil deformation induced by wetting. In order to simulate this problem a simple calculation has been done including the following elements: a collapsible soil, a geomembrane, a reservoir and two point sources simulating pipelines having water loses (water at atmospheric pressure). Small water loses may start the collapse deformation which may lead to further damage of the pipeline
q (MPa)
20
(a)
2 Deviatoric stress (MPa) 1.5 1
Confining pressure = 0.1 MPa
Model Experimental
Confining pressure = 0.3 MPa
Figure 4. Photography of a typical reservoir constructed with a geomembrane for artificial snow production.
0.5 0 0 2 4 6 8 10 Axial strain (%) 12 14 16
geomembrane
(b)
2.5 Water pressure (MPa) 2 1.5 1 0.5 0 -0.5
0.2 Vertical displacements (m) . 0
J40
Settlements due to dam construction and impoundment Settlements due to rain period
reservoir pipelines
-0.2 -0.4 -0.6 -0.8 -1

Measurements start
(a)
0 -0.05 Vertical Displacement (m) -0.1 -0.1 -0.15 -0.15 -0.2 -0.25 -0.2 0 -0.05 Vertical stress (MPa)
Accumulated Rainfall (mm)
-1.2
Water pressure in the rockfill
-1
14000 12000 10000 8000 6000 4000 2000 0 0 360 720 1080 1440 1800 2160 2520 2880 3240 3600
Days
(c)
-0.3 0 20 40 60 80 Distance (m) 100 120
-0.25 140
0.6 0.5 0.4 q (MPa) 0.3 0.2 0.1 0 0 0.1 0.2 0.3 p (MPa) 0.4 0.5 0.6 0.7
Dry strength (end of construction) Wet strength (operation) Inner rockfill downstream (2) Inner rockfill upstream (1)
(b)
0.25
1 2
0.2 Deviatoric stress (q)

Plastification begins
Stress path Yield curve for initial state Yield curve during wetting Yield curve after wetting
Plastification during wetting: collapse
0.15
0.1
End of collapse
0.05
(d)
0.05
0.1
0.15
0.2
0.25
0.3
0.35
Mean stress (p)
(c)
Figure 3. Beliche Dam: (a) Sketch of the stress-suction path for a triaxial test including inundation, (b) Triaxial test on rockfill (experimental and simulation), (c) Vertical displacements and accumulated rainfall on the surface marker J40 of Beliche Dam, (d) Stress path during construction and operation for a points in the inner rockfill.
Figure 5. a) Water pressure distribution in a 2D scheme including the ground, the membrane and the water, b) vertical displacements and vertical stresses along the geomembrane, c) stress path in a point near the pipeline (loading by water level rise followed by collapse due to wetting).
and membrane which in turn may lead to large water loses. The process is catastrophic in the sense that it is self-accelerated. Figure 5 shows (a) the distribution of water pressures, (b) the displacements along the membrane and the vertical stresses and (c) the stress path for a point near the water source.
Due to collapse deformation, the stress distribution changes decreasing the load that is transmitted on the zone where the soil is collapsing. The stress path shows the phase of loading due to water level rise at the reservoir, including plastic deformations in two ways: stress changes and suction changes.
21
Swelling and expansivity of the soils takes place for a more dense materials (overconsolidated clays) and it is a process that is desired in some cases. Construction of isolation barriers requires expansive materials and this is expected to happen during humidification or hydration. In the last 20 years, a number of experimental investigations both in the laboratory and in situ have been performed in order to understand the behavior of such waste isolation schemes. The FEBEX experiment (Gens et al 2009) has been a successful exercise including a comprehensive number of investigations to learn about the response of the bentonite barriers. Another investigation in this field of isolation of waste is the EB experiment which consists in the hydration of a barrier made of pellets. The experiment consists in the artificial hydration of an isolation barrier made of bentonite blocks and pellets (figure 6a). The artificial hydration was achieved using a system of pipelines which permitted to inject water at different locations within the blocks and pellets (figure 6b). Further details can be found in Alonso and Hoffmann (2001) including modeling of the EDZ. One of the key issues in this experiment was to discriminate the ratio the injected water was stored in the micropores and macropores. The model used for the modeling presented in this paper is a relatively simple extension of BBM model which uses a suction micro (sm) to calculate the swelling deformations. The suction micro is calculated from the suction that prevails in the macropores (sM) according to a transfer equation: w micromicro(sm sM) (16)
Opalinus Clay
Bentonite pellets
W
WB2
SECTION A1
Simulated canister
Bentonite blocks
a)
b)
0.40 0.35 0.30 Porosity 0.25 0.20 0.15 0.10 0.05 0.00 0 100 200 300 400 500 600 Time (d) 700 800 900 1000 1100 Macroporosity (blocks) Macroporosity (pellets) Microporosity (pellets and blocks)
c)
1350 1200 Injected volume (l/m) 1050 900 750 600 450 300 150 0 0 250 500 750 1000 Time (d)
Calculated
870 l/m
18852
20000 Injected volume (l) 15000 10000
Measured
13500
From deformations, the porosity variations are considered with the two components, the micro and the macroporosity. These can be decomposed using the following equation for which micro porosity may be defined in two ways: M (1 M)m or M m with the corresponding equivalence. This is represented in figure 6c. Although the components of porosity undergo significant variations, the overall porosity shows little variations. Actually, the total porosity may change only if some zones compress and other expand because the total volume is practically constant in this problem, as it is limited by the drift walls. Both in the pellets and in the blocks the assumed microporosity increases while the macroporosity reduces. The impact on permeability is expected to be significant (it was not considered in this modeling). d v
Injected volume of water per 1 m of vertical half section
5000
0 1250 1500 1750
d)
Figure 6. EB in situ test: a) Scheme in vertical section, b) Calculated water pressure during hydration (artificial), c) Calculated evolution of porosity (micro and macro) in the pellets and in the blocks, d) Measured and calculated volume of injected water,
p dp s dsm 1 e p 1 e sm patm
(17)
The possibility to calculate these two components of porosity from the model permits to obtain a more accurate estimate of the water inflow. This is because a higher density of the water in the micropores can be considered. Using a density of 1400 kg/m3 for the water in the micropores the calculated water inflow was much more consistent with the measured values
22
(figure 6d). This is considered a confirmation of the assumed and calculated distribution of porosity between micro and macro. This is an evidence of the necessity of the double structure models because it not only influences the rate at which the water hydrates the clay components but also controls the amount of water that inflows. And this is important for the swelling process. As mentioned above, a general formulation that captures in a very appropriate way the phenomena taking place in these double structured materials has been developed by Sanchez et al (2005, 2008). The formulation includes an elastoplastic model for expansive materials which is formulated by means the generalized plasticity approach and an extended form of the balance equations for double porosity materials. The formulation has been applied to model in situ and mock up large scale tests involving THM processes. The application of this model to the Mock-up test has lead to a large improvement of swelling stress development induced by swelling under nonisothermal conditions. Figure 7 shows an sketch of the 1:1 scale mock-up experiment together with the calculated and measured relative humidities and stresses (Sanchez et al 2008). The heating up to 100oC of the simulated heaters produces evaporation of water (decrease of relative humidity) but this is compensated by artificial hydration on the outer boundary. Overall, the bentonitic clay tends to full saturation and therefore swells progressively. The comparison of the model predictions of stresses with measurements was greatly improved by the incorporation of the double structure model. Swelling and collapse deformation appear in unsaturated soils in different magnitude depending on the pore size distribution, average and variance. In general the collapse processes are related to capillary forces while the swelling are more explained by water migration towards the micropores (double layer) driven by osmotic suction. There are other possibilities, for instance the swelling induced by freezing water in a soil in the permafrost region. One of the relevant equations that permits to simulate the freezing process in a saturated soil is obtained by combination of a retention type curve (in this case van Genuchten) and the Clausius-Clapeyron equation:
1 1 i P l ln T 1 l i 273.15 l Sl 1 P
Figure 7. Mock up test. Simulation and measurements of relative humidity and radial stress. Modelling is performed using a double structure model.
(18)
Where T is temperature; P, are the parameters of the model; l is the latent heat of ice fusion; and l and i are the density for liquid water and ice, respectively. Figure 8 shows the swelling induced by freezing around a pipeline which was maintained at 8.5 oC for a long period of time. The approach includes a particular case of BBM model for freezing soils.
23
HYDRO-MECHANICAL ANALYSES IN UNSATURATED SOILS. LARGE INTRINSIC PERMEABILITY VARIATIONS
a)
80 Heave of pipeline [cm] 60 40 20 0
b)
Monthly varying air temp. Constant air temp. Berm constructed D G D G C C
Observed Simulated 0 200 400 600 Time [Days] 800 1000
c)
Figure 8. Deformed geometry and porosity predicted for Control section at (a) day 300 and (b) day 1000 for constant air temperature. c) Simulated and observed heave of pipeline (C: Control section, D: Deep burial section, G: Gravel section).
This approach is also useful to explain the ice lens formation when a freezing front is developed in the soil. After ice thawing a consolidation process leads to soil compression and porosity reduction. The analogies between the unsaturated soils and the freezing soils not only include the retention curve but may be extended to thermal conductivity and relative permeability. Permeability undergoes variations induced by swelling (more pore space available for water) and by freezing (less pore space available for liquid water). The combination of these two process controls the swelling process by accumulation of water in the frozen zone. The mechanical constitutive model to be considered should be investigated as the mechanical properties of the ice would have to be incorporated in the mechanical constitutive model of the mixture. Details of the formulation considered and analyses performed can be found in Nishimura et al (2009).
Various problems involving unsaturated soils require considering variable intrinsic permeability in modeling. Porosity variations induced by volumetric deformations of collapse or swelling cannot explain the large intrinsic permeability variations taking place in soils. The large variations are caused either by changes in soil structure for instance in Olivella and Gens (2001) it is shown that intrinsic permeability of low permeability clays changes dramatically as water content changes); by formation of highly permeable preferential paths induced by dilatancy and damage (Rummel and Weber, 2000) and by gas migration (Harrington and Horseman, 2003; Arnedo et al 2008) or gasfrac processes. An appropriate constitutive model is essential for calculating intrinsic permeability as a function of variables that depend on deformations or changes in structure. First, a model of an embedded discontinuity is applied to model and slope in an unsaturated soil and afterwards the model will be applied to problems of gas migration coupled to deformations. In the case of dilatance deformations in zones of localization, it is known that permeability increments may be significant (see for instance Rummel and Webber, 2000) and this in turn may lead to important consequences on the deformation process itself. Ehlers et al (2004) have presented a similar formulation and solved some slope stability problems including localization phenomena. Viscoplasticity is used as a regularization technique. The authors discuss cases of slope failure in the context of embankments subjected to water level increase and decrease at the upstream of an embankment dam. Several problems of slope stability are associated to rainfall events. A slope under unsaturated conditions is assumed and three scenarios are considered: constant permeability with no rain conditions (hydrostatic pressure profile), constant permeability with rain and variable permeability with rain (rain consists in assuming atmospheric pressure of water on the soil surface). Other climate events could be considered but this was not the objective of this exercise. The formation of the slope is simulated by the excavation of the soil in three layers. Figure 9 shows the displacements, the deformations and the liquid pressure corresponding to approximately 170 days. The three models give substantially different results. It is clear from figure 9 that the slope is stable but some irreversible
24
Figure 9. Distribution of displacements (scale with maximum of 0.03 m), irreversible deformations (scale with maximum of 0.2) and pressures. No rain and constant permeability case.
Figure 10. Distribution of displacements (scale with maximum of 0.3 m), irreversible deformations (scale with maximum of 0.2) and pressures. Constant rain and constant intrinsic permeability.
deformations occur. Displacements are very small, in general below 2 cm. Figure 10 shows the case with a rain saturating the surface and penetrating in the soil leading to an advancing saturation front. Results are also presented for 170 days. The hydraulic parameters in this case are a key factor and the suction decreases relatively slowly in the ground thus causing a slow development of the shear deformations. Finally, figure 11 shows the results for a case in which the permeability changes strongly with dilatancy. The results can only be presented for a time before 60 days as deformations accelerate and failure takes place. Actually, if one looks at Figure 12, the time of failure is well defined in the case for variable permeability. In contrast, the case with constant permeability shows an evolution of displacement approximately at constant velocity and this is a consequence of pressure evolution in the zone of irreversible deformations. Looking at the suction evolution (figure 13) the movements are associated to suction variations, and the slow or fast movement depends on the variation of suction which depends on the intrinsic permeability variations. The incorporation of the model simulating a discontinuity in the form of embedded element
Figure 11. Distribution of displacements (scale with maximum of 0.3 m), irreversible deformations and pressures. Rain and variable intrinsic permeability.
25
0.4
No rain, Constant permeability Rain, Constant permeability
0.3 displacement (m)
Rain, Variable permeability
progresive movement
failure 0.2
0.1
stable 0 0 20 40 60 80 100 Time 120 140 160 180 200
Figure 12. Evolution of movements at the slope crest for the cases analysed: no rain (hydrostatic suction profile), rain and constant intrinsic permeability (i.e. permeability only changes by saturation) and rain and variable intrinsic permeability (deformations influence the intrinsic permeability).
0.5
0.4
suction (kPa)
0.3
slow movement
0.2 failure 0.1

No rain, constant permeability Rain, Constant permeability Rain, Variable permeability
0 0 20 40 60 80 100 Time 120 140 160 180 200
Figure 13. Evolution of suction on the slope for the cases analysed: no rain (hydrostatic suction profile), rain and constant intrinsic permeability (i.e. permeability only changes by saturation) and rain and variable intrinsic permeability (deformations influence the intrinsic permeability).
Figure 14. Picture of the GMT in situ test together with a schematic representation. A concrete silo simulating a waste package is surrounded by a buffer material (sandbentonite) in contact with the host granitic rock.
was actually motivated by the necessity to simulate gas migration in barriers. An interesting example has been developed and is described in Olivella and Alonso (2008). The simulation of the GMT in situ test (Figure 14) was very challenging as several phenomena were combined including high rates of water and gas flow. The presence of the discontinuities was a key factor for both, the water flow taking place induced by an artificial water pressurization period and the gas flow taking place during gas injection. The test consists in various events of water and gas injection with different objectives. The most important ones are: the pressurization and gas injection tests. The first was carried out with the objective of achieving full saturation and a level of pressure in shorter time than under natural conditions. The gas injection tests simulate the pos-
sibility of gas generation by corrosion and have the objective to understand the process regarding preferential path development and coupling with mechanical interactions. Figure 15a shows the development of water fluxes during water injection which took place in a ring above the layers in the basis of the upper cavity. Figure 15b shows the development of gas fluxes which begin at the upper part of the concrete silo. Both processes are strongly influenced by the presence of discontinuities between the buffer and the host rock and by the presence of a permeable zone that hits the upper cavity. The assumption of anisotropy was a limiting factor of this model. Several measurements were used for comparison with the model results. Figure 16 shows some of the comparison plots, indicating that an overall picture of the problem was achieved. The pressure
26
900 800
PE/9/7/0
CAVERN PRESSURIZATION RGI1
RGI2
RGI3 RGI4
0.9 0.8 0.6 0.5 0.4 0.3 0.2 0.1 0 0.7
600 500 400 300 200 100 0
Calculated Gas Pressure Layer 9
29-may-01
24-may-02
19-may-03
13-may-04
27-ago-01
22-ago-02
17-ago-03
1000
1090
1180
1270
1360
1450
1540
1630
1720
1810
1900
1990
2080
2170
0.35 0.30 Volumetric water content 0.25 0.20 0.15 0.10

28-feb-01 18-feb-03 23-feb-02 25-nov-01 30-nov-00 20-nov-02 15-nov-03 13-feb-04 27-ago-01 17-ago-03 22-ago-02 24-may-02 29-may-01 19-may-03 13-may-04 11-ago-04 01-sep-00
2260
820
910
TV/8/3/3 TV/8/3/4 Calaculated Layer 8
Date
1000
1090
1180
1270
1360
1450
1540
1630
1720
1810
1900
1990
2080
2170
1000 900 800 700 600 500 400 300 200 100 0
01/09/2000 30/11/2000
2260
820
910
11-ago-04
0.35 0.30 0.25 0.20 0.15 0.10
28-feb-01
23-feb-02
18-feb-03
30-nov-00
25-nov-01
20-nov-02
15-nov-03
13-feb-04
1-sep-00
Horizontal stress (kPa)
TC/6/7/5 Horizontal stress - Layer 7 Horizontal stress - Layer 6
-1.0 -0.9 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
17/08/2003 13/02/2004 15/11/2003 13/05/2004 22/08/2002 20/11/2002 18/02/2003 19/05/2003 11/08/2004
28/02/2001
23/02/2002
29/05/2001
27/08/2001
25/11/2001
24/05/2002
Date
1.E-08 Intrinsic permeability (m2) . 1.E-09 1.E-10 1.E-11 1.E-12 1.E-13 1.E-14 1.E-15 1.E-16 1.E-17 1450 1540 1630 1720 1810 1900 2080 2170 1090 1180 1270 1360 1000 1990 2260 820 910 1.95 25 1.95 24.7 1.95 24.1 1.93 23
CAVERN PRESSURIZATION RGI1 RGI3 RGI2 RGI4
Time (days)
Figure 16. Evolution of pressures, water content, stresses and permeabilities for GMT in situ test. Figure 15. (a) Contour fields of water fluxes (m/s) during pressurization and (b) gas fluxes (m/s) during gas injection tests.
CHEMO-HYDRO-MECHANICAL ANALYSES IN UNSATURATED SOILS
evolution at injection points, the water content and the normal stresses are compared with in situ measurements. In addition, the calculated permeability is shown for the zones that undergo larger variations. Water pressurization phase and gas injection phases can be clearly identified in the pressure, stress and permeability plots.
Recently applications which include chemical aspects of unsaturated soils have appeared. Formulations that couple the geochemical aspects in geological materials and multiphase flow in deformable media have been developed and transformed into its numerical form (Saaltink et al 2004; Guimaraes et al, 2007). Geochemical aspects cover a wide range of problems: composition of groundwater and the
27
Pressure (MPa)
Pressure (kPa)
700
Calculated Liq Pressure Layer 9
interaction with minerals, contamination problems, influence of leakages on clay stability, osmotic effects and semipermeable membranes, variable density caused by solutes, precipitation and dissolution, among others. Coupling with deformations of the soils and rocks is in some cases essential. Gran et al (2005) have investigated the effects of salinity on multiphase flow using open soil columns under evaporation conditions. In this section, an example of applications dealing with chemical effects coupled with thermohydro-mechanical processes in a rock is presented. It concerns the heave of the platform of a tunnel excavated in suphate-bearing clay (Figure 17). Excavations in anhydritic-gypsiferous claystones may experience severe swelling phenomena. The intensity of the expansive behaviour observed in these materials is higher than the expansion developed in other expansive sulphate-free rocks and soils. Tunnel functionality and stability is impaired in most cases. Field observations in Lilla tunnel as well as laboratory tests results have demonstrated that the swelling process is associated with gypsum crystal growth in discontinuities. Before finalization of the high velocity train infrastructure, it was observed a large swelling on floor of the tunnel. This motivated several investigations and modifications of the construction. The coupled modelling of crystal formation driven by water evaporation and producing expansion of the rock was considered a possible cause of the swelling. This analysis helps to develop correction measures to avoid future difficulties (Berdugo, 2006). Plaza (2008) has shown how the mechanisms of permeability increase around a tunnel can enhance the water flow and chemical processes. Figure 18 is a sketch for permeability increase in a fractured rock around a tunnel which is then combined with crystal formation and further discontinuity opening. It is assumed that crystallization may cause discontinuity opening. At the same time, crystals are subjected to stresses and this affects its chemical potential and thus modifies the solubility. The causes of crystallization may be various. Here, evaporation caused by the prevailing atmospheric conditions in the drift is the driving force for crystallization. Evaporation induces increase of concentration and at some point it causes supersaturation and the consequent crystallization. Other causes of precipitation will be investigated in case that evaporation cannot explain completely the actual behavior. This model is motivated by the need to understand the causes of soil heave in tunnels excavated for high velocity train development in Spain in zones where sulphate rocks are encountered. Figure 18 shows the mesh used for the coupled THMC calculations and the movements of heave obtained in the rock below the floor.
800 700 Vertical displacement (mm) 600 500 400 300 200 100 0 0
12/1/02
12/19/02
Construction of the invert
411+880
1/17/03
411+420 411+900 411+540 411+380 412+545 411+920 411+940 412+240

Slab axis reference at 9/20/02
100
200 300 Time (days)
400
12/15/03
9/20/02
3/20/03
6/19/03
9/15/03
500
Figure 17. Floor heave during construction in Lilla tunnel (Lower Ebro Basin (Spain), Flat Slab) and sample of rock showing the formation of a crystal vein.
Apart from the transport of solutes and the precipitation/dissolution term, an equation relating the crystallization and the deformation is required. This is written as a function of crystal density and precipitation/dissolution mass rate: 1 dm d c dt c dt
min ( max min )e Pc (1 ( 0 ))
Pc k
(19)
28
2/2/04
SOLERA PLANA Desplazamiento vertical bajo solera (m)

-0.05 0 0.5 1 1.5 0.00 0.05 0.10 0.15
Profundidad (m)
2 2.5 3 3.5 4 4.5 5
Figure 18. Conceptual model for permeability increase associated to excavation and crystallization of salts.
t = 0.1 aos t = 0.5 aos t = 1 ao
which assumes that crystal growth depends on stress level or pressure on the crystal which is in turn a function of porosity, assuming that the lower the porosity the lower the contact stress between crystals. Olivella and Gens (2002) have demonstrated the sensitivity of dissolution to stress level when investigating the compaction process of wet crushed NaCl salts. The above mentioned movements are associated with changes in permeability and consequent flow enhancement. In this case, the permeability is first increased during excavation and later further developed by crystallization of salts induced by evaporation. Figure 19 shows the variations of permeability as predicted by the model. Due to excavation, permeability increases by a factor of one order of magnitude approximately and 5 m of rock are affected. This is typical and it is known as EDZ (Excavation Disturbed Zone) development.
t = 1.5 aos t = 2 aos
Figure 19. Finite element mesh and development of vertical displacements induced by excavation and crystallization.
The study of EDZ is of particular interest when drifts are used for waste storage as it may become a preferential path for migration of contaminated fluid. In this problem, it permits easier movement of water and solutes. Crystallization of salts could produce larger fracture aperture variations thus leading to variations of permeability larger than three orders of magnitude according to model predictions (Figure 19). These permeability variations are
29
1.E-15 0 1 2 3
1.E-14
1.E-1
4 5 6 7 8 9 10
Permeabilidad (m2)
1.E-12 0 0.5 1 1.E-11 1.E-10 1.E-09 1.E-08
Permeability is a function of matrix permeability and disctontinuity aperture (cubic law). Research in this field is in progress including the possible explanation of the swelling by means of crystal formation in the anhydrite-gypsum system (Ramon et al 2009). Back to the environmental field, osmotic effects due to the presence of solutes in waste materials have been investigated by Mokni et al 2008. The problem of water attraction by salt crystals that takes place in bituminized waste requires a coupled formulation in which the activity of water is modified by the presence of salts thus generating a water attraction. These materials are to be disposed in clay formations and the water that may flow from the rock may induce chemo-hydro-mechanical processes such as swelling and permeability increase. Figure 20 shows the evolution of pore pressure development induced by presence of salts of a simulated sample which can swell under zero and 0.2 MPa of stress. The presence of crystals and the subsequent water pressure development induces the swelling that is plotted in figure 21.
0.35 0.3 0.25 Pressure (MPa)
Profundidad (m)
Free swelling Swelling under Syy = 0.2 MPa
1.5
0.2 0.15 0.1 0.05
Profundidad (m)
2 2.5 3 3.5 4 4.5
0 0.0001
0.001
0.01
0.1 Time (hour)
10
100
Figure 21. zone.

3.50
Fluid pressure evolution in the high concentration
Permeabilidad (m2)
displacement (mm)
3.00 2.50 2.00 1.50 1.00 0.50 0.00 -0.50 -1.00 0 10 20 30 40 50 time (hour) 60 70 80 90 100
Figure 20. Permeability increase in the rock below the floor due to exacavation and crystallization (after 2 years).
Free swelling Swelling under Syy = 0.2 MPa
simulated by means the same discontinuity embedded (considered in a diffuse way and without anisotropy permitted) that has been described in Olivella and Alonso (2008) and used for GMT modeling (see above). Here fracture opening is generated by dilatancy induced by shear or by rock expansion induced by crystallization.
Figure 22. Induced displacement on top of the sample (for a sample with 20 mm height).
30
This numerical example concerns a simulation of an oedometric swelling test with a bituminized waste with membrane efficiency of 0.75 containing NaNO3 with porosity and permeability of 0.2 and k 1.1019 m2. It is assumed that within the bituminized waste, salt crystals are dissolved, so that the medium is considered to be saturated with NaNO3 brine with an initial mass fraction of w 0.47 (high concentration), corresponding roughly to NaNO3 solubility. The material is put in contact with pure water. These simulations show the model response, calibration of the model with experimental results from laboratory tests carried out at SCK is under way. Simulation is extremely useful to define new experimental work to understand the complex behaviour of such material. The problem of salts at high concentrations in an unsaturated porous media requires to take into account a number of questions such us: osmotic and ultrafiltration fluxes, semipermeable behavior of the medium, changes in permeability and diffusivity, double structure effects combined to osmotic processes. Mechanical response of clays under the effects of osmotic and capillary suctions has been investigated to a limited extend. 6 CONCLUSIONS
ACKNOWLEDGMENTS This paper shows a number of research works performed at UPC, all the contributors including professors, researchers and doctorate students are acknowledged. REFERENCES
Acero P., C. Ayora, M.W. Saaltink, S. Olivella, J. Carrera, 2008. Modeling the interaction between geochemical and hydraulic processes during evaporation in tailings. Canada.Geochimica et Cosmochimica Acta, 72, A3-A3. Alonso, E.E., Gens, A. and Josa, A. (1990). A constitutive model for partially saturated soils, Geotechnique, Vol. 40, No. 3, pp. 405430. Alonso E, Vaunat J, Gens A. (1999). Modelling the mechanical behaviour of expansive clays. Engineering Geology 54:173183. Alonso, E., S. Olivella, N-M. Pinyol (2005). A review of Beliche dam. Gotechnique, 55(4): 267285. Alonso E.E. and C. Hoffmann (2007). Modelling the field behaviour of a granular expansive barrier. Physics and Chemistry of the Earth 32, 850865. Alonso E.E., S. Olivella, E. Romero, (2008). EB in situ test: experimental and numerical investigations, Progress Report, Department of Geotechnical Engineering and Geosciences, UPC. Alonso E.E., Romero, E. and S. Olivella, (2009). Estudio Balsas de Infiltracin. Report for ACDLL. Arnedo D., E.E. Alonso, S. Olivella, E. Romero (2008). Gas injection tests on sand/bentonite mixtures in the laboratory. Experimental results and numerical modelling, Physics and Chemistry of the Earth 33, S237S247. Barbour, S.L. and Yang, N., (1993). A review of the influence of clay-brine interactions on the geotechnical properties of ca-montmorillonitic clay soils from Western Canada. Canadian Geotechnical Journal, 30, 920934. Berdugo, I.R., (2007). Tunnelling in sulfate bearing rocksexpansive phenomena. Doctoral Thesis. Universidad Politcnica de Catalua. Di Maio, C., (1996). Exposure of bentonite to salt solution: osmotic and mechanical effects. Geotechnique 46 (4), 695707. Ehlers, W., T. Graff, M. Ammann (2004). Deformation and localization analysis of partially saturated soil. Comput. Methods Appl. Mech. Engng., 193: 28852910. Gawin, D, F. Pesavento, B.A. Schrefler, (2003). Modelling of hygro-thermal behaviour of concrete at high temperature with thermo-chemical and mechanical material degradation, Computer Methods in Applied Mechanics and Engineering, Volume 192, Issues 1314, 28 Pages 17311771. Gens, A., M. Snchez, L. Do, N. Guimares, E.E. Alonso, A. Lloret, S. Olivella, M.V. Villar, F. Huertas, (2009). A full-scale in situ heating test for high-level nuclear waste disposal: observations, analysis and interpretation, Gotechnique, Volume 59, Issue 4, pages 377399.
The applications presented in this paper demonstrated that coupled THM problems of a wide variety are solved at present. The modeling is of great help to understand the processes, to make predictions in short and long term, to modify engineering designs, to study the safety of waste disposal. Advances are required in different directions for instance to investigate in detail the couplings between chemical and thermo-hidromechanical problems at different levels (mechanical constitutive model, hydraulic model, transport properties, osmotic effects, etc). This may help to understand real problems and to help to improve designs. Another field of research is the gas migration and transport in soils and rocks. This is motivated by various problems; one of them is the gas generation induced by corrosion in problems of waste storage. This problem is numerically highly demanding and non-continuous. Probably different numerical techniques are required which is actually related to strain localization and therefore using alternative numerical methods would help to solve different difficulties within the THMC coupled problem. Finally the THMC problems with highly nonlinear calculations in three dimensions will be performed and normally time steps are small
31
Gran, M., J. Carrera, S. Olivella, J. Massana, M. Saaltink, C. Ayora and A. Lloret, (2005). Effects of salinity on multiphase flow using open soil columns under evaporation conditions. European Geosciences Union EGU General Assembly 2005, Viena, Austria, 2529 April. Guimaraes, L.M., A. Gens and S. Olivella, (2007). Coupled Thermo-Hydro-Mechanical and Chemical Analysis of Expansive Clay Subjected to Heating and Hydration, Transport in Porous Media, (2007) 66:341372. Harrington, J.F. & Horseman, S.T. (2003). Gas migration in KBS-3 buffer bentonite: Sensitivity of test parameters to experimental boundary conditions, SKB Technical Report TR 03-02. Stockholm:SKB. Hueckel T., 1997. Chemo-plasticity of clays subjected to stress and flow of a single contaminant. International journal for numerical and analytical methods in geomechanics, 21: 4372. Lloret A. and E.E. Alonso (1980). Consolidation of unsaturated soils including swelling and collapse, Geotechnique, 30, 4, 449477. Li, X., Z. Liu & R.W. Lewis (2005). Mixed finite element method for coupled thermo-hydromechanical process in poro-elasto-plastic media at large strains. Int. J. Numer. Meth. Engng., 64: 667708. Liu, Z, N. Boukpeti, X. Li, F. Collin, J.-P. Radu, T. Hueckel and R. Charlier, (2005). Modelling chemohydro-mechanical behaviour of unsaturated clays: a feasibility study, Int. J. Numer. Anal. Meth. Geomech., 29:919940. Mokni, N., Olivella, S., Li, X., Smets, S., and Valcke, E.: Deformation induced by dissolution of salts in porous media J. Phys. Chem. Earth. (2008). doi: 10.1016/j. pce.2008.10.008. Nishimura, S., A. Gens, S. Olivella, R.J. Jardine, (2009). THM-Coupled Finite Element Analysis of Frozen Soil: Formulation and Application. Geotechnique. Oldecop, L.A. & Alonso, E.E. (2001). A model for rockfill compressibility. Gotechnique 51, No. 2, 127139. Oldecop, L. and Alonso, E.E. (2003). Suction effects on rockfill compressibility. Gotechnique. 53(2): 289292. Olivella, S., Carrera, J., Gens, A. & Alonso, E.E. (1994). Non-isothermal multiphase flow of brine and gas through saline media. Transport in Porous Media, 15, No. 3, 271293. Olivella S. and A. Gens (2002). A constitutive model for crushed salt, Int. Journal for Numerical and Analytical Methods in Geomechanics. 26: 719746.
Olivella S. and Alonso, E.E. (2008). Gas flow through clay barriers, Geotechnique, Geotechnique 58, No. 3, 157176. Plaza, A. (2008). Anlisis del comportamiento de un tnel excavado en rocas altamente expansivas, Graduation Thesis at ETSECCPB, UPC. Perzyna, P. (1966). Fundamental problems in viscoplasticity. Advances in Applied Mechanics, Academic Press, New York, 9, 244368. Rummel, F. & Weber, U. (2000). GS experiment: Rock-mechanical testing on gasfrac self-healing effect, Mont Terri Project TN99-36. Pinyol, N.M., E.E. Alonso, and S. Olivella (2008). Rapid drawdown in slopes and embankments, Water Resour. Res., 44. Pruess, K.: 1987, TOUGH Users Guide, Lawrence Berkeley Laboratory. Ramon, A., S. Olivella and E.E. Alonso (2009). Swelling of a gypsiferous claystone and its modeling. Proceedings of the 17th Internacional Conference on Soil Mechanics and Geotechnical Engineering, Alexandrie, M. Hamza et al (Eds), Vol. 1, p. 730733. Rummel, F. & Weber, U. (2000). GS experiment: Rockmechanical testing on gasfrac self-healing effect, Mont Terri Project, TN9936. Saaltink, M.W., F. Batlle, C. Ayora, J. Carrera and S. Olivella. (2004). RETRASO, a code for modeling reactive transport in saturated and unsaturated porous media, Geologica Acta Vol. 2, N. 3, 2004, 235251. Sanchez, M., A. Gens, L.D. Guimaraes and S. Olivella, (2005). A double structure generalized plasticity model for expansive materials,, Int. Journal for Numerical and Analytical Methods in Geomechanics, 29, 751787. Sanchez, M. A. Gens, L. Guimaraes, S. Olivella, (2008). Implementation algorithm of a generalized plasticity model for swelling clays, Computers and Geotechnics, 35, 860871. Vilarrasa, V., D. Bolster, M. Dentz, S. Olivella, J. Carrera, (2010). Effects of CO2 Compressibility on CO2 Storage in Deep Saline Aquifers, Transp Porous Med, Published on line. Zandarin, M.T E Alonso, S Olivella (2010). Influence of suction on the mechanical behaviour of rock joints, UNSAT.
32
Energy geotechnology: Implications of mixed fluid conditions

J. Carlos Santamarina & Jaewon Jang
School of Civil and Environmental Engineering, Georgia Institute of Technology, Atlanta, GA, USA
ABSTRACT: Multi phase fluids are common in energy-related geotechnical problems, including gaswater, gas-oil, ice-water, hydrate-water, and oil-water fluid conditions. The generalization of classical unsaturated soil mechanics concepts to energy geotechnology requires physical understanding of surface tension, contact angle, capillary pressure, solubility and nucleation. Eventually, these pore-level processes affect the granular skeleton. Together, pore and particle-scale interactions upscale through the sediment structure to affect its macroscale response. Possible emergent phenomena include fluid percolation, residual saturation and recovery efficiency; fluid driven fractures, lenses, fingering and pipe formation; bubble migration and bottom blow up.
INTRODUCTION
Table 1. Mixed geotechnology.
fluid
conditions
in
energy
Energy geotechnology involves geotechnical phenomena and processes related to energy, from resource recovery to infrastructure and waste management. Energy resources include fossil fuels (90% of all primary sourcescoal, petroleum, and gas), nuclear, hydroelectric, and other renewable sources (wind, geothermal, solar, tidal, biomas). The most critical energy-related waste storages include: CO2 geological storage (from fossil fuels), fly ash (from coal), nuclear waste, and coal-mining waste. Resource recovery, energy infrastructure and waste management often involve multi-phase fluid conditions (Table 1classical infrastructure related conditions are not addressed in this manuscript). The most relevant cases are: L-G: water-air, water-CO2 and water-methane interfaces (as well as other biogenic and thermogenic gases). The liquid L has molecules of the gas in solution, and the gas contains molecules of the liquid. L1-L2: water-liquid CO2 (geological C-storage), and water-oil (petroleum reservoirs). Both liquids include molecules of the other liquid in solution. L-I: water-ice and water-hydrate. Related analyses can often be interpreted as the solid ice or hydrate phase behaving as a high viscosity fluid. The purpose of this manuscript is to extend fundamental concepts in unsaturated soil behavior to address mixed-fluid conditions in energy geotechnology. First, we explore interfacial processes at the atomic scale; then, we identify emergent phenomena that affect field-scale applications. Concepts
Fossil fuels (oil, coal, gas. Unconventional: coal-bed methane, shale-gas, tight-gas sandstone, CH4 hydrates) Recovery water(l), oil(l) CO2(g), CO2(l) CH4(g), CH4(h) water(l), oil(l) CO2 storage CO2(g), CO2(l), CO2(h) CH4(h) Nuclear Spent fuel storage Renewable: solar, wind, tidal Compressed air storage Renewable: bio, geothermal Production air/vapor(g), water(l) air/vapor(g), water(l) steam(g), water(l)
Note: Mixed fluid conditions in infrastructure are not listed. Phases shown in parenthesis (g: gas, l: liquid, h: hydrate).
presented in the following section capture the essential characteristics; references are provided for detailed information. 2 ATOMIC-SCALE PHENOMENA
Geotechnical implications of mixed-fluid conditions arise from interactions at the atomic scale where surface tension and contact angle are defined. Interfaces are in a state of dynamic equilibrium: molecules are continuously jumping from one phase to the other. The average residing time for a
33
molecule at the surface of a liquid is in the order of 106 s (Shaw 1992). Interfacial phenomena are understood at the atomic level. The van der Waals force is the result of Coulombian interactions (Fig. 1a) between atoms or molecules that experience instantaneous polarization. A dipole experiences orientation polarization next to an ion (Fig. 1b); this situation is extended to the mutually induced rotation of two interacting dipoles (KeesomFig. 1c). An atom experiences displacement of the electron cloud relative to the nucleus in the vicinity of a charge or ion (Fig. 1d). Furthermore, a dipole may polarize an otherwise non polar molecule (DebyeFig. 1e), and two non polar molecules may become mutually polarized because the instantaneous position of electrons confers an atom or molecule with some instantaneous polarity even when the timeaveraged polarity is zero (London dispersion Fig. 1f). The resultant Coulombian force for the instantaneous charge configuration shows mutual attraction in all these cases. The van der Waals force includes Debye, Keesom and London contributions (Fig. 1c, 1e, & 1f). The atomic analysis of van der Waals interactions helps explain surface tension and contact angle. 2.1 Surface tension
(a) low Pg
(b) high Pg
Figure 2. Effect of gas pressure on the development of surface tension along a gas-liquid interface. (a) Low gas pressure. (b) High gas pressure.
The time-average van der Waals attraction is isotropic when the molecule is away from the boundary. However, induced polarizations become anisotropic for molecules along the boundary and higher van der Waals attraction develops along the interface (Fig. 2). This situation alters the molecular organization in fluids near the interfaces. The altered molecular arrangement extends for about 5-to-10 mono-layers away from the interface. For example, water molecules at the water-vapor interface prefer to be oriented with their negative side towards the vapor phase, while random bulk con-
ditions are attained at a distance of 12 nm from the interface (Butt et al. 2006). Molecular dynamics simulations show the preferential alignment of water molecules near interface ions (Bhatt et al. 2004) and of water and CO2 molecules at the interface (da Rocha et al. 2001, Kuznetsova & Kvamme 2002, Kvamme et al. 2007). These molecular-scale phenomena cause the emergence of a contractile membrane along the interface and a measureable surface tension Ts [mN/m]. This situation applies to L-G, L1-L2 and L-I interfaces. The effect of dissolved species. Foreign species modify the electrical field within the liquid and alter the interfacial tension. Solutes may either be repelled by the interface or attracted to it (van Oss et al. 2002). Variations in interfacial tension Ts [mN/m] with solute concentration c [mol/L] are anticipated in terms of surface excess of solute [mol/m2] (Butt et al. 2006, Pegram & Record 2007, Tuckermann 2007), Ts RT c T c (1)
(a) Ion - Ion (Coulomb)
(d) Ion - Polarized Molecule
(b) Ion - Dipole
(e) Dipole - Polarized Molecule (Debye)
(c) Dipole - Dipole (Keesom)
(f) Two Polarized Molecule (London)
Figure 1. Electrical forces between ions and molecules. The van der Waals force combines Debye, Keesom and London contributions (Santamarina et al. 2001).
where [dimensionless] is the solute activity coefficient and T [K] is temperature. Ions are depleted at the interface 0 in inorganic solutions, but there is enrichment of organic species 0 at the interface when organic compounds are present. In the case of water-CO2, there is high concentration of dissolved CO2 near the interface, 0, causing a drop in interfacial tension (Chun & Wilkinson 1995, Massoudi & King 1974, Sutjiadi-Sia et al. 2008). The effect of temperature. The balance between thermal activity and molecular forces determines the evolution of surface tension with temperature. In general, surface tension decreases with increasing temperature.
34
The effect of pressure. fluid density. Consider a liquid-gas LG interface. The proximity to and the number of near-neighbor charges depends on gas density. Hence, higher interaction and lower interfacial tension is expected with increasing gas pressure and density (Sugden-Macleod equation Ts f() (Chun & Wilkinson 1995)). Likewise, the interaction with the external fluid remains relatively constant once the pressure exceeds the vapor-liquid boundary. Data for water-CO2 in Figure 3 shows the pronounced sensitivity of interfacial tension with pressure and the relatively constant Ts values when pressure exceeds the CO2 L-V boundary. Critical point. A distinct separation between a liquid and its vapor ceases to exist and surface tension vanishes at P-T conditions above the critical point (Bauer & Patel 2009). The critical point for water is at 22 MPa and 647 K, and for CO2 at and 7.4 MPa and 304 K. Supercritical fluid conditions develop above this pressure and temperature. Values. Figure 3 and Table 2 show interfacial tension data for mixed fluid conditions relevant to energy geotechnology, including water-vapor, water-CO2, water-CH4, water-oil, water-ice, and water-hydrate. 2.2 Contact angle
Table 2.
Interfacial tension and contact angle. Interfacial tension [mN/m] Contact angle []
Water-CO2 (g) 72a (0.1 MPa, 298 K) 38a on calcite (0.1 MPa, 298 K) Water-CO2 (l) 30a (7 MPa, 298 K) 30a on Calcite (7 MPa, 298 K) Water-CH4 (g) 72b (0.1 MPa, 298 K) 105a on PTFE 64b (10 MPa, 298 K) (0.1 MPa, 298 K) Water-Oil 33c H2O-benzene 98180d on 49c H2O-mineral oil mineral (308366 K) Water-Ice 32e 0f (water on ice) Water-CH4 (h) 39g no data found 32e no data found Water-CO2 (h) 30e a-Espinoza & Santamarina 2010, b-Ren et al. 2000, c-Kim & Burgess 2001, d-Treiber et al. 1972, e-Anderson et al. 2003, f-Knight 1971, g-Uchida et al. 1999.
vapor-solid TVS, liquid-solid TLS, and liquid-vapor TLV tensions. The contact angle is computed from the sum of forces parallel to the solid surface (Fig. 4; Adamson & Gast 1997) cos TVS TLS TLV Youngs Equation (2)
The contact angle formed by two adjacent fluids resting on a mineral substrate reflects the mutual interactions between the three neighboring phases. Contact angle and interfacial tension. Fluid-fluid and fluid-solid interfaces attempt to shrink because of the corresponding interfacial tensions. The contact angle reflects the equilibrium between these forces. In particular, if a liquid is surrounded by its own vapor, the relevant surface tensions are the
100 Interfacial tension [mN/m] 80 60 40 20 0 0
CO2 L-V boundary

at 295K Gaseous CO2 at 298K Liquid CO2
H2O - PTFE H2O - PTFE Brine - PTFE Brine - PTFE H2O - oil-wet quartz H2O - oil-wet quartz H2O - oil-wet quartz Experimental fitting (a) Exp. data points (b) Exp. data points (c) Exp. data points (d)
Experimental error 2.5 mN/m
10 Pressure [MPa]
15
20
Figure 3. Interfacial tension between water and CO2. Lines indicate values reported in the literature for deionized water at 298 K. Collected data and measured values reported in Espinoza & Santamarina (2010).
The crystal structure of minerals explains the different affinity for fluids. For example, when crystallization takes place in an aqueous environment, groups with water affinity develop on the surface, and the contact angle that water forms on the crystal is lower than the angle observed in the same mineral crystallized in air (Shaw 1992). Oil-wet and water-wet mineral surfaces. The assessment of wettability from contact angle measurements is depicted in Figure 4. By convention, the contact angle is measured with respect to water. Usually, mineral surfaces in soils and rocks are hydrophilic and water-wet conditions prevail. The oil-wet condition may result when a water-repellent agent (for example, silicone) is absorbed onto the mineral surface. Chemisorption of organic matter on the mineral surfaces may also create a hydrophobic surface. However, since mineral surfaces are hydrophilic in nature, the oil-wet condition can be considered a temporary condition. Thus, given enough time, organic fluids tend to be displaced by water. In short-time processes, wettability has a significant effect on multiphase flow, such as during enhanced oil recovery (Morrow 1990). Hysteresis. The effects of temperature, pressure, chemical composition, and pH on interfacial tension and wetting of solid surfaces can be
35
equilibrium analysis (Fig. 6). The CO2-substrate interfacial tension is also affected by CO2 pressure, and combines with changes in TCO2-H2O to fully explain results in Figure 5. Values. Table 2 and Figure 5 provide contact angle data for mixed fluid conditions relevant to energy geotechnology, including water-vapor, water-CO2, water-CH4, water-oil, water-ice, and water-hydrate. 2.3 Capillarity Consider a water droplet surrounded by its vapor (Fig. 7a). The contractile interfacial membrane compresses the droplet increasing its internal pressure. The free body diagram shows that the difference between the internal liquid pressure and the external vapor pressure u uL-uV is related to the liquid-vapor interfacial tension TLV as u 2TLV/R (Fig. 7a). More generally, the pressure difference between the two interacting fluids is related to the curvature of the interface: 1 1 u TLV Young-Laplace R1 R2 (3)
Figure 4. WettabilityInterfacial tension and contact angle. Ranges for water-wet, neutrally wet, or oil-wet surfaces. Note that water-wet or oil-wet conditions are established by testing the substrate with the two fluids present at the same time (from Francisca et al. 2003).
investigated by means of contact angle measurements (see techniques in Shaw 1992, Dullien 1992, Kwok & Neumann 1999). The apparent simplicity in the determination of contact angle is misleading. In particular, there is hysteresis in the contact angle with respect to the direction of fluid motion. There are two possible causes for contact angle hysteresis (Adamson & Gast 1997, Extrand 1998). First, there is the effect of surface roughness, whereby the macroscopically observed angle differs from the contact angle at the scale of asperities, which is the scale relevant to Youngs equation. Second, the contact angle is affected by surface heterogeneity due to the alteration of surface properties by fluids and the presence of surfactants or surface-active agents. Pressure effects. Changes in interfacial tensions TVS, TLS, and TLV (in particular) with pressure will alter the contact angle, particularly in liquid-gasmineral systems such as water-CO2-substrate and water-CH4-substrate systems. For example, an increase in pressure from P 0.1 MPa to 8 MPa causes an increase in contact angle of 45 (water on hydrophobized glass pressurized by CO2), of 50 (Water droplet on Teflon-PTFE pressurized by CO2), of 25 (CO2 droplet on muscovite mica pressurized by brine), and of 60 (water on coal pressurized by CO2) (Dickson et al. 2006, Siemons et al. 2006, Chiquet et al. 2007, Chi et al. 1988). Additional data are shown in Figure 5 for water-CO2-mineral systems. Note that contact angle decreases on hydrophilic surfaces but increases on hydrophobic surfaces when TLV decreases with increasing CO2 pressure, as anticipated from force
where R1 and R2 are the two principal radii of curvature. The same formulation permits computing
CO2 L-V boundary
at 295K Gaseous CO2 at 298K Liquid CO2 PTFE
160 140 120 Contact Angle [] 100 80 60 40
CO2-wet substrates Oil-wet quartz H2O-wet substrates
Calcite
20 0 0 5 10 Pressure [MPa] 15 20
Quartz
Figure 5. Contact angle evolution with pressure for a water droplet surrounded by CO2 and resting on hydrophobic substrates (oil-wet quartz and PTFE) and hydrophilic substrates (quartz and calcite). Continuous lines: deionized water; dashed lines: brine (from Espinoza & Santamarina 2010).
36
(a)Non-wetting liquid
Liquid
(b)Wetting liquid
LV
LV VS LS
Liquid
systems is affected by temperature T (Clennel et al. 1999, Coussy 2005, Coussy & Monteiro 2007, Gens 2010): u m(Tm T) (6)
VS
LS
Mineral
Mineral
cos 180
LV
TVS TVS TLV

cos
TVS TLS TLV

LV
Figure 6. The effect of changes in interfacial tension on contact angle. The dotted line shows the droplet geometry when TLV decreases.
where Tm is the melting temperature and m is the melting entropy (for ice: Tm 273.15 K and m 1.2 MPa/K. We can combine these equations for the case of cylindrical tubes to obtain the size of pores invaded by the non-wetting phase at given relative humidity or temperature conditions: R 2 TLV M cos water-vapor 1 RT ln hr 2TLV cos ice or hydrate m Tm T (7)
Air (a) Water Droplet

TLV TLV
uV
uL
2R
(uL uV ) R2 = 2 RTLV
u =
TLV
2TLV R
R
(8)
(b)
2R
uV
uL
(uV uL ) R2 = 2 RTLV cos
u =
2TLV cos R
Figure 7. Fluid pressure and capillary phenomena as a function of interfacial tension and contact angle. (a) Water droplet. (b) Tube.
Equations 7 & 8 relate invaded pore size to either relative humidity or temperature, while Equations 3 & 4 relate pore size to capillary pressure. These equations together with pore structure (connectivity and spatial variability) define the characteristic curves u S for the following systems under consideration: gas-water uG uW, gas-oil uG uO, ice-water uI uW, hydrate-water uH uW, and oilwater, uO uW. 2.4 SolubilitySaturation
the pressure difference between gas-water uG uW, gas-oil uG uO, ice-water uI uW, hydrate-water uH uW, and oil-water, uO uW. Equation 3 anticipates that the pressure difference u can be large when small pores are involved. Implications on the development of discontinuities are explored later in this manuscript in the context of fractures and lenses. In a cylindrical pore, the pressure difference between the wetting and the non-wetting fluids is (Fig. 7b) u 2TLV cos R (4)
Phase A may dissolve or come out of solution from a neighboring phase B depending on pressure and temperature. During this process, interfaces may appear or disappear. When solubility limits for A-in-B and B-in-A are reached, an equal number of molecules from each specie travel from one phase to the other. P-T dependent solubility. The P-T dependent concentration of a certain specie in another phase MP,T [mol/m3] can be approximated using a linear function of pressure H 1 1 0 M P ,T Papplied kH exp R T T 298.15 K (9)
The vapor pressure is affected by the curvature of the liquid-vapor interface. Relative humidity and capillary pressure relate as (Defay & Prigogine 1966): u 1 RT ln M hr (5)
where T is absolute temperature, R is the universal gas constant R 8.31 N m/mol K, M is the molecular weight of the liquid, and is the liquid density. Equilibrium in water-ice and water-hydrate
The enthalpy of the solution is H 14130 [J/ mol] for CH4 in water, and H 19940 [J/mol] for CO2 in water (Wilhelm et al. 1977). Typical values of Henrys constant for different gas species are listed in Table 3. Hence, the solubility of gas increases with increasing pressure and decreasing temperature, indicating an increased preference for dissolved gas rather than a separate phase at high pressure.
37
Table 3. Henrys law constants (data from Wilhelm et al. 1977). Gas Carbon dioxide Methane Hydrogen Nitrogen Oxygen Sulfur dioxide CO2 CH4 H2 N2 O2 SO2 Henrys constant kH0 [M/atm] 3.4 102 1.4 103 7.8 104 6.5 104 1.3 103 1.2
Table 4. Measured values of supersaturation needed to cause bubble nucleation in aqueous solutions. Compiled from Lubetkin (2003). Gas Carbon dioxide Methane Hydrogen Nitrogen CO2 CH4 H2 N2 Measured supersaturation 4.6220 80 8090 19140
Competing solutes. Solubility is affected by the emergence of new phases or the presence of competing solutes. For example, the presence of hydrates facilitates further hydrate formation and the equilibrium concentration of gas in water decreases in the presence of hydrate (Waite et al. 2009). Salt is a competing solute for dissolved gas and lowers the solubility of gas in water (Davie et al. 2004, Sun & Duan 2007, Tishchenko et al. 2005, Zatsepina & Buffett 1998). Size-dependent solubility. Smaller water droplets in vapor, hydrate crystals in water, or gas bubbles in water have higher pressure, i.e., chemical potentials, and require higher concentration (water or methane) in the surrounding fluid to balance that chemical potential (Henry et al. 1999, Kwon et al. 2008). This process is significant in small pores (e.g., smaller than 40 nm for hydrate in seawater (Sun & Duan 2007). Supersaturation. Solutions are often supersaturated at prevailing pressure and temperature conditions. The degree of supersaturation is defined as:
cactual 1 cequilib
(10)
Experimentally determined supersaturation values for different gases in water are listed in Table 4. The degree of supersaturation decreases if the solute nucleates as a separate phase and in the presence of certain impurities that facilitate (rather than hinder) nucleation. 2.5 Nucleation
Gas bubble formation, salt precipitation and hydrate nucleation phenomena are end-conditions to solubility. Critical nuclei size and heterogeneous nucleation play a critical role in the formation of a separate phase. Heterogeneous nucleation. Supersaturation thresholds for homogeneous nucleation in the bulk liquid are a function of molecular interactions between the liquid and the dissolved gas. The
presence of impurities and surface imperfections may facilitate nucleation and lower the degree of supersaturation. For example, the presence of mineral surfaces tends to favor heterogeneous bubble nucleation at substantially lower supersaturations. Nucleation centers in sediments include microcavities, irregularities and impurities at mineral surfaces (Dominguez et al. 2000, Blander 1979, Gerth & Hemmingsen 1980, Pease & Blinks 1947). The heterogeneous nucleation of hydrate and ice on mineral surfaces is prompted by reduced thermal activity and spatial distribution of water molecules in the vicinity of mineral surfaces. However, ice or hydrate cages do necessarily form against the mineral surface as hydrogen bonding is needed to form the crystals. We can observe that: (1) the boundary layer may extend to few monolayers; (2) water molecules are not in a disordered liquid form, (3) water molecules have limited mobility, (4) diffusion through this layer is necessarily lower than in the bulk fluid, and (5) the hydrate (or ice) and mineral interface may sustain a tensile strength. Critical nuclei size. Molecules continuously jump from one phase to the other, e.g., gas molecules jump from the bubble into the liquid and vice versa (numerical simulations in Walsh et al. 2009). The effect of size on solubility implies that molecules could jump out of small embryos faster than they could jump back into them. There is a critical nuclei size d* for which nuclei are in equilibrium with the solution so that nuclei smaller than d* will tend to dissolve (Nvlt et al. 1985, Finkelstein & Tamir 1985, La Mer 1952, Ward et al. 1970, Lubetkin 2003). Typically, stable nuclei are several nanometers in size and d* decreases with supersaturation. Ostwald ripening. The higher saturation predicted around smaller gas bubbles or crystalline nuclei promote diffusion from small nuclei towards large ones. Therefore a large bubble or crystal will tend to grow at the expense of smaller neighboring nuclei. This process is called Ostwald ripening. Ripening is a diffusion-controlled aging process and is important in fast crystallization processes that produce small crystal sizes. Ostwald ripening
38
will alter the crystal size distribution with time (Myerson 2002). Gas nucleation. Gas molecules occupy cavities between water molecules or between organic molecules in oil, until the fluid reaches the supersaturation threshold that prompts bubble nucleation (Ronen et al. 1989). Spontaneous bubble nucleation can result from: (1) depressurization of a pure liquid below the vapor pressure, (2) temperature increase for a pure liquid until the vapor becomes more stable than the pure liquid, (3) by gas coming out of solution from a supersaturated liquid (e.g., due to depressurization as per Henrys law; biogenic gas at constant pressure), or (4) by hydrate dissociation when P-T conditions migrate outside the stability field (Hemmingsen 1975 & 1977, Lubetkin 2003, Rebata-Landa & Santamarina 2010, Santamarina & Jang 2009). Ice and hydrate nucleation. Ice forms at low temperature when the thermal activity of water molecules decreases to allow a solid structure. The presence of molecules such as CH4 or CO2 helps the nucleation of water cages around these gas molecules. The resulting crystalline structure is known as gas hydrate. Typical structures are shown in Figure 8a. The P-T phase boundaries for ice, CO2 hydrate and CH4 hydrate are shown in Figure 8b and c. Ice and hydrates share similar properties. There are, however, two important differences. First, the phase boundary for ice is at almost constant temperature independently of pressure, while the phase boundary for CH4 and CO2 hydrates is pressure and temperature dependent. Differences in P-T phase boundaries define the location of permafrost (cold regions) versus the location of hydrates (offshore & below deep permafrost, i.e., high pressure & low temperature). Second, water expands during ice formation; conversely, ice contracts as it melts (8.3% volume contraction). The case of hydrates is more complex: the initial volume of water Vw increases to Vhyd 1.26Vw when methane hydrates form; upon dissociation, hydrates form water and gas phases and the initial volume of hydrate Vhyd expands to occupy a volume Vhyd where is P-T dependent: for P 10 MPa and T 288.15 K, 2.26, the water occupies Vw 0.79Vhyd and the rest is the volume of gas Vg 1.47Vhyd. These pronounced volume expansion upon dissociation and the formation of separate gas and liquid phases anticipate significant implications of dissociation. Gases and non-wetting fluids invade pores by displacing water, starting at large pores and gradually advancing into smaller pores, according to the interconnected pore structure. Ice and hydrate also invade the pore space along the larger-size interconnected porosity, but do so by crystallizing in pores.
(b) 10
Pressure [MPa]
8 6 4 2
Ice Ice Ice + Hydrate Hydrate + Gas Hydrate + Gas Water + Gas
Water + Gas
0 263
268
273 278 Temperature [K]
283
288
(c) 10
Pressure [MPa]
8 6 4 2 0 263
CO 2 hydrate
268
273
278
283
288
Temperature [K]
Figure 8. Ice and gas hydrate. (a) Molecular structure (figures from Sloan & Koh 2008 and Heriot-Watt University. (b) Phase boundary for methane hydrate. (c) Liquidvapor and hydrate phase boundaries for carbon dioxide.
PORE AND PARTICLE-SCALE ANALYSES
Interfacial tension, contact angle, solution / dissolution and formation / dissociation processes occur at the pore scale and affect the sediment response through particle-scale interactions. Related concepts and phenomena are analyzed next. 3.1 Pores and pore throats SizeStatistics. The evolution of mixed-fluid phenomena in porous/granular media is intimately related to pore size distribution and spatial correlation. The mean pore size (dp) can be analytically estimated as a function of the void ratio e, the specific surface Ss [m2/g] and the mineral density [g/cm3]
dp k

e Ss
(11)
39
where the shape factor is k 5 for spherical particles, k 2 for dispersed clays and k 4 for edge-toface configuration. The void ratio depends on the effective stress as e e1kPa Cc log 1kPa (12)
allow us to study fluid flow even under mixed fluid flow conditions. On the other hand, pore-networks are better suited for the study of unsaturation; in this case, each pore is connected to its neighboring pores through throats that satisfy geometric constraints relative to the log-normal distributed pores. A comparative summary is presented in Figure 9. 3.2 Particle sizeParticle level forces
where e1 kPa is the void ratio at 1 kPa and Cc the compressibility coefficient of the sediment. We have compiled published pore size distribution data for clays, silts, and sands. Results show that pore size distribution statistics (mean and standard deviation) increase with particle size. When data are fitted with a log normal distribution, the standard deviation of ln(dp/[m]) is bound between (Phadnis & Santamarina 2010) dp ln 0.4 0.2 m (13)
Finally, we note that pores are connected through pore throats. The size of pore throats is related to the size of the two connected pores; assuming a simple cubic configuration, dth 2 1 min d p1, d p 2 3 1 ~ 0.6 min d p1, d p 2
(14)
Maximum capillary pressure. The pressure difference between the non-wetting and wetting phases umax is limited by the largest pore throat along the boundary between wetting and non-wetting fluids umax unw uw 4Ts max dth (15)
Sediments are particulate materials. Particle displacement, hence the macroscale sediment mechanical response, is determined by the forces particles experience. These include forces generated at the boundary and transmitted through the skeleton (i.e., due to effective stress), forces that result from the particle volume (weight and buoyancy), forces that develop along the particle surface (hydrodynamic and capillary) and contact level forces (electrical and cementation-reactive). Asymptotic, order-of-magnitude estimates of the forces most relevant to this manuscript are summarized in Table 5. The normal contact force N related to the applied effective stress scales with the square of the particle diameter N d2 (assuming a simple cubic packing of equal size spheres). Weight and buoyancy combine to determine the submerged weight of a particle of unit weight m submerged in a fluid of unit weight f, S (mf)d3/6. The capillary force a particle experiences when the sediment is subjected to mixed fluid conditions can be estimated using sketches shown in Figure 10 (Note: this analysis applies to vapor-water, gas-oil, oil-water, ice-water and hydrate-water mixtures). There are two force components. One is exerted by the interfacial membrane in the direction of the
Tube-network Pore-network Water
This pressure difference can develop in gas, oil, ice or hydrate phases in water-saturated sediments. Upscalingnetwork models. Pore-scale phenomena upscale through the interconnected pore space. The interconnected porosity can be captured using tube-only, pore-only, and pore tube network models. Tube-network models are used to compute flow conditions, assuming Poiseuille flow within each tube
Gas
Node Tube Tubes are connected at nodes
Pore Throat Pores are connected at throats
R 4 P q tube 8 L
qc0
2T cos u s Rtube
Pg Pw Pc
u
2Tscos Rth
(16)
where the flow rate through a tube q [m3/s] is a function of fluid viscosity [N s/m2], tube radius Rtube [m], tube length L [m], and pressure difference between end nodes P [N/m2]. These models
Figure 9. Networks of tube and of pores used for mixed fluid studies (details in Jang and Santamarina 2010a&bSee also Blunt 2001 for a comprehensive review of network models).
40
Table 5. Particle-level forces. Skeletal Weight Buoyant Hydrodynamic Capillary Electrical Attraction Electrical Repulsion Cementation N d2 W Gswd3/6 U d3/6 Fd 3vd Fc 2Tsd Att Ahd/24t2 Rep 0.0024c00.5exp(108tc00.5)d T tentd
important phenomena that take place in sediments in the context of energy geotechnology. 4.1 Fluid invasionMixed fluid conditions
Fluid invasion under mixed fluid conditions is controlled by capillary Fc, viscous drag Fd and weight/ bouyancy W or Fb forces. These forces can be combined into dimensionless numbers (Pennel et al. 1996, Lenormand et al. 1988Note: the defending and invading fluid viscosities are def and inv): NC NM Fd 3 d invv 3invv Fc dTs cos Ts cos (18) (19)
Re.: Cho et al (2001), Santamarina (2001); Gili & Alonso 2002.

unw
uw
inv def
Fd 3 d invv 18invv 3 W Gs w d 6 Gs w d 2 Fb d 3 6 d 2 Fc dTs cos 6Ts cos
Ts
NG
(20)
unwuw
Figure 10. Forces on particles under mixed fluid flow condition (Shin & Santamarina 2010b).
NB
(21)
wetting fluid, as the interfacial membrane clings to the mineral around the grain surface. The second component is caused by the pressure difference between the wetting and the non-wetting phase against the cross section of the grain. Then, the capillary force acting on the grain and transmitted to the sediment skeleton is (compressive towards the side of the wetting fluid). Fcap d Ts cos
Invasion conditions and emergent phenomena can be identified in the dimensionless space of these -ratios. Figure 11 shows a 2D-slice of this space. Conditions relevant to energy geotechnology are identified next: Water invades a dry sediment: slow, capillary driven invasion by a wetting fluid follows the connectivity among the smallest pores.
log NC
d2 unw uw 4 d 2 Ts d Ts cos max dth 2 d Ts ( for uniform grain size )
(17)
Viscous fingering
-4
Stable displacement
Upscalinggrain models. Particle-scale forces determine the equilibrium of each particle and its eventual displacement. Sediment-scale implications can be explored using particle models, which are based on Newtonian mechanics. The resultant force F acting on a given particle of mass m causes it to accelerate with a F/m. After a time interval t, the particle is at a new location x v0tat2/2 and experiences a new set of forces. 4 MACRO-SCALE PHENOMENA: MIXED FLUID FLOW
-8
-4
log NM
-4
-8
Capillary fingering
Pore-scale and particle-scale conditions identified in previous sections are used herein to explore
Figure 11. Space for viscous fingering, capillary fingering, and stable displacement in terms of dimensionless ratios NM and NCrefer to text (modified from Lenormand et al. 1988).
41
Oil invades a water-saturated sediment: slow invasion by a non-wetting high viscosity fluid takes place by overcoming capillary resistance, i.e., invades largest pores first. Gas storage into a water saturated sediment and gas production during hydrate dissociation: invasion by a non-wetting low viscosity fluid. Liquid CO2 is injected into a water saturated sediment: the non-wetting liquid CO2 has a viscosity two orders of magnitude smaller than water (Jung et al. 2010). Viscous fingering may develop (high Nc and low Nm in Figure 11see Qi et al. 2009). Ice and hydrate growth: it resembles the slow invasion of a high viscosity non-wetting phase. 4.2 Breakthrough pressure The long-term storage of CO2 is a quasi-static condition controlled by capillary forces at pore throats. Similar conditions apply to gas and oil storage. The distribution of pore size discussed in Section 3.1 is invoked in this analysis. For a given pore structure, the breakthrough pressure pc* determined by the pressure-dependent interfacial tension Ts and contact angle is given by
* pc
4Ts cos d* p
(22)
where dp* is the minimum pore diameter along a percolating path across the seal layer. The pore size dp* can be related to the mean by a factor of the standard deviation. For a log-normally distributed pore size, d* Rp dp p ln ln ln nm nm m (23)
The analysis of gas breakthrough data in Hildenbrand et al. (2002), in Hildenbrand et al. (2004) and in Horseman et al. (1999) leads us to conclude that percolating paths are made of pores larger than the mean, and that the second term in Equation 23 is relatively independent of effective stress (Espinoza & Santamarina 2010). We note that this analysis does not take into consideration poro-mechanical effects (pore size is affected by the percolating immiscible fluid and the ensuing capillary forcesSection 4.6) and reactive fluid transport (which also affects the pore size along the percolating pathSection 4.9). 4.3 Gas invasion -vs.- internal gas nucleation
Expressions for degree of saturation and permeability as a function of capillary pressure have been
derived for typical drying soil conditions where water vapor is in a continuous phase with the atmosphere. These u-S expressions can be used to investigate similar conditions in energy geotechnology, such as the injection of liquid CO2 into watersaturated sediments. However, gas comes out of solution and bubbles grow within the sediment in various energy-related applications, from oil recovery to methane production from hydrate bearing sediments (as well as frequent seepage conditions downstream of earth dams). These two cases are referred to internal gas drive process for the case of nucleation and gas liberation during depressurization, and external gas drive process or gas injection for the case of forced invasion (Yortsos & Parlar 1989, Poulsen et al. 2001, Nyre et al. 2008). We use 3D network models to explore the effect of gas invasion versus nucleation on the evolution of the characteristic saturation curves and relative permeabilities (details in Jang & Santamarina 2010c). We assume slow invasion and nucleation so that the process is controlled by capillary forces both viscous and gravity forces are disregarded (refer to Section 4.1). Gas invasion is enforced from a peripheral gas-water interface, while gas nucleation is initiated at randomly selected internal nodes; the algorithm and tube size distribution are identical in both invasion and nucleation simulations. Results are described in terms of the defending fluid retention and relative permeabilities. Characteristic curve in random media. The characteristic curve captures the causal link between water saturation and capillary pressure (See also Wilkinson & Willemsen 1983, Lenormand & Zarcone 1984). Pore throat size distribution and spatial correlation govern the shape of the characteristic curve (Francisca & Arduino 2007). The lower bound characteristic curve is determined by sorting pores and gradually invading from the largest pore to the smallest one; conversely, the upper bound is obtained by invading the smallest pore first and blasting through the sediment as gradually larger pores are invaded. Real cases fall in between these two extremes. The characteristic curves for spatially-uncorrelated random-distributed pores are shown in Figure 12a (Simulation details are indicated in the figure caption). Results are almost identical for gas invasion and gas nucleation. Water has been displaced in most tubes when the capillary pressure corresponds to the mean pore size, which in this case is u 0.14 MPa for (Rtube) 1m. Relative permeabilities. Gas and water conductivity during gas invasion and nucleation are calculated at each saturation. Computed water conductivities are normalized by the water conductivity of the fully water saturated network (Figure 12b). Gas conductivities are normalized by the gas conductivity obtained when gas invasion
42
0.30 Capillary pressure [MPa] 0.25 0.20 0.15 0.10 0.05 0.00 0.0 0.2 0.4 0.6 0.8 1.0 Saturation [ ]
initial oil saturation Soi and the residual oil saturation Sor as EO
(a)
Soi Sor Soi
(24)
1.0 Relative permeabilities [ ] 0.8 0.6 0.4 0.2 0.0 0.0 0.2 0.4 0.6 0.8 1.0 Saturation [ ]
(b)
k rg
k rw
Residual oil saturation after water flooding ranges from 10% to 40%, depending on the sediment heterogeneity and flooding conditions (Morrow 1990). Gas hydrates. Internal gas nucleation and gas expansion creates conditions that are different to oil extraction. We use pore-network simulations to investigate gas recovery efficiency and residual gas saturation. An analytical result can also be obtained for gas recovery efficiency EG using macro-scale analyses (details in Jang & Santamarina 2010b). EG Vgrec Vgdis
1.0 0.8 0.6 0.4 0.2 0.0 1 2 3 4
Gas recovery efficiency
Figure 12. Characteristic curves and relative permeabilitiesGas invasion vs. gas nucleation. (a) Characteristic curves. Symbols: ( ) gas invasion through 132 nodes on one side, () gas invasion into the network sorted by tube size; gas invasion through multiple nodes distributed inside network model: ( ) 132, () 2 132, () 3 132 nodes. (b) Relative conductivity of water krw and gas krg. Results obtained using a three dimensional tube-network model. Details: 13 13 13 nodes, 5460 tubes, coordination number cn 6, log-normal distribution of tube radius R, the mean tube size (R) 1m, and the standard deviation in tube radius (ln(R/[m])) 0.4. Parameters for Young equation Pc 2TLVcos/R: TLV 72 mN/m, 0.
Vgdis Vgres Vgdis
Vhini 0.79Vhini Vgres Vhini 0.79Vhini

res (0.21Sh Sg 1) (25)
Shyd 40% 30% 20% 15% 10%
res Shyd 0.79 Sg
0.79
(a) (Rp)=1m
5%
process is completed. The normalized water conductivities are almost same for both gas invasion and nucleation. However, gas conductivity during nucleation is much lower than during gas invasion. Similar conclusions were reached by Poulsen et al. (2001) using different model characteristics. Other implications. There are important porogranular mechanical distinctions between gas invasion and nucleation conditions. First, we observe that isolated bubbles can exist at different gas. Second, a capillary front evolving at the sediment boundary (e.g., during desiccating) tends to cause sediment compaction, while internal nucleation promotes expansion. These two cases are discussed later in this section, in the context of desiccation cracks and bubble migration. 4.4 Residual saturationrecovery efficiency
Expansion factor
10
Residual gas saturation
0.5 0.4 0.3 0.2
40%
d
Shy
30% 20% 15% 5%
0.1 10% 0.0 1
(b) (Rp)=1m
2 3 4
Expansion factor
10%
10
Isolated gas saturation
0.5 0.4 0.3 0.2 0.1 0.0
(c) (Rp)=1m
30% 20% 15%
Shyd=5%
40%
1 2 3 4 5 6 7 8 9 10
Expansion factor
Oil recovery efficiency is controlled by pore connectivity and water invasion. The recovery of gas from hydrate bearing sediments depends on volume expansion upon dissociation. Oil recovery. The displacement efficiency EO during oil recovery can be defined in terms of the
Figure 13. Gas production from hydrate bearing sediments. (a) Recovery efficiency. (b) Residual gas saturation. (c) Isolated gas saturation. Each point is an average of 20 realizations. Three-dimensional pore-network model: 15 15 15 pores. Randomly distributed pore radius with constant mean (Rp) 1m and standard deviation [ln(Rp)] 0.4. Pore throat Rth 0.5min(Rp1, Rp2). Periodic boundary condition is used. Parameters for Young equation Pc 2TLVcos/R: TLV 72 mN/m, 0 (from Jang & Santamarina 2010b).
43
where the Vhini is initial hydrate volume and Vgres is residual gas saturation per unit volume of sediment, Shyd is initial hydrate saturation, represents the expansion of dissociated gas and water from initial volume of hydrate. Numerical and analytical results presented in Figure 13. Results show that gas recovery efficiency and residual gas saturation during hydrate dissociation are functions of initial hydrate saturation, mean pore size, and the variability in pore size distribution. 5 MACRO-SCALE PHENOMENA: MECHANICAL EFFECTS
u0 unsat. ub H L sat.
The ratio Fc/N 2Ts/(d) refer to Table 5- defines two extreme regimes: mixed fluid conditions do not affect the mechanical behavior of the sediment if Fc/N1, conversely capillary phenomena will have a profound effect on the sediment mechanical response when Fc/N1 (Fig. 14). Clearly, this analysis can be extended to include other particle level forces, such as cementation and electrical attraction. Salient phenomena are discussed next. 5.1 Seepage and gas nucleation: bottom blow up Bubble nucleation during depressurization and associated changes in hydraulic conductivity may combine to create unwanted failure conditions (we acknowledge studies by Japanese researchers in the early 1990s). Consider upward flowing liquid through a soil plug of length H. The liquid pressure at the lower inlet is ui, and uo at the upper outlet. The liquid pressure when gas comes out of solution is the bubble point ub (Note: the liquid is not gas saturated at pressure ui). Lets assume that equilibrium has been reached so that the continuity of water flow is maintained throughout the saturated and unsaturated soil layers qw(sat) qw(unsat), and that the hydraulic conductivities are related as kunsat ksat (Fig. 15). Then, the elevation L where gas comes out of solution can be computed from the following
ui
Figure 15. Gas dissolution, changes in permeability and the potential development of glow-up conditions.
quadratic equation in L (the explicit solution for L is not shown here) qw

sat
qw
unsat
ui ub ub u L L o H H L L
(26)
The hydraulic gradient is higher in the unsaturated zone, and uplift may cause blow up. The effective stress becomes zero at the elevation of the bubble point L when ui ub L tot (1 ) L (1 ) 5.2 Development of discontinuitiesgranular mechanics effects (27)
0.1 1.0
Localizations: s s 10
Fc N
2Ts 'd
Figure 14. Capillary force vs. skeletal force. Localizations develop at high Fc/N ratio.
Desiccation cracksImmiscible fluid cracks. The fundamental particle-level mechanism for desiccation crack initiation and growth is centered on the air-water interface membrane. Initially the air-water interface membrane resists invading pores, capillary suction increases, the effective stress increases, the soil consolidates, and the skeletal stiffness increases. Eventually, the air-water interface membrane invades the largest pores and causes particle displacement away from the invasion point. The void ratio increases at the tip and facilitates further membrane invasion and crack growth (Fig. 10 & 16). The same mechanism is
44
2
N100 kPa
Force
mN N nN
Fc
Figure 16. Surface defects and crack initiation. Local void ratio evolution during desiccation (Shin & Santamarina 2010a).
mm Diameter d
Figure 17. Skeletal force, capillary force, and weight of particles of different size (from Santamarina 2001).
responsible for fluid-driven fractures when other immiscible fluids are involved, such as gas-oil, gaswater, and oil-water systems. Ice and hydrate lenses. The formation of segregated ice and hydrate lenses responds to similar force-balance conditions and the granular-mechanics effects at the tip: lens growth causes an increase in porosity at the tip, fluid migrate toward the tip and further nucleation is favored. The growth of ice lenses is temperature controlled, and they typically develop parallel to the soil surface in frozen ground due to the advancing cold front, that is normal to the 1-direction (see Gens 2010). However, hydrate growth is gas-limited in most cases; in other words, sediments are already within the P-T stability field and hydrate grows as methane becomes available. In this case, hydrate lens growth is facilitated normal to the minor principal stress 3 (evidence in Yun 2005). Localization regime. As the particle size decreases, the particle weight and the skeletal force decrease faster than either the capillary force and the force exerted by the ice or hydrate lens. Hence, fine-grained soils are more susceptible to fluiddriven fractures and lens formation (Fig. 17see also Jain & Juanes 2009). 5.3 Buoyancy and bubble migration
water gw. If the capillary pressure at the bottom of the bubble is ubottom, the capillary pressure at the top is equal to utop ubottom Lw (Fig. 18). On the other hand, the effective stress increases with depth, (z) (z L) and both porosity and pore size decrease with depth (Eq. 12). Large capillary pressure and pore size at shallower depths explain the upwards propagate of bubbles The analysis can be extended to other immiscible buoyant fluids. Other observations are anticipated: (1) migrating bubbles coalesce with other bubbles found along their trajectory; coalescence and lower water pressure towards the surface lead to marked bubble expansion as the bubble approaches the sediment surface; (2) thinner bubbles will develop in sediments of higher stiffness; (3) long tubular bubbles cause higher capillary pressure for the same gas volume and may favor propagation in strong sediments; (4) a continuous bubble may not connect the source to the free sediment surface: discontinuous bubbles are needed to build up gas pressure. Therefore, gas bubble migration through sediments is inherently intermittent. 5.4 Monitoring: bulk stiffness
Bubbles larger than the critical size of nuclei d d* can coalesce to form even larger bubbles. These large bubbles become trapped at pore throats defining Laplacian capillary surfaces. As the bubble size increases, so does the buoyancy drive and its tendency to migrate upwards (experimental evidence in Boudreau et al. 2009). Buoyancy drive applies to other light and immiscible fluids. Consider a gas bubble that extends from depth z to depth z L, composed of a gas of unit weight much lower than the unit weight of
The monitoring of energy related processes is often conducted with P-wave propagation techniques. The stiffness of small bubbles has to be given special attention. Bubbles much smaller than soil particles can fit within the pore space without distorting the soil structure. Thus, the presence of gas bubbles only changes the compressibility of the pore fluid (Wheeler 1988). Even relatively small size bubbles are sufficient to significantly change the pore fluid bulk stiffness (Sparks 1963, Santamarina et al. 2001). The pore fluid bulk stiffness f depends on the
45
u top u bottom L w
g
w
z+L
g w
u bottom
Figure 18. Elongated gas bubble subjected to different capillary pressure in saturated sediments.
degree of saturation S, the bulk stiffness of water w (2.2GPa) and the bulk stiffness of gas bubbles b. The bulk stiffness of bubbles is intimately related to the gas pressure, which is governed by the surface tension and the bubble size db. We can show that (Rebata-Landa & Santamarina 2010):
b
4Ts 3 db
(28)
Then, the fluid bulk stiffness of water becomes:
f
1 3d 1 S 1 S b w 4Ts
(29)
Finally, the bulk stiffness of sediments in the presence of disseminated gas bubbles is:
soil sk
1 1 3d 1 n n S 1 S b 4Ts g w
(30)
albeit the low saturation concentration. Mineral dissolution by acidic water flow is most relevant to long-term geological storage of CO2. Similar situations apply to other energy geotechnology systems including hydrate dissolution by unsaturated water flow. Three potential emergent phenomena are discussed next. Dissolution pipes (Hydro-chemical coupling). A positive feedback condition develops between dissolution and preferential fluid flow paths. Eventually, high conductivity flow channels or wormholes may emerge (Hoefner & Fogler 1988; Fredd & Fogler 1998). Damkhlers number Da dissolution rate / advective transport rate, and Peclets number Pe advective transport rate / diffusion transport rate are used to explore the phenomenon: wormholes form at high Pe 1 and Da 1 numbers (Golfier et al. 2002). Change in capillary forces (mixed-fluid skeleton coupling). Pore size enlargement during mineral dissolution or pore size reduction during precipitation alters capillary forces when immiscible fluids are involved. Change in internal skeletal stresses (chemo-mechanical coupling). Mineral dissolution produces a pronounced drop in horizontal effective stress under zero lateral strain conditions. The state of stress decreases from a ko-state and reaches the active shear failure ka-state after 5% mineral dissolution (Shin & Santamarina 2009). Furthermore, shear localization may take place during mineral dissolution in sediments that exhibit postpeak strain softening behavior (in agreement with Rudnicki & Rice 1975Details in Shin et al. 2008).
In laboratory and field applications, the global bulk stiffness of the soil soil and its skeletal stiffnes sk can be inferred from S-wave velocity measurements. 5.5 Reactive fluid transport
CONCLUSIONS
Fluids react with the mineral walls and alter the sediment structure. Consider carbon dioxide: CO2 dissolves in water to form carbonic acid H2CO3, which eventually ionizes into H and CO32 lowering the pH of water. Minerals dissolve in acid water with dissolution rates that are proportional to the pH difference with respect to the buffer pH. The solubility of CO2 increases with pressure (Stumm & Morgan 1996). Conversely, a reduction in pressure produces the nucleation of CO2 gas and the precipitation of dissolved minerals. Therefore, both dissolution and re-precipitation can take place in the sediment. Mineral dissolution in the pore fluid satisfies similar concepts to those discussed in Section 2.4,
Problems related to energy geotechnology often involve mixed-fluid conditions. However, contrary to near-surface unsaturated soil mechanics, most energy geotechnology problems are at high fluid pressure and in deep sediments. This leads to phenomena seldom encountered in the more classical unsaturated soil literature. Processes reviewed in this manuscript are relevant to oil and gas, hydrate bearing sediments and carbon geological storage. The most important observations follow. Surface tension is pressure-dependent. Contact angle is determined by the interfacial tension of neighboring phases, thus, it is pressure dependent as well. Capillary phenomena can be extended to include ice and hydrates. While the phase boundary for ice is almost constant with pressure, the stability field for hydrates is pressure and temperature dependent.
46
Solubility is pressure and temperature dependent, and it is affected by the presence of other phase or solutes. Supersaturation facilitates nucleation. The higher saturation around smaller nuclei leads to diffusion-controlled ripening. Ice formation and melting is accompanied by volume change and affects the granular skeleton. Volume change is much more pronounced when gas hydrates are involved. The nucleation of a gas phase in pores reduces the sediment bulk stiffness, alters the hydraulic conductivity, and may lead to a zero effective stress condition, as observed in bottom blowup situations. Capillary pressure is a function of interfacial tension and curvature, and is related to relative humidity (water-vapor) or temperature (ice or hydrate). Hence, the size of pores invaded by the non-wetting phase can be related to relative humidity (vapor) or temperature (ice and hydrate). Pore size distribution and spatial connectivity govern fluid flow, percolation, break-through pressure, residual saturation and resource recovery efficiency. The percolating path for an immiscible fluid is made up of pores within the upper-side tail of the pore size distribution. The characteristic saturation curves u-S show no significant difference for gas invasion and gas nucleation, and there is almost no difference for liquid relative permeability in these two cases. However, the gas relative permeability is lower for internal gas nucleation compared to gas invasion. Particle size distribution determines pore size and particle-level forces, which eventually define the sediment response. Forces are of gravitational, skeletal (related to effective stress), viscous, capillary, and electrical origin. These forces can be combined into dimensionless ratios to facilitate the identification of physical regimes and emergent phenomena. Pore size changes as immiscible fluids cause particle-level capillary forces and particle displacement. Immiscible fluid-driven fractures, desiccation cracks, and gas bubble migration reflect these particle-scale mechanisms. The invading fluid may react with minerals. Reactive fluid transport alters the sediment structure, induces localized flow, forms high conductivity flow channel, and changes skeletal stresses. Storativity and the performance of seal layers is limited by these emergent phenomena. ACKNOWLEDGEMENTS Support for this research has been provided by the Department of Energy, the National Science Foundation and the Goizueta Foundation. We
are grateful to A. Sivaram for proof-reading the manuscript. NOTATION c Cc d db dp d* dp* dth Da e EG EO Fcap kH0 M MP,T N n Pc p c* Pe q r R Rtube S Sgres Shyd Soi Sor T TLV TLV Tm Ts TVS uL uV Vgdis Vgrec Vgres Vhini f g m w H Solute concentration Compressibility coefficient Particle diameter Bubble diameter Pore diameter Critical nuclei size Minimum pore diameter Throat diameter Damkhlers number Void ratio Gas recovery efficiency Oil recovery efficiency Capillary force Henrys constants Molecular weight Solubility Normal contact force Porosity Capillary pressure Breakthrough pressure Peclets number Flow rate Gas bubble radius Gas constant R 8.31 N m/ (mol K) Tube radius Saturation Residual gas saturation Hydrate saturation Initial oil saturation Residual oil saturation Temperature Interfacial tension liquid-vapor Interfacial tension liquid-vapor Melting temperature Interfacial tension Interfacial tension vapor-solid Liquid pressure Vapor pressure Dissociated gas volume Recovered gas volume Residual gas volume Initial hydrate volume Fluid expansion factor Solute activity coefficient Unit weight of fluid Unit weight of gas Unit weight of mineral Unit weight of water Degree of supersaturation Enthalpy of solution Contact angle [mol/L] [] [m] [m] [m] [m] [m] [m] [] [] [] [N] [M/atm] [g/mol] [mol/m3] [N] [] [N/m2] [N/m2] [m3/s] [m] [m] [] [] [] [] [] [K] [mN/m] [mN/m] [K] [mN/m] [mN/m] [N/m2] [N/m2] [m3] [m3] [m3] [m3] [] [] [N/m3] [N/m3] [N/m3] [N/m3] [] [J/mol] []
47
b f g sk soil w m
Bulk stiffness of gas bubble Bulk stiffness of pore fluid Bulk stiffness of soil particle Sulk stiffness of soil skeletal Sulk stiffness of soil Bulk stiffness of water Surface excess of solute Viscosity Density Effective stress Melting entropy
[GPa] [GPa] [GPa] [GPa] [GPa] [GPa] [mol/m2] [N s/m2] [kg/m3] [N/m2] [MPa/K]
REFERENCES
Adamson, A.W., & Gast, A.P. 1997. Physical chemistry of surfaces. 6th Ed., New York: Wiley. Anderson, R., Llamedo, M., Tohidi, B., & Burgass, R.W. 2003. Experimental measurement of methane and carbon dioxide clathrate hydrate equilibria in mesoporous silica, J. of physical chemistry B. 107: 35073514. Bauer, B.A., & Patel, S. 2009. Properties of water along the liquid-vapor coexistence curve via molecular dynamics simulations using the polarizable TIP4PQDP-LJ water model. The J. of chemical physics. 131: 084709-1:16 Bhatt, D., Newman, J., & Radke, C.J. 2004. Molecular dynamics simulations of surface tensions of aqueous electrolytic solutions. J. of physical chemistry B. 108: 90779084. Blander, M. 1979. Bubble nucleation in liquids. Advances in colloid and interface science. 10: 132. Blunt, M.J. 2001. Flow in porous mediapore-network models and multiphase flow. Current opinion in colloid & interface science. 6: 197207. Boudreau, B.P., Algar, C., Johnson, B.D., Croudace, I., Reed, A., Furukawa, Y., Dorgan, K.E., Jumars, P.A., Grader, A.S., & Gardiner, B.S. 2009. Bubble growth and rise in soft sediments, Geology, 33: 517520. Butt, H.-J., Graf, K., & Kappl, M. 2006. Physics and chemistry of interface. 2nd Ed. Weinheim: Wiley-VCH Verlag GmbH & Co. KGaA. Chi, S.-M., Morsi, B.I., Klinzing, G.E., & Chiang, S.-H. 1988. Study of interfacial properties in the liquid CO2water-coal system. Energy & Fuels. 2: 141145. Chiquet, P., Broseta, D., & Thibeau, S. 2007. Wettability alteration of caprock minerals by carbon dioxide. Geofluids. 7: 112122. Cho, G.C., & Santamarina, J.C. 2001. Unsaturated particulate materialsParticle-level studies. J. of geotechnical and geoenvironmental engineering. 127: 8496. Chun, B.-S., & Wilkinson, G.T. 1995. Interfacial tension in high-pressure carbon dioxide mixtures. Industrial & engineering chemistry research. 34: 43714377. Clennell, M.B., Hovland, M., Booth, J.S., Henry, P., & Winters, W.J. 1999. Formation of natural gas hydrates in marine sediments 1. Conceptual model of gas hydrate growth conditioned by host sediment properties. J. of geophysical research. 104: 2298523003. Coussy, O. 2005. Poromechanics of freezing materials. J. of the mechanics and physics of solids, 53: 16891718.
Coussy, O., & Monteiro, P. 2007. Unsaturated poroelasticity for crystallization in pores. Computers and geotechnics, 34: 279290. da Rocha, S.R.P., Johnston, K.P., Westacott, R.E., & Rossky, P.J. 2001. Molecular structure of the water supercritical CO2 interface. J. of physical chemistry B. 105: 1209212104. Davie, M.K., Zatsepina, O.Y., & Buffett, B.A. 2004. Methane solubility in marine hydrate environments. Marine geology. 230: 177184. Defay, R., & Prigogine, I. 1966. Surface tension and adsorption. New York: John Wiley & Sons. Dickson, J.L., Gupta, G., Horozov, T.S., Binks, B.P., & Johnson, K.P. 2006. Wetting phenomena at the CO2/ water/glass interface. Langmuir. 22: 21612170. Dominguez, A., Bories, S., & Prat, M. 2000. Gas cluster growth by solute diffusion in porous media. Experiments and automation simulation on pore network. International J. of multiphase flow. 26: 19511979. Dullien, F.A.L. 1992, Porous media fluid transport and pore structure, 2nd Ed. New York: Academic. Espinoza, D.N., & Santamarina, J.C. 2010. Water-CO2mineral systems: interfacial tension, contact angle and diffusionImplications to CO2 geological storage. Water resources research. Accepted for publication. Extrand, C.W. 1998. A thermodynamic model for contact angle hysteresis. J. of colloid and interface science. 207: 1119. Finkelstein, Y., & Tamir, A. 1985. Formation of gas bubbles in supersaturated solutions of gases in water. American institute of chemical engineers J. 31: 14091419. Francisca, F.M., & Arduino, P. 2007. Immiscible displacement model for anisotropic and correlated porous media. International J. of geomechancis. 7: 311317. Francisca, F.M., Rinaldi, V.A., & Santamarina, J.C. 2003. Instability of hydrocarbon films over mineral surfaces: Microscale experimental studies. J. of environmental engineering. 120: 11201128. Fredd, C.N., & Fogler, H.S. 1998. Influence of transport and reaction on wormhole formation in porous media. American institute of chemical engineers J. 44: 19331949. Gens, A. 2010. Soil-environment interactions in geotechnical engineering, Gotechnique. 60: 374. Gerth, W.A., & Hemmingsen, E.A. 1980. Heterogeneous nucleation of bubbles at solid surfaces in gas-supersaturated aqueous solutions. J. of colloid and interface science. 74: 8089. Gili, J.A., & Alonso, E.E. 2002. Microstructural deformation mechanisms of unsaturated granular soils, International J. for numerical and analytical methods in geomechanics, 26: 433468. Golfier, F., Zarcone, C., Bazin, B., Lenormand, R., Lasseux, D., & Quintard, M. 2002. On the ability of a Darcy-scale model to capture wormhole formation during the dissolution of a porous medium, J. of fluid mechanics, 457: 213254. Hemmingsen, E.A. 1975. Cavitation in gas-supersaturated solutions. J. of applied physics. 46: 213218. Hemmingsen, F.A. 1977. Spontaneous formation of bubbles in gas-supersaturated water. Nature. 267: 141142.
48
Henry, P., Thomas, M., & Clennell, M.B. 1999. Formation of natural gas hydrates in marine sediments 2. Thermodynamic calculations of stability conditions in porous sediments. J. of geophysical research. 104: 2300523022. Hildenbrand, A., Schlomer, S., & Krooss, B.M. 2002. Gas breakthrough experiments on fine-grained sedimentary rocks. Geofluids. 2: 323. Hildernbrand, A., Schlomer, S., Krooss, B.M., & Littke, R. 2004. Gas breakthrough experiments on politic rocks: comparative study with N2, CO2 and CH4. Geofluids. 4: 6180. Hoefner, M.L., & Fogler, H.S. 1988. Pore evolution and channel formation during flow and reaction in porous media. American institute of chemical engineers J.. 34: 4554. Horseman, S.T., Harrington, J.F., & Sellin, P. 1999. Gas migration in clay barriers. Engineering geology. 54: 139149. Jain, A.K., & Juanes, R. 2009, Preferential Mode of gas invasion in sediments: Grain-scale mechanistic model of coupled multiphase fluid flow and sediment mechanics, J. of geophysical research, 114: B08101. Jang, J., & Santamarina, J.C. 2010a. Hydraulic conductivity in spatially varying mediaA pore-scale investigation. Geophysical J. international. Under review. Jang, J., & Santamarina, J.C. 2010b. Gas production from hydrate bearing sediments: Pore-network model simulation and macro-scale analyses. In preparation. Jang, J., & Santamarina, J.C. 2010c. Evolution of water and gas saturation and relative permeabilities during gas invasion and gas nucleation. In preparation. Jung, J.W., Espinoza, D.N., & Santamarina, J.C. 2010. Properties and phenomena relevant to CH4-CO2 replacement in hydrate bearing sediments. J. of geophysical research. Under review. Kim, H., & Burgess, D. 2001. Prediction of interfacial tension between oil mixtures and water. J. of colloid and interface science. 241: 509513. Knight, C.A. 1971. Experiments on the contact angle of water on ice. Philosophical magazine. 23: 153165. Kuznetsova, T., & Kvamme, B. 2002. Thermodynamic properties and interfacial tension of a model watercarbon dioxide system. Physical chemistry chemical physics. 4: 937941. Kvamme, B., Kuznetsova, T., Hebach, A., Oberhof, A., & Lunde, E. 2007. Measurements and modelling of interfacial tension for watercarbon dioxide systems at elevated pressure. Computational material science. 38: 506513. Kwok, D.Y. & Neumann, A.W. 1999. Contact angle measurement and contact angle interpretation. Advances in colloid and interface science. 81: 167249. Kwon, T.-H., Cho, G.-C., & Santamarina, J.C. 2008. Gas hydrate dissociation in sediments: Pressuretemperature evolution. Geochemistry Geophysics Geosystems. 9: Q03019, doi:10.1029/2007GC001920. La Mer, V.K. 1952. Nucleation in phase transitions. Industrial and engineering chemistry. 44: 12701277. Lenormand, R, touboul, E., & Zarcone, C. 1988. Numerical models and experiments on immiscible displacements in porous media. J. of fluid mechanics, 189: 165187.
Lenormand, R., & Zarcone, C. 1985. Invasion percolation in an etched network: measurement of a fractal dimension. Physical review letters. 54: 22262229. Lubetkin, S.D. 2003. Why is it much easier to nucleate gas bubbles than theory predicts? Langmuir. 19: 25752587. Massoudi, R., & King, A.D. 1974. Effect of pressure on the surface tension of water. Adsorption of low molecular weight gases on water at 25C. The J. of physical chemistry. 78: 22622266. Morrow, N.R. 1990. Wettability and its effect on oil recovery. J. of petroleum technology. 42: 14761484. Myerson, A.S. 2002. Handbook of industrial crystallization, 2nd Ed. Boston: Butterworth-Heinemann. Nyre, A.N., McDougall, S.R., & Skauge, A. 2008. Effect of depressurization on trapped saturations and fluid flow functions. Improved oil recovery symposium. SPE 113394. Tulsa, Oklahoma, 1923 April. Nvlt, J., Sohnel, O., Matuchova, M., & Broul, M. 1985. The kinetics of industrial crystallization. Amsterdam: Elsevier. Pease, D.C., & Blinks, L.R. 1947. Cavitation from solid surfaces in the absence of gas nuclei. J. of physical and colloid chemistry. 51: 556567. Pegram, L.M., & Record, M.T. 2007. Hofmeister salt effects on surface tension arise from partitioning of anions and cations between bulk water and the air-water interface. J. of physical chemistry B. 111: 54115417. Pennell, K.D., Pope, G.A., & Abriola, L.M. 1996. Influence of viscous and buoyancy forces on the mobilization of residual tetrachloroethylene during surfactant flushing. Environmental science technology, 30: 13281335. Phadnis, H.S., & Santamarina, J.C. 2010. Pore size limited lift in sediments. In preparation. Poulsen, S., McDougall, S.R., Sorbie, K.S., & Skauge, A. 2001. Network modeling of internal and external gas drive. Proceeding of the 2001 International symposium of the society of core analysts (SCA). Edinburgh. Sept. Qi, R., LaForce, T.C., & Blunt M.J. 2009. A three-phase four-component streamline-based simulator to study carbon dioxide storage, Computational Geosciences, 13(4), 493509. Rebata-Landa, V., & Santamarina, J.C. 2010. Biogenic gas bubbles in soils: Skemptons B parameter, P-wave velocity and liquefaction resistance. J. of geotechnical and geoenvironmental engineering. In print. Ren, Q.-Y., Chen, G.-J., Yan, W., & Guo, T.-M. 2000. Interfacial tension of (CO2CH4)water from 298 K to 373 K and pressures up to 30 MPa. J. of chemical and engineering data. 45: 610612. Ronen, D., Berkowitz, B., & Magaritz, M. 1989. The development and influence of gas bubbles in phreatic aquifers under natural flow conditions. Trans in porous media. 4: 295306. Rudnicki, J.W., & Rice, J.R. 1975. Conditions for the localization of deformation in pressure-sensitive dilatant materials. J. of the mechanics and physics of solids. 23: 371394. Santamarina, J.C. 2001. Soil behavior at the microscale: particle forces, Proceedings of symposium of soil behavior and soft ground construction, October, MIT.
49
Santamarina, J.C., & Jang, J. 2009. Gas production from hydrate bearing sediments: Geomechanical implication, NETL methane hydrate newsletter: Fire in the Ice. 9: 1822. (http://www.netl.doe.gov). Santamarina, J.C., Klein, K.A., & Fam, M.A. 2001. Soils and waves. London: John Wiley. Shaw, D.J. 1992. Introduction to colloid & surface chemistry, 4th Ed. London: Butterworth-Heinemann. Shin, H., & Santamarina, J.C. 2009. Mineral dissolution and the evolution of k0. J. of geotechnical and geoenvironmental engineering. 135: 11411147. Shin, H., & Santamarina, J.C. 2010a. Desiccation cracks in saturated fine-grained soils: particle level phenomena and effective stress analysis. Geotechnique. Under review. Shin, H., & Santamarina, J.C. 2010b. Fluid-driven fractures in uncemented sediments: Underlying particlelevel processes. Earth and plenatary science letters. Under review. Shin, H., Santamarina, J.C., & Cartwright, J.A. 2008. Contraction-driven shear failure in compacting uncemented sediments. Geology. 36: 931934. Siemons, N., Bruining, H., Castelijns, H., & Wolf, K.-H. 2006. Pressure dependence of the contact angle in a CO2-H2O-coal system. J colloid and interface science. 297: 755761. Sloan, E.D., & Koh, C.A. 2008. Clathrate hydrates of natural gases, 3rd Ed. Boca Raton: CRC Press. Sparks, A.D.W. 1963. Theoretical considerations of stress equations for partly saturated soils, Third regional conference for Africa on soil mechanics and foundation engineering. Salisbury. Rhodesia. 29 June. 215218. Stumm, W., & Morgan, J.J. 1996. Aquatic chemistry: chemical equilibria and rates in natural waters, 3rd Ed. New York: John Wiley & Sons. Sun, R., & Duan, Z. 2007. An accurate model to predict the thermodynamic stability of methane hydrate and methane solubility in marine environments. Chemical geology. 244: 248262. Sutjiadi-Sia, Y., Marckmann, H., Eggers, R., Holzknecht, C., & Kabelac, S. (2007), The effect of in liquids dissolved dense gases on interfacial and wetting characteristics, Forschung im Ingenieurwesen, 71, 2945. Sutjiadi-Sia, Y., Jaeger, P., & Eggers, R. 2008. Interfacial tension of solid materials against dense carbon dioxide. J. of colloid and interface science. 320: 268274. Tishchenko, P., Hensen, C., Wallmann, K., & Won, C.S. 2005. Calculation of the stability and solubility of methane hydrate in seawater. Chemical geology. 219: 3752.
Treiber, L.E., Archer, D.L., & Owens, W.W. 1972. A laboratory evaluation of the wettability of fifty oilproducing reservoirs. Society of petroleum engineers J.. 12: 531540. Tuckermann, R. 2007. Surface tension of aqueous solutions of water-soluble organic and inorganic compounds. Atmospheric environment. 41: 62656275. Uchida, T., Ebinuma, T., & Ishizaki, T. 1999. Dissociation condition measurements of methane hydrate in confined small pores of porous glass, J. of physical chemistry B. 103: 36593662 van Oss, C.J., Docoslis, A., & Giese, R.F. 2002. Role of the water-air interface in determining the surface tension of aqueous solutions of sugars, polysaccharides, proteins and surfactants, in Contact angle, wettability and adhesion, Volume 2, 312. Waite, W.F., Santamarina, J.C., Corte, D.D., Dugan, B., Espinoza, D.N., Germaine, J., Jang, J., Jung, J.W., Kneafsey, T.J., Shin, H., Soga, K., Winters, W.J., & Yun, T.-S. 2009. Physical properties of hydrate-bearing sediment. Reviews of geophysics. 47: RG4003, doi:10.1029/2008RG00279. Walsh, M.R., Koh, C.A., Sloan, E.D., Sum, A.K., & Wu, D.T. 2009. Microsecond simulations of spontaneous methane hydrate nucleation and growth. Science. 326: 10951098. Ward, C.A., Balakris, A., & Hooper, F.C. 1970. On thermodynamics of nucleation in weak gas-liquid solutions. J. of basic engineeringTransactions of the ASME. 92: 695704. Wheeler, S.J. 1988. A conceptual model for soils containing larger gas bubbles. Geotechnique. 38: 389397. Wilhelm, E., Battino, R., & Wilcock, R.J. 1977. Lowpressure solubility of gases in liquid water. Chemical reviews. 77: 219262. Wilkinson, D., & Willemsen, J.F. 1983. Invasion percolation: a new form of percolation theory. J. of physics A: Mathematical and general. 16: 33653376. Yortsos, Y.C., & Parlar, M. 1989. Phase change in binary systems in porous media: application to solution-gas drive. Society of petroleum engineers. SPE 19697. Yun, T.-S. 2005. Mechanical and thermal study of hydrate bearing sediments. Ph.D. thesis. Georgia Tech. Zatsepina, O.Y., & Buffett, B.A. 1998. Thermodynamic conditions for the stability of gas hydrate in the seafloor. J. of geophysical research. 103: 2412724139.
50
General reports
Geotechnical problems and case histories from the perspective of unsaturated soil mechanics
C.W.W. Ng & A.K. Leung
Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, HKSAR
ABSTRACT: This document consists of three major parts. In the first part, a total of eight and nine papers under the themes Geotechnical problems and Case histories are highlighted, respectively. In the second part, some state-of-the-art (SOA) findings revealing non-liquefied slide and liquefied flow mechanisms of man-made loose fill slopes are presented. These findings are based on a series of comprehensive triaxial tests on saturated and unsaturated loose decomposed geomaterials and centrifuge model tests conducted on loose and dense fill slopes. In the last part, some SOA findings on possible landslide triggering mechanisms of a natural hillslope are described and reported. These findings are obtained from laboratory and field tests and a comprehensive field monitoring programme of a natural saprolitic hillslope (decomposed tuff) in Hong Kong. The field monitoring programme is designed specifically to investigate the influence of two constitutive variables (i.e., matrix suction and net stress) on the natural hillslope. The responses of the hillslope subjected to the interaction between matrix suction and net stress resulting from seasonal climatic variations are explored and discussed in this paper.
INTRODUCTION
As the request of the Organisers, this document not only reports some selected papers grouped under Topic I: Geotechnical Problems and Case Histories, but it also provides two SOA reviews on man-made fill and natural slopes that the authors consider to be relevant and interesting. Therefore, this document is divided into three major parts. In the first part, based on the review of fifty five papers grouped under Topic I, some selected papers are highlighted. Then, some stateof-the-art findings from a specific geotechnical problemdeformation and failure mechanisms such as static liquefaction of loose fill slopes are described and discussed in the second part of this paper. In the final part, some state-of-the-art findings from a case history of field monitoring and testing in Hong Kong are reported and highlighted. Among the fifty five papers reviewed under Topic I (see Table 1), a total of eight and nine of them under the themes Geotechnical problems and Case histories respectively are selected to be reported and discussed in this paper. According to the technical contents, the former eight papers may be further categorised into two sub-themes, namely (i) Soil-atmosphere interaction and (ii) Flow and Infiltration. On the other hand, for the latter nine papers, it may be appropriate to similarly group
these under two sub-themes, namely, (i) Field monitoring and (ii) Engineering applications. Due to the page limit, this document does not intend to cover details of each selected paper but only highlight the main features and key-findings, and raise any queries and concerns of the reviewed papers as appropriate. In the second part (SOA reviews on geotechnical problems), the deformation and failure mechanisms such as static liquefaction of man-made loose fill slopes are explored and discussed. The explorations are based on centrifuge model tests conducted independently at the University of Cambridge and the Hong Kong University of Science and Technology. The fill materials used in the tests include Leighton Buzzard sand and completely decomposed granite (CDG) or silty sand. To assist in the interpretation of centrifuge test results, a series of comprehensive triaxial element tests are carried out on saturated and unsaturated CDG subjected to various stress paths. Key results from the element tests are highlighted and illustrated whether they are consistent with the centrifuge model tests or not. In the last part (SOA reviews on case histories), the field monitoring of an initially unsaturated saprolitic hillslope underlain by decomposed volcanic rocks in Lantau Island of Hong Kong is described and reported. The comprehensive monitoring programme is purposely designed to monitor
53
Table 1.
A summary of fifty five papers in the Topic I: Geotechnical problems and Case Histories. Author Country
Title of paper Theme: Geotechnical problems Sub-theme: Embankment and dams 1 Investigation of the water flow in an embankment built of organic soil: Experimental results and large scale field test 2 Numerical analysis of the behaviour of embankments constructed with expansive soils 3 Banking simulation with a constitutive model for unsaturated soil Sub-theme: Flow and infiltration 4 On the inverse problem solution for laboratory determination of unsaturated water hydraulic conductivities 5 Influences of spatially varying porosity on unsaturated flow through earth structures 6 Back-analysis of infiltration characteristic of a saprolitic hillslope by considering permeability heterogeneity 7 Study of stratification resistant to salt damage
Birle, E., Boso, M. & Heyer, D. Ortega, J.I., Lloret, A., Olivella, S. & Romero, E. Tanaka, M., Kawai, K., Kanazawa, S., Iizuka, A., Tachibana, S., Ohno, S. & Takeyama, T. Bicalho, K.V., Fleureau, J.M., Cui, Y.J. & Romanel, C. Le, T.M.H., Gallipoli, D., Wheeler, S. & Sanchez, M. Leung, A.K. & Ng, C.W.W. Nomura, S., Kawai, K., Kanazawa, S., Iizuka, A., Ohno, S., Thirapong, P. & Tachibana, S. Soltani, F. & Naeemifar, I. Trinh, V.N., Tang, A.M., Cui, Y.J., Dupla, J.C., Canou, J., Calon, N., Robinet, A., Lambert, L. & Schoen, O. Andr de Almeida, M., Miguel, M.G. & Teixeira, S.H.C. Bryant, J.T. & Haque, M.A. Mohamed, F.M.O., Vanapalli, S.K. & Saatcioglu, M. Schanz, T., Lins, Y. & Vanapalli, S.K. Stanier, S.A. & Tarantino, A.
Germany Spain Japan
Brazil/France UK/USA Hong Kong, China Japan
8 9
The study of permeability of incompressible fluids in infiltrating into the unsaturated compacted soil Unsaturated hydraulic properties of fine-grained soil from the blanket layer of old railway lines in France
Iran France
Sub-theme: Foundations 10 Pile horizontal bearing capacity in tropical soil under distinct moisture contents 11 12 13 14 15 16 Performance and design of foundations on unsaturated expansive soil Comparison of bearing capacity of unsaturated sand using cone penetration tests (CPT) and plate load tests (PLT) Bearing capacity of a strip footing on an unsaturated sand Active earth force in cohesionless unsaturated soils using bound theorems of plasticity A technique for estimating the shaft resistance of test piles in unsaturated soils
Brazil USA Canada Germany/ Canada UK
Interpretation of the bearing capacity of unsaturated soils extending the effective and the total stress approaches 17 Bearing capacity of an unsaturated sand from model footing tests Sub-theme: Geoenvironment problems 18 Coupled solute mass transport and evaporation in saline soil 19 Laboratory studies of the addition of bentonite to landfill cover coil 20 Evaluation of a capillary barrier at the rocky mountain arsenal
Vanapalli, S.K., Eigenbrod, K.D., Canada Taylan, Z.N., Catana, C., Oh, W.T., & Garven, E. Vanapalli, S.K. & Oh, W.T. Canada Vanapalli, S.K., Sun, R. & Li, X. Dunmola, A.S. & Simms, P.H. Izzo, R.L.S., Mahler, C.F. & Huse, K. Williams, L.O., Hoyt, D.L., Hargreaves, G.A., Dwyer. S.F. & Zornberg, J.G. Canada
Canada Brazil USA
(Continued )
54
Table 1. (Continued ) Sub-theme: Isolation barrier 21 Gas migration in sand/bentonite mixtures through preferential paths 22 An experimental root barrier to protect the foundations of a building from the desiccating effects of tree roots 23 Direct shear strength of compacted bentonite under different suctions Sub-theme: Pavements 24 Water migration in pavement subgrade 25 Hydromechanical behaviour of a natural sand using different suction controlled techniques 26 Unsaturated soil mechanics implementation during pavement construction quality assurance Sub-theme: Soil-atmosphere interaction 27 Influences of the soil-atmosphere interface interactions on the soil water content and temperature profiles 28 Soil heat and evaporation from soil 29 Shrinkage during wetting of fine-pored materials: does this accord with the principle of effective stress? 30 A prototype soil column to calibrate numerical models accounting for soilatmosphere interaction 31 Use of satellite observations for estimating the degree of saturation of superficial soils 32 An environmental chamber for studying the soil-atmosphere Interaction Sub-theme: Soil cover systems for waste containment 33 Hydro-mechanical response of an expansive silt-bentonite mixture 34 35 36 Geotechnical issues in the design and construction of PMOBs Water infiltration and methane emission through three different cover layers of an experimental municipal waste landfill at Muribeca, Recife, Pernambuco, Brazil Unsaturated hydraulic characteristics of soil with significant oversize particles Arnedo, D., Alonso, E.E., Olivella, S. & Romero, E. Blight, G.E. Nishimura, T., Rahardjo, H. & Koseki, J. Nelson, E.J. & Overton, D.D. Nowamooz, H. & Arsenie, M.I. Siekmeier, J. & Dot, M. France South Africa Japan/ Singapore USA France Not known
Bicalho, K.V., Vivacqua, G.P.D. & Cui, Y.J. Blight, G.E. Blight, G.E. Caruso, M. & Jommi, C.
Brazil/France South Africa South Africa Italy
Montrasio, L., Valentio, R. & Italy Quintavalla, C. Ta, A.N., Tang, A.M., Cui, Y.J. & France Thiriat, J. Ajdari, M., Nowamooz, H., Masrouri, F., Habibagahi, G. & Ghahramani, A. Cabral, A.R., Ltourneau, M., Yanful, E., Song, Q., McCartney, J.S. & Parks, J. Lopes, R.L., Alves, M.C.M. & Juc, J.F.T. Somasundaram, S., Shenthan, T., Benson, C. & Nannapaneni, S. France/Iran Canada/USA Brazil USA
Sub-theme: Slope stability 37 Anti-slide piles reinforcement of expansive soil landslide Cao, X. & Wei, J. 38 Influence of seasonable water pressure variations on Franch, F.A.J. & Futai, M.M unsaturated slopes stability 39 Slope stability problems in NorthEast part of Tehran, Iran Khanlari, G.R. & Hadi-mosleh, M. 40 The significance of curved shear strength envelope in Noor, M.J. Md. & Hadi, B.A. modelling rainfall induced shallow failure of slopes 41 Test site experience on mechanisms triggering mudflows in Pirone, M., Papa, R. & unsaturated pyroclastic soils in southern Italy Nicotera, M.V. 42 Performance of an instrumented slope under a capillary Rahardjo, H., Hua, C.J., barrier system Leong, E.C. & Santoso, V.A. 43 Analysis of influencing factors in strengthening expansive Yuan, J.P. & Wu, J.H. & soil slope with the coverage of geo-membrane Yin, Z.Z. Sub-theme: Thermal problems 44 Thermo-hydro-mechanical behaviour of an expansive bentonite/silt mixture using different suction controlled techniques Nowamooz, H., Gomez-Espina, R., Villar, M.V. & Masrouri, F.
China Brasil Iran Malaysia Italy Singapore China
France/Spain
55
(Continued )
Table 1.
(Continued ) Author Country
Title of paper Theme: Case histories Sub-theme: Engineering applications 45 Identification, characterization and analysis of deep foundation in collapsible soilregional prison of Eunpolis, Bahia/Brazil. 46 The effect of ripping clay soil on swell behaviour 47 48 49 50 Evaluation of landfill gas emission in experimental cover layers in Brazil Behaviour of wet fill using residual soil Improvement and controlling deformation of the expansive soil ground Failure of peat dykes by drought
Coutinho, R.Q., Castro, B.P.F. & Brazil Dourado, K.C.A. Fityus, S., Buzzi, O., Holt, M. & Gunther, T. Maciel, F.J., Lopes, R.L. & Juc, J.F.T. Marinho, F.A.M. & Massad, F. Miao, L., Wang, F. & Cui, Y.J. Nterekas, D., Oostveen, J.P., Mathijssen, F.A.J.M., Brinkgreve, R.B.J., Hicks, M.A. & Molenkamp, F. Oldecop, L.A., Garino, L., Muoz, L., Rodriguez, R. & Garca, C. Shibata, M. S, Kawai, K., Kanazawa, S., Iizuka, A., Tachibana, S., Ohno, S. & Honda, M. Australia Brazil So Paulo China Netherlands
51 52
Unsaturated behaviour of mine tailings in low precipitation areas Simulation of static compaction with unsaturated soil/water coupled F.E. analysis
Argentina/ Spain Japan
Sub-theme: Field monitoring 53 Suction changes during the construction of an embankment from A10 Motorway in Portugal 54 The use of filter paper method for field suction investigation 55 Matric suction in unsaturated pyroclastic slopes
Cardoso, R., Maranha das Portugal Neves, E. & Almeida, S.P. Nacinovic, M.G.G., Mahler, C.F., Brazil Izzo, R.L.S. & Brandt, P.F.S. Scotto di Santolo, A. & Italy Evangelista, A.
the responses of the hillslope due to changes of two constitutive variables (matrix suction and net stress) resulting from climatic variations. In addition to in situ infiltration tests using the Instantaneous Profile (IP) method, a series of laboratory experiments, with the help of some newly developed apparatus, are carried out to measure the stress-dependent soil-water characteristic curve (SDSWCC) and permeability function, k(), of the completely decomposed tuff (CDT) specimens. Field observations and laboratory experiment results are compared and discrepancies are highlighted and discussed. 2 2.1 REPORT ON SOME REVIEWED PAPERS Geotechnical problems
surfaces in detail. The balance equation is modified and is expressed as follows: LE (Rn Ro) WE (G H P) S (1)
2.1.1 Soil-atmosphere interaction Blight addresses and discusses three main difficulties in applying the conventional solar energy balance equation of measuring evaporation from soil
where LE is latent heat consumed in evaporation; Rn is net incoming vertical short wave radiation; Ro is outgoing long wave head radiation; WE is wind energy; G is energy expended in heating the nearsurface soil; H is energy consumed in heating the near-surface air; P is energy used in plant photosynthesis; and S is energy storage term. Unfortunately, the measurement method of each term is not clearly presented, probably due to the papers page limit. To verify the modified equation, a series of tests is carried out using water and partially saturated and fully saturated fine sand contained in plastic pots. Each test pot is exposed to the environment and the evaporation amount, soil and water temperature is recorded every 2 hours (see Fig. 1(a)). Figure 1(b) shows that evaporation continued to occur after sunset but at a reduced amount. Since there is no wind after sunset, the author concludes that the
56
Figure 2. Experimental setup of the large-scale environmental chamber (Ta et al.).

Air Soil t = 0 day t = 5 days t = 10 days t = 15 days t = 20 days t = 25 days t = 30 days 17 18 19 20 21 22 23 24
only driving component of evaporation is the S in equation (1). It is the solar energy absorbed by soil or water during sunlight, which then partly converts into evaporation after sunset. As shown in Fig. 1(c), it can be seen that under a given change of E, the decreasing rate of G is the greatest for the partially saturated sand. It indicates that the conversion of soil heat to evaporation becomes less efficient as the sand desaturates. This study improves the fundamental understanding of the interaction between soil and atmosphere from a theoretical point of view. Ta et al. implement a comprehensive instrumentation scheme on a large scale environmental chamber to study the effects of air conditions on evaporation rate in the laboratory. Figure 2 shows the experiment setup of the environmental chamber. Apart from measuring the air temperature, relative humidity and wind speed at the soil surface, volumetric water content (VWC), suction and soil temperature at various depths along the soil column are monitored. The development of cracks upon the 30-day drying test is also recorded using a high definition camera with a specific lens. Figure 3 depicts the variation of temperature profiles during the 30-day drying period. Under the influence of external temperature, the temperature of the soil specimen increases uniformly by about
Figure 3. Variation of temperature profiles during the 30-day drying period (Ta et al.).
2C. Based on the recorded flow rate, temperature and relative humidity of the air, the actual evaporation rate (mm/day) and accumulated amount of evaporation (mm) are calculated. The heavily instrumented soil column and its comprehensive measurements provide valuable data to analyse the mechanisms involved in the soil-atmosphere interaction. Moreover, the pore-air pressure in the soil column may be of significance since the soil temperature changes substantially (see Fig. 3). Therefore, the investigation would be ideal if pore-air pressure in the soil column can be measured. To determine which meteorological data are relevant and appropriate to assign as boundary conditions in a numerical analysis, a prototype soil column apparatus equipped with a complete meteorological station is designed by Caruso & Jommi. The height and the internal diameter of the soil
57
Level (mm)
Figure 1. Recorded (a) increase of soil and water temperature, (b) cumulated amount of evaporation and (c) relationship between calculated water heat, WH, (and G) and E (Blight).
50 -50 -150 -250 -350 -450 -550 -650 -750 -850 -950
column are 1200 mm and 290 mm, respectively. The meteorological station is installed at the same height as the column surface while five capacitive soil water sensors are installed at various depths along the soil column. The entire setup is then left in an environmentally uncontrolled atmospheric condition for a two month time period. This paper presents an intensive and comprehensive set of meteorological data as shown in Fig. 4. However, it is rather unfortunate that suction, one of the constitutive variables for unsaturated soils (Gens 2009), is not measured in this monitoring programme. As a result, detailed assessment of the measured data and their interaction could not be conducted from a geotechnical point of view. Montrasio et al. study the possible correlation between the daily output of a sequential data assimilation system called ACHAB (Assimilation Code for HeAt and moisture Balance) and effective degree of saturation, Se of a soil. The system is based on integrated use of remotely sensed land surface temperature and other common meteorological data over a territory from satellite acquisition. The daily output considered in this analysis is the evaporative fraction, EF, which is defined as the ratio between the energy consumed for evapotranspiration and the net available energy. A site in Italy is chosen for investigation where soil VWC is monitored by Time Domain Reflectometry devices at shallow depths (up to 180 cm) for three years. Assuming the residual VWC to be zero, Se is numerically equal to degree of saturation, Sr. Based on the data comparisons between the measured EF and the calculated Sr, an empirical correlation is introduced and is expressed as follows: EF a Sr exp b
250 200 150 8 100 4 50 0 0 20/10/09 30/10/09 09/11/09 19/11/09 29/11/09 09/12/09 19/12/09 Date 20 200 160 10 120 80 0 40 -10 0 20/10/09 30/10/09 09/11/09 19/11/09 29/11/09 09/12/09 19/12/09 Date 100 4 80 3 60 2 40 1 20 12 0 0 20/10/09 30/10/09 09/11/09 19/11/09 29/11/09 09/12/09 19/12/09 Date 16
where a and b are calibration parameters. Figure 5 shows the comparisons between the measured and estimated Sr for the monitoring period from March to December 2005. The seasonal variation of the estimated Sr is then used as an input parameter for subsequent real-time stability analysis. This study proposes an innovative and high-tech methodology to estimate Sr of superficial soil through equation (2) and to assess real-time slope stability over any particular site on a territorial scale. However, more research is needed to improve the predictability of Sr from theoretical considerations. Other possible combinations between the daily output of the ACHAB system and the indicators of soil moisture could also be sought for better correlation. 2.1.2 Flow and infiltration To examine the unrealistic assumption of material homogeneity often made in geotechnical practice, Le et al. present a stochastic study of unsaturated seepage through a heterogeneous embankment with a random porosity field. The total flow rate Q and the time to reach the steady-state of Tss, which are the main parameters frequently used as embankment design, are examined. A total of 1015 steady-state seepage analyses are performed with the same geometry and boundary conditions but with different random porosity fields. Porosity is assumed to follow a log-normal distribution with mean equal to 0.3, standard deviation equal to 0.1 and a correlation length of 1 m, and is generated by the Monte Carlo approach. Figure 6 shows the frequency diagram of Q and Tss. The results show that both these quantities appear to follow a lognormal probability distribution function. It also demonstrates that if the material is assumed to be homogeneous, the Q and Tss could be under-predicted and over-predicted respectively, which are not conservative in engineering designs. Arnedo et al. investigate the mechanism of gas flow in a compacted sand/bentonite mixture
1.0 0.9 0.8 100 80 60 40 20 Daily rainfall 120
(2)
Temperature (C)
Cumulated rain (mm)
Rain (mm/h)
Solar radiation (W/m2)
0.7 0.6 0.5 0.4 0.3 0.2 0.1 Sr Rainfall Sr estim. Sr meas. (-10 cm) Sr meas. (-25 cm)
Relative humidity (%)
Wind velocity (m/s)
0.0 0 17/02/05 08/04/05 28/05/05 17/07/05 05/09/05 25/10/05 14/12/05 02/02/06 Date
Figure 4. Measured meteorological data for two months (Caruso & Jommi).
Figure 5. Comparisons between the measured and estimated Sr for monitoring period from March to December 2005 (Montrasio et al.).
58
1 year after
2 years after
3 years after
Concentration
Figure 8. Salt concentration distributions in the model ground for 3 years (Nomura et al.).
1 year after 0.5 m 0.2 m Y 2 years after 3 years after 4 years after Concentration
Figure 6. Frequency histogram of (a) the total flow rate Q and (b) the time to reach steady-state Tss and their fitted log-normal probability distribution function (Le et al.).
Figure 9. Salt concentration distributions in the model ground using improved mulching technique (Nomura et al.).
Figure 7. Comparisons between the computed and measured injection and recovery pressures (Arnedo et al.).
through laboratory testing and numerical simulation. The experiment results show that preferential air flow paths are developed along (i) the openings of compaction interfaces and (ii) the contact between the sample and the cell wall. In particular, the formation of preferential paths through mechanism (ii) seems to be unrealistic in situ. Efforts to modify the sample-wall contact upon sample preparation may be needed to minimise this undesirable preferential gas migration. To take into account the possible formation of preferential flow paths upon gas flow in the 3D coupled thermo-hydro-mechanical finite element simulation, the existing permeability law is modified in this study. The initial porosity of the model sample is generated randomly prior to analysis. Figure 7 shows the comparison between the measured and computed injection and recovery pressures. The relatively large discrepancies between the computed and measured recovery pressures
may be attributed to the undesirable preferential air flows through mechanism (ii) which is not able to be simulated in the numerical analysis. Nomura et al. perform an unusual but useful numerical analysis to simulate salt transport in an unsaturated zone under the effects of evaporation and precipitation. An unsaturated soil/water coupled finite element code is used in this study. The climate data recorded from Thailand and a constant salt concentration ratio, C, of 0.03 is applied on the top and the bottom boundary of the model ground, respectively. In this study, the actual evaporation is determined by multiplying the field-measured evaporation rate with the effective degree of saturation. This calculation method may be oversimplified since it neglects the effects due to soil and air temperature and relative humidity (Wilson et al. 1997). The simulated results show that the salt continuously migrates upwards in the model ground until steady-state condition is reached, i.e., C 0.03 for the entire model ground (see Fig. 8). The authors further evaluate the effectiveness of using improved mulching technique in deterring salt damage. The mulching layer has a relatively large particle size and is overlaid on the model ground at a certain depth. Sensitivity analysis demonstrates that a mulching layer of 20 cm which embeds at a depth of 50 cm is effective for the inhibition of salt accumulation (see Fig. 9). 2.2 Case histories
2.2.1 Field monitoring Scotto di Santolo & Evangelista instrument four pyroclastic slopes in Italy for 5 years (2000 to
59
0 -20 Matrix suction (kPa) -60 -80 0 -20 -40 -60 -80 (b) 0 500 1000 Time (day) 1500 2000
Volumetric Water Content (%)
60 50 40 30 20 10 (a) 0 1
Characteristic curve-Forest Recovery Plot (30 cm)

Papel Filtro - Suco Mtrica - Amostra 9 - n = 63% van Genuchten - = 0,40 - n = 1,34 - m = 0,25 Papel Filtro - Suco Total - Amostra 9 - n = 63% Papel Filtro - Suco Mtrica - Amostra 22 - n = 62% Papel Filtro - Suco Total - Amostra 22 - n = 62%
-40 (a)
Figure 10. Measured and fitted seasonal variation of matrix suction for 5 years at depth (a) 0.5 m and (b) 2 m (Scotto di Santolo & Evangelista).
10
100
1,000
10,000
100,000
Suction (kPa) 60 50 40 30 20 10 0 (b) 1 10 100 1,000 Suction (kPa) 10,000 100,000 Characteristic Curves-Pasture Plot
Pasture - 30 cm Pasture - 60 cm Pasture - 90 cm
2005), to monitor their responses under seasonal variations. In situ measurements of matrix suction by tensiometers at six different depths (from 0.5 to 4 m) are presented and are best-fitted by a sinusoidal function with period equal to a year (see Fig. 10). It is interesting to know that the measured and fitted matrix suctions at some depths are quite consistent with each other. In the second part of this study, the fitted time history of matrix suction at depths of 0.5 and 4 m are used as the top and the bottom boundary condition for one-dimensional numerical simulations, respectively. The results seem to indicate that the seasonal variation of matrix suction can be predicted by using a sinusoidal function. Nacinovic et al. present and describe the procedures to install the filter paper and to measure the total suction in situ. The filter papers used in this study are of different batches and sizes whereas only one single calibration curve is used (ASTM 2003). Total suction in the four experimental plots is monitored for half a year and the measured data appear to correlate with the precipitation data. In the second part of the study, characteristic curves of four undisturbed samples are measured using filter papers and Fredlund SWCC Device. The measured characteristic curves are then fitted using the semi-empirical fitting method proposed by van Genuchten (1980) (see Fig. 11). However, some measured data points are not shown in some characteristic curves. The consistency using one single calibration curve can be evaluated if all the measured data are presented. To investigate the effectiveness and performance of a capillary barrier system (CBS) as a slope cover, Rahardjo et al. perform a one-year field monitoring on three slopes (original slope, grassed slope with CBS and bare slope without CBS) in Singapore. The variation of matrix suction and groundwater level at the crest, middle and toe of each slope are measured by tensiometers and piezometers, respectively. The monitoring results show that the porewater pressures (PWPs) in the grassed slope with CBS are generally lower than that in the original
Figure 11. (a) Measured and fitted characteristic curves at forest recovery plot at 30 cm depth and (b) fitted characteristic curves at Pasture plot at depths of 30, 60 and 90 cm (Nacinovic et al.).
Figure 12. Response of PWP and water table for the grassed slope with CBS (Rahardjo et al.).
slope, illustrating the effectiveness of CBS in minimizing rainfall infiltration into the slope (see Figs. 12 and 13). The authors also show that similar field observations can be simulated by conducting transient seepage analysis. This study provides valuable field data in demonstrating the performance of the CBS under seasonal variations. However, unsaturated hydraulic properties, i.e., SWCC and k(), of each material (the coarse-grained and the fine-grained layers of the CBS and the underlying soil), which may be useful to assist readers to
60
700 600 Leachate at the bottom (wet side) 500 400 300 200 100 0 29.10.08 Precipitation
20 Cumulative leachate / interflow (mm) 18 16 14 12 Interflow in the drainage mat 10 (wet side) 8 6 4 2 0 28.12.08 26.02.09 27.04.09 26.06.09 25.08.09
Cumulative precipitation (mm)
Interflow in the drainage mat (dry side) Leachate at the bottom (dry side)
Figure 13. Response of PWP and water table for the original slope (Rahardjo et al.).
N II I
Figure 15. Comparisons of the cumulated leachate at each side of the embankment and the cumulated precipitation (Birle et al.).
"dry" compacted material "wet" compacted material 10
extensometers
test areas vertical clay barriers II 10 5 I 5 10
Figure 14. et al.).
Plane view of the test embankment (Birle
understand how CBS helps to preserve suction in a slope, are not measured but rather predicted by semi-empirical equations. Birle et al. perform a full scale field monitoring to investigate water flow in an embankment which is built with organic soil under the effects of climate. The plane view of the test embankment is shown in Fig. 14. In order to study the effects of initial water content on the response of the test embankment, the west and east sides of the embankment, which are separated by vertical clay barriers, are compacted at an initial water content of 55% and 75%, respectively. A drainage mat which acts as a capillary barrier is then installed on top of the compacted embankment before a non-compacted top layer is placed. Apart from the measurements of matrix suction and VWC, the leachate discharge at the bottom of each side of the embankment is also measured. The climate data, including air temperature, relative humidity, effective radiation and wind velocity and direction, is measured by a meteorological station close to the test embankment. The results show that the leachate discharge rate at the west (wetter) side of the embankment
is always significantly higher than that at the east (drier) side, where the leachate at the west side is found to mainly come from the consolidation of the compacted organic soil (see Fig. 15). The slight reduction of the measured VWC of the compacted soil at the east side of the embankment demonstrates that the top drainage mat can effectively reduce the rainfall infiltration into the compacted embankment. In order to better analyse the response of the unsaturated compacted embankment, it would be more useful if the measured matrix suction is presented. Moreover, although a comprehensive instrumentation scheme is implemented in this study, it is rather unfortunate that other measured climate data are not presented. It is believed that the measured data is extremely useful in investigating the soil-atmosphere interaction as well as its effect on the overall behaviour of the test embankment. Maciel et al. and Lope et al. investigate landfill gas (LFG) surface emission using three different cover layers in an experimental cell (36,659 t) in Brazil, namely Methanotrophic Layer (MET), Capillary Barrier Layer (BAC) and Conventional Layer (CONV) (see Figure 16). Gas volume concentration of CH4, CO2 and O2, CH4 oxidation potential, pressure and temperature in each cover layer are measured internally using the static chamber technique described by Maciel (2003). The results show that the average CH4 flux rate can be greatly reduced by using BAC (15.95 g/m2 d) and MET (22.16 g/m2 d), which is about 7 times lower than that using CONV (161.49 g/m2 d). The lower emission rate using the BAC and MET is likely attributed to the more propitious environment for methanotrophic bacteria to grow, favouring CH4 oxidation within the soil cover. Concerning the gas migration problems in an unsaturated porous medium for some geo-environmental problems, there may be a need for further research on the unsaturated air permeability while possible in situ measurements of poreair pressure would be more relevant.
10
61
Metanotrophic layer 0,3-0,45 m 0,3 m 75% soil + 25% compost Compacted soil Waste 0,1-0,2 m 0,3 m
Table 2. et al.).
Infiltration rate for each cover layer (Lope Infiltration MET 32,0% 12,2% 15,8% 0,0% 1,6% 0,6% 0,9% BAC 96,7% 86,7% 75,3% 0,0% 0,0% 0,5% 0,3% CONV 12,3% 0,1% 0,0% 0,0% 0,0% 0,0% 0,0%
Month jun/09 jul/09 aug/09 sep/09 oct/09 nov/09 dec/09

1.3 Dry density d (g/cm3) 1.28 1.26 1.24 1.22
Rainfall (mm) 333,0 387,4 288,4 83,3 16,3 49,9 47,8
Capillary layer 0,55-0,7 m 0,2 m 0,3-0,55 m 0,2 m
Compacted soil geotextile gravel Waste Conventional layer
0,75-0,85 m
Compacted layer
0,5-0,65 m
Waste
: Simple compaction : Cyclic compaction
Figure 16. Schematic diagram of the three top cover layers (Maciel et al. and Lopes et al.).
Lopes et al. further evaluate the three cover layers described as shown in Fig. 16 by assessing and monitoring the amount of water infiltrated under precipitation for 1 year. A lysimeter which is placed beneath each top cover is used to measure the amount of water infiltrated daily. Three pairs of temperature and water content sensors are also installed in each cover layer. Table 2 summarises the rainfall amount and the measured infiltration rate for each cover layer in different months. Despite the lower CH4 emission using either the BAC or the MET, the high infiltration rate indicates that both layers are ineffective in preventing the production of leachate when the infiltrated water mixes with waste beneath. Based on the measurements from the two studies, both Maciel et al. and Lopes et al. propose a cover profile from top to bottom, a soil/ compost mix followed by a compacted clay layer and a permeable layer. 2.2.2 Engineering applications In order to elucidate the compaction mechanism using principles of soil mechanics, Shibata et al. simulate static and cyclic compaction processes using an unsaturated soil/water coupled finite element code. The constitutive models for soil and water are exactly the same as those described by Nomura et al. in the Section 2.1.2 of this report. In this study, the model specimens are prepared at the same initial void ratio with different initial gravimetric water contents (GWC) while both the static and the cyclic compaction is performed under a fully undrained condition. It should be pointed out that the simulations do not consider the compression of air and its influence on volume change, which may be crucial in this exercise.
1.2 0
10
20 30 40 Water content w (%)
50
Figure 17. Simulated static and cyclic compaction curve of the model specimen (Shibata et al.).
Figure 17 shows the comparison between the static and the cyclic compaction curve for the model specimen. With the consideration of the hysteresis of soil-water retention characteristic curve (SWRCC) in the coupled analysis, volumetric strain accumulates. The dry density thus becomes larger than that of the static compaction curve as the water content increases. The authors claim that the simulated results show good agreement with the measured data by Kawai et al. (2003), although details are not presented and described in the paper. Concerning the needs of understanding the PWP development during embankment construction, a series of laboratory tests are conducted by Marinho & Massad. They investigate the effects of initial water content, i.e., optimum water content (O1), dry of optimum (D1) and wet of optimum (W1, W2, W3 and W4), on PWP generation when each triaxial specimen is subjected to undrained isotropic compression. A high capacity tensiometer is installed at the base of each specimen to monitor the change of PWP throughout each test. Although it is well-known that the increase of total pressure would generate excess PWP under undrained conditions, this study provides useful information for embankment engineers in quantifying the amount of PWP generated at various confining pressures
62
(or equivalent heights of an embankment) (see Fig. 18). The investigations would be more ideal if transient measurements of both pore-air and PWP can be measured simultaneously. Vanapalli et al. propose a general expression to estimate the ultimate shaft resistance of piles in unsaturated sands, Qf(us), which is shown as follows: Q f us Q f Q f u
i 1 i n
a uw
Table 3. Comparison between the measured and calculated ultimate shaft resistance (compression test) for silty sand ( 0.35) (Vanapalli et al.). Shaft length intervals Test # (m) T1 T2 0.45 0.45 0.60 0.45 0.45 0.60 0.45 0.45 0.60 Degree Cal. Meas. Matric of suction satura- Q(ua-uw) Qf(us) Qf(us) (kPa) tion (S) (kN) (kN) (kN) 13.5 15.0 14.0 1.50 1.00 0.90 11.0 16.0 16.0 0.780 0.765 0.775 0.985 0.999 0.999 0.810 0.755 0.755
( z )
dL
(3)
[(ua uw )S (tan )] dL where Ki tan ; Ki is coefficient of earth pressure along pile shaft; is angle of friction along the soil/pile interface; (ua uw) is matrix suction; S is degree of saturation; and is fitting parameter. The contribution of ultimate shaft resistance due to matrix suction comes from the second term on the right hand side of the equation. A series of small-scale model tests under one gravity (1 g) is then undertaken in the laboratory to determine the shaft resistance of model jacked pipes (0.65 m in diameter and 1.5 m in length) embedded in sandy soils under saturated and unsaturated conditions. Each pile is tested in axial compression and tension. For each test, jet-filled tensiometers are installed to measure the matrix suction changes in the soil stratum up to 1.2 m depth. Table 3 summarises the measured shaft resistance and those calculated using equation (3). The authors conclude that the contribution of matrix suction to the ultimate shaft resistance is found to be in the range of 35% to 40% of the total shaft capacity in silty sand. It should be noted that the use of equation (3) to estimate the ultimate shaft resistance in field conditions should be treated with caution because the test results are calibrated at relatively low stress levels. At a given density, soil at a deeper region is subjected to a higher confining pressure
Equivalent Height (m) 20 30
T4

1.84 0
3.1
4.7
0.74 0.9
1.82
3.11 5.1
where contractive behaviour (instead of dilative) is more likely to occur. 3 SOA: SLOPE INSTABILITY PROBLEM DEFORMATION AND FAILURE MECHANISMS OF MAN-MADE FILL SLOPES Background
3.1
400 300 200 100 0 -100 -200 -300
10
40 W1 W2 W3 W4 O1 D1
50
200
400 600 Confining Pressure (kPa)
800
1000
Figure 18. Measured PWPs at various confining pressures for specimens prepared at different initial water content (Marinho & Massad).
Hong Kong is located in a sub-tropical region with a high landslide risk. This is because of its steep topography and prolonged rainfall in the wet season. In order to meet the demands for flat lands for housing and transportation developments in Hong Kong between the 1960s and the 1970s, a large number of man-made fill slopes were formed. These fill slopes were often constructed by dumping construction waste and decomposed geomaterials such as completely decomposed granitic (CDG) and completely decomposed volcanic (CDV) soils. No proper compaction was required prior to the establishment of the Geotechnical Control Office (GCO) in 1977. These pre-GCO slopes are generally in a loose state and highly vulnerable to landslides during and after a rainstorm. Two major catastrophic loose fill slopes failed at Sau Mau Ping in Hong Kong, resulting in the deaths of 71 and 18 people in 1972 and 1976, respectively (Malone 1998; Ng 2007). After the 1976 Sau Mau Ping failure, an international panel investigated and concluded that the failure was caused by the loss of shear strength of the loose fill, resulting in so-called static liquefaction (Knill et al. 1976). However, according to some eye-witnesses, the fill slope travelled downhill like a carpet when it failed (Hong Kong Government, 1977). The actual failure mechanism
Pore Water Pressure (kPa)
63
has generated considerable debate in Hong Kong over the last decade as to whether static liquefaction could occur in CDG and also whether soil nails can be used to stabilise existing sub-standard unsaturated loose fill slopes or not. The debate resulted from the lack of fundamental understanding of firstly the static liquefaction failure mechanism and secondly the material behaviour of CDG and CDV at both unsaturated and saturated states. In order to understand the failure mechanisms of loose fill slopes and explore possible stabilisation methods such as the use of soil nails, a series of comprehensive triaxial tests were carried out on saturated and unsaturated loose CDG and CDV specimens. Moreover, centrifuge model tests were conducted on CDG fill slopes by the University of Cambridge and the Hong Kong University of Science and Technology (HKUST) independently to investigate possible failure mechanisms of static liquefaction, if any. In this section, some selected triaxial and centrifuge model tests are described and reported. 3.2 Clarification of terminologies relating to static liquefaction
(a)
Deviator stress
B Strain hardening Dilation Strain softening C Strain hardening Limited liquefaction
A Liquefaction Strain softening (b)
Axial strain
B Dilation Undrained strength increase Phase due to dilative transformation C tendency point Limited liquefaction contractive tendency A Liquefaction Undrained strength reduction due to contractive tendency Mean effective stress Contractive tendency Liquefaction A
Deviator stress Excess pore pressure

(c)
Axial strain C Limited liquefaction contractive tendency B Dilation Dilative tendency
It is rather unfortunate that the term static liquefaction is frequently used in various literatures to describe material behaviour from an element test such as a triaxial test. Figure 19 shows some typical results from monotonic triaxial tests on saturated, anisotropically consolidated sand specimens. As shown in Fig. 19(a), a very loose sand specimen, A, exhibits a peak undrained shear strength at a relatively small shear strain and then collapses to a much smaller shear strength at large strains. This behaviour is often causally referred to as liquefaction or flow liquefaction by many researchers and engineers. No matter whether it is called flow liquefaction or liquefaction, the terminology to describe the material behaviour observed in the laboratory is rather confusing and, strictly speaking, incorrect. Would it be clearer and more precise to describe the material behaviour of the loose specimen, A, and a dense specimen, B, as strainsoftening and strain-hardening, respectively, in the deviator stress-axial strain space (see Fig. 19(a)). In the mean effective stress-deviator stress space (see Fig. 19(b)), would it be more precise to use the terms undrained strength reduction (or so-called collapse (Sladen et al. 1985)) and undrained strength increase to describe the strength changes of specimen A and specimen B, respectively? Of course, it is well-recognised that a reduction and an increase in undrained shear strength are caused by the respective tendency of sample contraction and dilation, leading to a respective increase and a reduction in pore water pressure (u) for specimens A and B during
Figure 19. Liquefaction, limited liquefaction, and dilation in monotonic loading tests (modified from Castro 1969, Kramer 1996).
undrained shearing (see relationship between u and axial strain in Fig. 19c). It must be pointed out that this is just material element behaviour that does not necessarily capture and represent the global behaviour of an entire fill slope or an earth structure. 3.3 Decomposed materials for element tests
The CDG samples used in triaxial tests were obtained in Shau Kei Wan (SKW) on Hong Kong Island (Ng et al. 2004) and a quarry located in Cha Kwo Ling (CKL) on the Kowloon Peninsular (Ng & Chiu 2003). Figure 20 shows the particle size distributions of the two CDG samples. The CDG is coarse-grained sand with an appreciable amount of silt and clay that can be classified as Grade V weathered soil (GEO 1988). After obtaining samples from each site, the freefall method as proposed by the Japanese Society of Soil Mechanics and Foundation Engineering (JSSMFE 1982) was used to break down small lumps of the soil samples consistently. The grading curve of the treated soil was obtained in accordance with GEO Rep. No. 36 (GEO 1996) for decomposed materials. The grading curves of the specimens from the two sites suggest that the treated CDG is silty sand. The treated CDG is predominately coarse- but well-graded and the fine contents are about 26% and 24% for the CKL and SKW sites. Based on the standard Proctor tests (BSI 1990) it was found that the maximum dry density for the
64
100 Percentage finer (%) 80 60 40 20 0 0.001
LB-Sand CKL-CDG SKW-CDG BH-CDG WTS-CDG VP-CDV
Deviator stress
Z' A' Matric suction F' B' O' C' X' Y'
D'
0.01 0.1 Particle Size (mm) 1 10
E'
Figure 20. Particle size distributions of LB sand, CDG and CDV (modified from Ng 2007).
CDG taken from SKW was 1650 kg/m3, which corresponded to the optimum moisture content of 21% (Ng et al. 2004); whereas the maximum dry density for the CDG taken from CKL was about 1670 kg/m3, which corresponded to the optimum moisture content of 20% (Ng & Chiu 2003). 3.4 Experimental programmes for element tests
Figure 21. Various stress paths for triaxial tests conducted on loosely compacted CDG from CKL (Ng & Chiu 2003).
are presented and discussed in this paper. Other details are reported by Ng & Chiu (2003). 3.5 Specimen preparation and test procedures
Triaxial compression tests were conducted to investigate the behaviour of saturated CDG from SKW and CKL and any key governing factors relevant to rain-induced slope instability problems. For the CDG samples taken from SKW, two saturated test series including isotropically consolidated undrained (CU) tests and anisotropically consolidated constant deviator stress tests with reducing mean effective stress (CQ) or field stress path tests are described and discussed in this paper. To investigate the undrained response of loose saturated CDG at SKW, five CU tests were carried out at relatively low initial consolidation pressures ranging from 50 kPa to 250 kPa, which were selected to simulate appropriate loading conditions of the fill slopes in Hong Kong (Ng et al. 2004). The CQ test series (eight tests on the SKW soil) considered a decreasing mean effective stress path under constant deviator stress conditions. The CQ drained tests were conducted by increasing pore water pressure at a very low rate with the top and bottom drainage lines connected to a pore pressure controller, whereas the applied total vertical and horizontal stresses were held constant throughout the tests. This type of stress path was intended to mimic the loading of soil elements within a slope subjected to a slow increase in pore water pressure or a reduction in soil suction due to rainfall or a rising ground water table (Ng et al. 2004). For soil samples taken from CKL, both saturated and unsaturated tests were carried out. Figure 21 summarises the test paths adopted for this site. Due to page limit, only some results from stress paths O C D E F (saturated), O X Y Z (unsaturatedconstant GWC) and O X Y A B (unsaturatedfield wetting path, i.e., rainfall infiltration)
In order to produce specimens with very loose structures, the moist-tamping technique (Sasitharan et al. 1993) was adopted. Specimens of 70 mm in diameter and 140 mm in height and 76 mm in diameter and 152 mm in height were prepared from the soil taken from SKW and CKL respectively. Due to slightly different sizes of triaxial cells, slightly different sizes of specimens were prepared from the two sites. In order to ensure the same initial microstructure and fabric, all the specimens were prepared at the same initial water content, which is the optimum moisture content of each CDG obtained by the standard Proctor compaction test. Each specimen was compacted in ten layers. In order to reduce end constraints on a test specimen, two enlarged lubricated end plates were used at both ends of the specimen. Other details of specimen preparation and test procedures are given by Ng & Chiu (2003) and Ng et al. (2004) for specimens from CKL and SKW respectively. 3.6 3.6.1 Behaviour of saturated CDG
Instability and strain softening behaviour of saturated CDG from SKW and CKL The results of strain-controlled CU monotonic undrained shearing tests on the loose CDG are shown in Fig. 22(a) and Fig. 22(b) in p q and q a plane, respectively. As shown in Figure 22(a), the tendency of contraction of the soil was evident as all the effective stress paths bend from the right to the left indicating an increase in pore water pressure during shearing. Each effective stress path is characterized by its initial increasing deviator stress
65
with decreasing mean effective stress. After a peak was reached, the deviator stress decreased with a further reduction in the mean effective stress. The slope of the critical state line (M) obtained from the undrained tests was 1.58 whereas the slope for the instability line, Minst, was 1.12. The instability line is the line joining the origin and the peak of each effective stress path. It provides the lower limit of the region of potential instability for a loose soil in an undrained test. The corresponding mobilized effective friction angles at the critical state, cs, and the peak, inst, are 38.7 and 28.1, respectively. As shown in Fig. 22(b), strain softening was observed in all the tests. The shear strength reduced after reaching an initial peak in the deviator stress at about 1% axial strain. The deviator stress stayed constant or changed very little once the axial strain reached 25%, at which point the critical state was attained. The maximum drop in the deviator stress after a peak, or liquefaction potential, (i.e., qmaxqmin) decreases with an increase in the confining pressure (Ng et al. 2004). Since the magnitude of the drop was caused by the increasing pore water pressure, which in turn was attributed to the contractive tendency of the specimen, it is clear that the larger the drop, the higher the tendency to contract. Figure 23 shows the test results of the CQ series in the p q plane, where p is mean effective stress and q is deviator stress. In the figure, the number in each bracket denotes stress ratio (c) (i.e., ranging from 0.47 to 1.00) obtained after anisotropic consolidation to various stress paths. For clarity, only
180 Deviator stress, q (kPa) 160 140 120 100 80 60 40 20 0 0 50 100 150 200 250 Mean effective stress, p' (kPa) CU100 CU250 CU150 b) 300 M = 1.58 CU050 CU200 CU100 CU250 CU150 a)
120 100 Deviator stress, q (kPa) 80 stress state at onset of large axial strain (at collapse) M = 1.58 CQ060-1 Minst = 1.12
CQ060-1 (1.00) CQ078 (0.47) CQ085 (0.53) CQ120 (0.57) CQ123 (0.50)
60 CQ123 CQ060-2 40 20 0 CQ085
20
40 60 80 100 Mean effective stress, p' (kPa)
120
140
Figure 23. Effective stress paths of constant deviator stress (CQ) tests (Ng et al. 2004).
Minst = 1.12
180 Deviator stress, q (kPa) 160 140 120 100 80 60 40 20 0 0 5 10 15 20 Axial strain, a (%) 25 30 CU050 CU200
some representative stress paths of the CQ tests are shown in the figure. Obviously, the stress state at the onset of large plastic strains or the initiation of the static instability does not represent the critical state, but it denotes the state of the soil element at which it is no longer able to support the external shear stress imposed on it. As some test pairs have the same deviator stress but with different mean effective stresses, different void ratios at collapse would be expected. This can be illustrated with the CQ060-1 and CQ123 tests which had the same initial deviator stress of 60 kPa but collapsed at different mean effective stresses and hence they mobilized different friction angles of 31.8 and 36.4, respectively, at collapse. Similar results were also observed for the CQ060-2 & CQ085, test pair. The observed different collapse points indicate the existence of non-unique yield surfaces, as a result of different extents of expansion of yield surfaces during pre-shearing (Ng et al. 2004). As shown in Fig. 23, the stress states corresponding to the initiation of static instability all lie above the instability line (Minst) determined from the CU tests (Ng et al. 2004). The mobilized friction angles at the initiation of the static instability (col) for all the CQ tests range from 31.8 to 36.4, higher than the instability angle of 28.1 from the CU tests, but less than the critical state angle of 38.7. This indicates that the use of inst in slope stability analysis would be conservative. 3.7 Shear behaviour of unsaturated CDG
Figure 22. (a) Effective stress paths, (b) stress-strain curves of CU tests on CDG taken from SKW at various consolidation pressures (Ng et al. 2004).
3.7.1 Results from constant water content tests The stress path followed is given in Fig. 21 (i.e., path O X Y Z ). After applying an initial suction of 40 kPa, four unsaturated specimens (ui1-g to ui4-g) from CKL were isotropically compressed and then sheared at a constant water content in a computer-controlled triaxial stress path apparatus for testing both saturated and unsaturated soils (Ng & Chiu 2001, Ng & Chen 2005). Figure 24(a) shows that the deviator stress (q) increases steadily
66
with axial strain and almost approaches a plateau at an axial strain over 25%, with no evidence of strain softening except at p 25 kPa (ui1-g). The deviator stress also increases with the applied net mean stress, p (Ng & Chiu 2003). Figure 24(b) shows that only contractive volumetric behaviour is observed for all specimens except specimen ui1-3 and the amount of contraction increases with the applied p. For specimens compressed to a p value of 25 kPa (ui1-g) or located on the dense side of the critical state (Ng & Chiu 2003), the small amount of contraction increases to a peak value at a low axial strain and then reduces from the maximum to an approximated constant value with increasing axial strain. This is a sign of the tendency to dilate after reaching the peak value. For the other specimens (ui2-g to ui4-g), the amount of contraction increases monotonically with the axial strain to a steady value. The volumetric behaviour may be governed by the compression of inter-particle air voids and the interlocking effect between individual soil particles during shearing. For specimens compressed at a low net mean stress (ui1-g), the state before shearing lies below and far away from the critical state line (Ng & Chiu 2003). This specimen behaves like a dense material. The volumetric compression of the interparticle air voids may be compensated by the volumetric expansion due to the interlocking between the soil particles. As a result, the amount of the overall volumetric contraction is relatively small. For specimens compressed to a higher net mean stress
600 Deviator stress, q (kPa) p = 25 kPa (ui1-g) p = 50 kPa (ui2-g) p = 100 kPa (ui3-g) p = 150 kPa (ui4-g)
(ui4-g), the state before shearing lies above the critical state line (Ng & Chiu 2003). The specimen behaves like a loose material. During shearing, the relatively large inter-particle air voids can be compressed and individual soil particles may roll into these interparticle air voids. Thus, the specimen exhibits substantial amounts of volumetric contraction. This implies that the design of any stabilization measures in loose unsaturated granitic fill slopes and foundations resting on the soil should take into account the significant shear-induced volume changes in the soil leading to potential serviceability problems. Figure 24(c) shows that all the tests exhibit a reduction in suction due to an increase in VWC and the amount of reduction increases with the applied p. The suction reduces steadily with axial strain and approaches a steady value at the end of the test. The suctions at 25% axial strain range from 25 to 33 kPa (Ng & Chiu 2003). As the applied p increases, the volumetric contraction of the specimen increases (refer to Fig. 24(b)). 3.7.2 Wetting behaviour of unsaturated specimens at CKL The test path followed is given in Fig. 21 (i.e., the field test path simulating rainfall infiltration O X Y A B ). After isotropic compression, four unsaturated specimens (ua1-g to ua4-g) were sheared at constant p and suction (s 40 kPa) until reaching a net stress ratio () of 1.4. Then, the specimens were wetted by reducing the soil suction while p and q were kept constant. The stress states of these specimens (ua1-g to ua4-g) before wetting lie below the CSL obtained from the constant water content tests with s 25 to 33 kPa but they are located above the CSL determined from s 0 kPa tests (Ng & Chiu 2003). Figure 25(a) shows the variation of the axial strain with suction for the wetting tests at a constant deviator stress. As the suction decreases from an initial value of 40 kPa to about 20 kPa, limited axial strain (less than 4%) is mobilised. As the suction continues to decrease, the rate of increase in the axial strain accelerates towards the end of the test. There is a substantial increase in axial strain for the last decrement of suction for all specimens. All the tests were stopped because the bottom piston of the stress path apparatus reached the travel limit. It is noted that the unsaturated specimens approach failure at lower suctions for higher applied net mean stresses. This agrees with the observation from the constant water content tests. As the confining pressure increases, the volumetric contraction of the specimen increases. As a result, the degree of saturation increases, which corresponds to a smaller value of suction. Figure 25(b) shows the variations of volumetric strain with suction. For unsaturated specimens compressed to a net mean stress higher than 50 kPa
400
200 (a) 0 p = 25 kPa (ui1-g) p = 50 kPa (ui2-g) 10 Axial strain (%) 12 p = 100 kPa (ui3-g) p = 150 kPa (ui4-g) 20 30
Volumetric strain (%)
10 8 6 4 2 0 -2 0 0
(b) 10 Axial strain (%) 20 30
Change of suction (kPa)
-4 -8 -12 -16
p = 25 kPa (ui1-g) p = 100 kPa (ui3-g)
p = 50 kPa (ui2-g) p = 150 kPa (ui4-g)
(c) 0 10 Axial strain (%) 20 30
Figure 24. Observed relationships of (a) stress-strain; (b) volumetric-axial strain; and (c) change of suction and axial strain of unsaturated CDG tested at constant GWCs (Ng & Chiu 2003).
67
30 Axial strain (%)
20
p = 25 kPa (ua1-g) p = 50 kPa (ua2-g) p = 100 kPa (ua3-g) p = 150 kPa (ua4-g)
(a)
10
0 40 4 Volumetric strain (%)
30
20 Suction, s (kPa)
10
0 (b)
p = 25 kPa (ua1-g) p = 50 kPa (ua2-g) p = 100 kPa (ua3-g) p = 150 kPa (ua4-g)
-2 40
30
20 Suction, s (kPa)
10
Figure 25. Relationships between (a) axial strain and suction and (b) volumetric strain for wetting tests at a constant deviator stress conducted on unsaturated CDG specimens (Ng & Chiu 2003).
(ua2-g and ua3-g), relatively small volumetric contraction is mobilised as the suction decreases from an initial value of 40 kPa to about 25 kPa. As the suction continues to decrease, the rate of increase in volumetric contraction increases towards the end of the test. For other specimens (ua1-g), dilative behaviour is observed and the volume reaches a maximum at a suction of 17 kPa. As the suction further decreases, the specimen commences to contract and the amount of volumetric dilation reduces. All four specimens exhibit an increase in the rate of volumetric contraction upon the application of the last decrement of suction, no matter if the specimen exhibits an overall volumetric contraction or dilation. It seems that this increase in the rate of volumetric contraction can be attributed to the collapse of the soil skeleton. The figure also shows that the unsaturated specimens exhibit an increasing tendency towards volumetric contraction as the applied net mean stress increases. This agrees with the results of the constant water content tests on unsaturated specimens and of the undrained tests on saturated specimens. However, the magnitude of induced volumetric strains at collapse during the wetting test path is considerably smaller than that induced during the constant water content tests (see Fig. 25(b)). This is because the wetting process (or reduction of soil suction) was conducted under a high constant net stress ratio (i.e., 1.4), which suppressed the amount of negative dilation (i.e., volumetric contraction) (Ng & Chiu 2003). This implies that anisotropic pre-shearing of a soil to a high net stress ratio appears to reduce the amount of contractive strains during the subsequent wetting. In other words, it
seems that steep slopes may be less vulnerable than gentle slopes to static liquefactions. Ng & Chiu (2001) also carried out tests on anisotropically consolidated unsaturated recompacted CDV subjected to a decreasing suction under a constant deviator stress. Figure 20 shows the particle size distribution of the CDV specimen. The CDV is a slightly sandy silt obtained from Victoria Peak (VP). Figure 26(a) shows the variation of axial strain with suction for the field stress path tests, which simulate rainfall infiltration, conducted on anisotropically consolidated unsaturated specimens. The deviator stress and the net mean stress were kept constant while the suction was decreased. As the suction decreases from an initial value of 15 to about 80 kPa, small axial strain (less than 4%) is mobilised for specimens consolidated to a net mean stress greater than 25 kPa (ua2 to ua4). As the suction continues to decrease, the rate of increase in axial strain accelerates towards the end of the test. For the specimen (ua1), the mobilised axial strain increases gradually as the suction decreases. It can be seen from the figure that the specimen approaches failure at higher suction for a higher applied net mean stress (Ng & Chui 2001). Figure 26(b) shows the variation of volumetric strain with suction for the field stress path tests. Relatively small volumetric strain is mobilised as the suction decreases from an initial value of 150 kPa to about 80 kPa. As the suction continues to decrease, contractive behaviour is observed for
25 20 Axial strain (%) 15 10 5 0 150 6 p = 25 kPa (ua1) p = 50 kPa (ua2) p = 100 kPa (ua3) p = 150 kPa (ua4) (a) 100 Suction, s (kPa) 50 0 p = 25 kPa (ua1) p = 50 kPa (ua2) p = 100 kPa (ua3) p = 150 kPa (ua4)
-2 150
(b) 100 Suction, s (kPa) 50 0
Figure 26. Relationships between (a) axial strain and suction, (b) volumetric strain and suction for stress path tests, which simulate rainfall infiltration, conducted on unsaturated CDV specimens (Ng & Chui 2001).
68
soil specimens consolidated to a net mean stress smaller than 100 kPa (ua1 and ua2) but dilative behaviour is observed for the other two specimens (ua3 and ua4). The anisotropically consolidated unsaturated specimens change from contractive to dilative behaviour as the applied net mean stress increases. It may be speculated that the stress states of ua3 and ua4 after anisotropic consolidation might fall on the dense side of their corresponding hyper critical state lines and so dilative behaviour is observed (Ng & Chui 2001). 3.8 3.8.1 Centrifuge modelling of liquefied flow of initially unsaturated loose sand fill slopes
800 700 600 500 q (kPa) e0 = 0.973 400 300 200 100 0 e0 = 0.983 e0 = 0.992 0 10 20 q (%) 30 e0 = 0.970 Quasi-steady state
(a)
Liquefaction potential (LP)
40
(b) 700
Model materialLeighton Buzzard (LB) sand Centrifuge model tests were carried out at the geotechnical centrifuge at HKUST (Ng et al. 2002a, Ng et al. 2006) to investigate the failure mechanisms of static liquefaction of loose fill slopes subjected to rainfall, rising ground water table and dynamic earthquake loadings (Zhang 2006, Zhang et al. 2006, Ng 2007). Leighton Buzzard (LB) Fraction E fine sand was selected as the fill material for the model tests. Figure 20 shows the gap-graded particle size distribution of LB sand. D10 and D50 of the sand were 125 m and 150 m, respectively. Following BSI (1990), the maximum and minimum void ratios of the LB sand were found to be 1.008 and 0.667, respectively (Cai 2001). The estimated saturated coefficient of permeability was 1.6 104 m/s. LB sand was chosen because of its pronounced strain-softening characteristics with its high liquefaction potential, LP, i.e., a substantial strength reduction in shear strength when it is subjected to undrained shearing (see Fig. 27(a)). The results from four loose specimens with different initial void ratios (eo) shown in the figure are obtained from isotropically consolidated undrained compression triaxial tests. The loose sand clearly shows pronounced strain-softening behaviour and substantial strength reduction in the deviator stress and shear strain (q-q) space and contractive responses in the mean effective stress (p) and deviator stress (q) space, i.e., p decreases continuously as q increases until the peak state is attained (see Fig. 27(b)), where p and q are equal to (1 2 3 ) / 3 and (1 3,) respectively. After the peak state, q drops (the soil collapses) with a large deformation developing until the quasisteady state (a shear strain of about 15%) or the critical state (shear strain 30%) is reached. The critical state friction angle (c) of the sand is 30 (Cai 2001). Following the approach proposed by Lade (1992), the angle of instability (ins) determined for the sand is 18.6. It is well-known that ins is dependent on void ratio and stress level
100
Figure 27. Contractive behaviour of loose LB sand under consolidated undrained tests (a) in the q q and (b) in p q planes (modified from Zhang 2006, data from Cai 2001).
(Chu & Leong 2002). For engineering assessment and design of remedial work for loose fill slopes, it may be reasonable to assume this angle is a constant as the first approximation. 3.8.2 Centrifuge model package and test procedures Figure 28 shows an instrumented 29.4 loose sand fill slope model together with locations of the pore water pressure transducers (PPTs) (Zhang & Ng 2003, Ng 2008, Ng et al. 2009). The model slope was prepared by moist tamping. The initial relative compaction was 68%. The body of the sand slope was instrumented with seven PPTs and arrays of surface markers were installed for image analysis of soil movements. Linear variable differential transformers (LVDTs) and a laser sensor were mounted at the crest of the slope to monitor its settlement. 3.8.3 Observed static liquefaction mechanism Although the initial angle of the loose slope was prepared at 29.4 at 1 g, the slope was densified to 80% of the maximum relative compaction due to self-weight compaction at 60 g. The slope angle was therefore flattened to 24 (see Fig. 29(a)),
69
LVDT LVDT & Laser sensor Drainage board
Model container Inlet hole PPT7 Reflector PPT5 PPT6 PPT4 y PPT2 PPT1 x Sand 305 Outlet hole PPT3
Temporary reservoir
Figure 28. Centrifuge model of a loose sand fill slope subjected to rising ground water table at 60 g (Zhang & Ng 2003).
which is steeper than 18.6, the angle of instability. This implies that the slope was vulnerable to instability, which could lead to liquefaction (see Fig. 27). At 60 g, the 18 m-height (prototype) slope was de-stabilised by rising ground water from the bottom of the model (Zhang 2006, Ng et al. 2009). The loose sand slope liquefied statically and flowed rapidly (see Fig. 29(b)), i.e., it followed a process in which the loose slope was sheared under undrained conditions, lost its undrained shear strength as a result of the induced high pore water pressure (see Fig. 30) and then flowed like a liquid, called liquefied flow. It is evident from the video images that an upper portion of the sand slope remained unsaturated before the liquefaction took place. Figure 30 shows the measured rapid increases in the excessive pore water pressure ratio (u/v) within about 25 seconds (prototype) at failure at a number of locations in the slope during the test. The maximum measured u/v was about 0.6, which would be much higher if a properly scaled viscous pore fluid were used to reduce the rate of dissipation of excess pore pressure in the centrifuge. This means that the slope would liquefy much more easily. As shown in Fig. 29(b), the completely liquefied slope inclines at about 4 to 7 to the horizontal after the test. The observed fluidization from in-flight video cameras and the significant rise in excessive pore water pressures during the test clearly demonstrate the static liquefaction of the loose sand fill slope. It should be noted that measurements of sudden and significant rise of excessive pore water pressures are essential to prove or verify the occurrence of static liquefaction of loose fill slopes if no video recording is available. The liquefaction of the loose sand slope was believed to be initially triggered by seepage forces in the test (Ng et al. 2009). It is obvious that soil nails cannot be used to stabilize a loose sand fill slope which has a high liquefaction potential (see Fig. 27(a)).
29.4
1130,7
Figure 29. Slope profile in a loose sand fill test (a) before rising ground water table; (b) after static liquefaction (Zhang & Ng 2003).
1.0 0.8 Excess pore pressure ratio (uw/v') 0.6 0.4 0.2 0.0 -0.2 -0.4 -0.6 -0.8 -1.0 37.8 38.2 38.6 39.0 39.4 39.8 40.2 40.6 41.0 41.4 41.8 42.2 42.6 43.0 Duration (min) PPT3 PPT2 PPT1 PPT6 Slope failure PPT7 PPT4 PPT5
Figure 30. Measured sudden and substantial increases in pore water pressure at seven locations inside the slope (Zhang & Ng 2003).
Figure 31 shows five postulated zones, Z1Z5, representing the sequence of the failure and liquefaction process of the slope (Ng et al. 2009). Z1 is a failure region de-stabilised by the loss of its toe due to the seepage force in the gully (drained failure). The soil mass at the toe of Z1 slid with the soil at the gully head to trigger the failure of Z2. The soil mass in Z2 collapsed rapidly (undrained) which was then followed by the collapse of Z3 (undrained) without inducing obvious deformation in the lower part. The collapses of Z2 and Z3 were due to the strain-softening
70
B Gully erosion Gully head

20 15 10 5 0
Water flow Sand movement Water
18.900 Prototype Scale 24.240
A
Z2
PPT7 PPT5 PPT4 PPT2 PPT1 PPT6
Z1 Z3 Z4
Slope profile Final slope profile before failure

PPT3
A
0 5 10
Z5 20 25 30 35 40 45 50 55 60 65 (m)
45
6.240 0.600 Scale
15
Figure 31. Postulated failure zones during the liquefaction of slope SG30 (Ng et al. 2009).
Unit in Metre
associated with the significant strength reduction (i.e. high liquefaction potential) of the loose LB sand as illustrated in Fig. 27. The rapid undrained collapses of Z2 and Z3 were evident from the measured large excess positive pore pressures at PPT7 (see Fig. 30). Subsequently, Z4 also collapsed as a result of the strain-softening associated with the significant strength reduction (high liquefaction potential) of the loose LB sand (see Fig. 27). The dotted line drawn between Z4 and Z5 in Fig. 31 represents the upper boundary of the stable region (Z5), monitored by markers and the small excess pore pressures at PPT1 and PPT2 (see Fig. 30) during the liquefaction process. Based on the observed static liquefaction mechanism, it is reasonable to deduce that soil nails cannot be used to stop any liquefied flow of loose sand fill slopes. However, the use of soil nails can reduce the magnitude of any excessive positive pore water pressure generated in a loose sand slope, minimize the chance of liquefaction and reduce damage after liquefaction (Zhang et al. 2006). 3.9 Investigation of deformation mechanisms of initially unsaturated CDG fill slopes in centrifuge
Figure 32. Displacement vectors in unreinforced loose CDG fill slope (CG45) (Ng 2007).
(2004) from their independent centrifuge model tests using the same loose CDG fill in Cambridge University, in which only non-liquefied slides were observed. In addition, a large-scale field test was reported by Tang & Lee (2003) on a loose CDG fill slope with an angle of 33 to the horizontal. The height and width were 4.75 m and 9 m respectively. It was constructed by the end-tipping method and resulted in a loose state with an initial dry density of 70 to 75% of the maximum dry density. It is considered that the stress state of this slope would represent reasonably well that of most of the existing fill slopes in Hong Kong. Only excessive settlement of the slope was recorded and no sign of liquefied flow and non-liquefied slide was observed. The observed liquefied flow in loose sand (see Fig. 29) and non-liquefied excessive settlements (refer to Fig. 32) in loose CDG fill slopes may be explained by the drop in undrained shear strength of CDG being substantially smaller than that of LB sand (see Figs. 22 and 27). 3.9.2 Non-liquefied slide of initially unsaturated loose shallow CDG fill slope The Housing Department of HKSAR has been actively looking for innovative methods to preserve the environment by minimizing the need for felling trees when improving the stability of existing shallow loose CDG fill slopes (Ng et al. 2007). Figure 33 shows an instrumented centrifuge model created to study the potential static liquefaction of a loose shallow CDG fill slope subjected to a rising ground water table. This model was intended to simulate a 1.5 m thick, 24 m high layered fill slope when tested at 60 g. The wooden block was used to represent an impermeable intact rock mass. The particle size distribution of the CDG used is denoted as WTS in Fig. 20. The initial fill density was 66%. In addition to laser sensors (LSs) installed for monitoring soil surface movements, pore pressure transducers (PPTs) were installed to measure excess pore water pressure during the test. Effects of layering were considered by tilting the model container during model preparation. The slope was destabilised by downward seepage created by a hydraulic gradient,
3.9.1 Excessive settlement of loose CDG fill slope Similar to the loose sand fill slope, centrifuge model tests on CDG fill slopes were also carried out using samples taken from Beacon Hills (BH) (Ng et al. 2002b, Zhang 2006). The particle size distribution of BH CDG samples is given in Figure 20. The main objective of these tests was to investigate failure mechanisms of static liquefaction (if any) in loose CDG fill slopes subjected to rising ground water and rainfall. The model slope was initially 45 to the horizontal and the initial relative compaction of the fill was less than 80%. At 60 g, a 300 mm high model slope was equivalent to an 18 m high slope at prototype. Figure 32 shows the measured displacement vectors of an unreinforced loose CDG fill slope de-stabilized by the rise of ground water. No matter whether rising ground water, rainfall or both were adopted to de-stabilise the loose CDG slopes in a centrifuge test, no sign of static liquefaction of the slopes was observed during and after the tests. Similar findings are also reported by Take et al.
71
PPT Unit: mm Model box Upstream drainage board Coarse soil
PPT2 PPT1 PPT3 Loose CDG (WTS) Upstream temporary Inlet hole Wood block PPT4 PPT5 PPT6 LS1 PPT7 PPTB PPT8 PPT 9 PPT C Coarse soil Outlet hole LS3 LS2 Downstream drainage board Downstream temporary
Figure 33. Model package of an instrumented shallow fill slope (Ng et al. 2007). Figure 35. Variations in the measured pore water pressure at the crest (PPT2) and at the toe (PPT7) of the slope with time (prototype) (Ng et al. 2007).
Figure 34. Top view of the model showing a nonliquefied slide (Ng et al. 2007).
which was controlled by the water level inside the upstream temporary reservoir and the conditions of the outlet hole located downstream (see Fig. 33). Two non-liquefied slides were induced in the test. Figures 34 and 35 show the occurrence of a nonliquefied slide and the measured excessive pore water pressure during the two slides, respectively. It can be seen from Fig. 35 that negative PWPs of 25 kPa and 36 kPa were recorded initially at the soil-rock interface near the crest (PPT2) and near the toe (PPT7), respectively. Once water was allowed to seep from the top to the toe controlled by the total head difference between the two locations, the negative pore water pressure was gradually reduced at PPT2, whereas the negative pore water pressure at PPT7 was rapidly destroyed after a prolonged delay. Two slides, initiated from the crest one after the other, were observed resulting in a small increase in excessive pore water pressures at both PPT2 and PPT7 each time. No static liquefaction was observed. It is evident from the images that a large portion of CDG remained unsaturated throughout the experiment. The results from this experiment are consistent with the centrifuge model tests by Take et al. (2004) on the same type
Figure 36. A slide-to-flow landslide triggering mechanism model (Take et al. 2004).
of loose CDG fill slopes, which are to be described later. 3.9.3 The slide-to-flow mechanism of initially unsaturated loose and dense shallow CDG fill slopes (Take et al. 2004) In an attempt to study a postulated mechanism of initially slow-moving slips driven by localized transient pore water pressure for loose fill slopes in Hong Kong, Take et al. (2004) carried out two centrifuge model tests on both initially unsaturated loose and dense CDG fill slopes. This attempt is intended to combine the two closely related pore water pressure generation scenarios of flow constriction and heterogeneity in the fill slopes. The CDG used in the tests was taken from Beacon Hill. Figure 36 shows the geometry adopted. The slope angle was 33o. At 30 g, the model corresponded to a fill slope of 9 m in height, with a vertical fill
72
depth of 3 m. The chosen soil profile for the model fill also represents an idealized case of layering in which the CDG fill material has been sieved and separated into its coarse and fine fractions and placed one on top of the other to form a layered backfill. The layer ends blindly at the toe of the slope to generate elevated transient pore pressures (Take et al. 2004). This ensures that the rate of arrival of the seepage water at the toe greatly exceeds that of the leakage, thereby ensuring a more rapid local transient build up of pore water pressures in this region than would have existed in the absence of layering. The density of the fill material in the first layered slope model was very loose, with an approximate relative compaction of 77%. Figure 37(a) shows the arrival of the transient pore water at the toe of the slope. Once the line source of seepage water was activated, the high transmissivity of the coarse layer quickly delivered water to the toe of the fill slope. As intended, the rate of water transfer into the toe region exceeded the seepage velocity through the model fill, causing a transient increase in the pore water pressure at the toe. The local pore water pressure (PWP) was observed to increase at a nearly constant rate reaching a maximum value of 16 kPa at point B in Fig. 37(a). Since the slope material is dry initially, the current position of the wetting front can be visually observed. As shown in Fig. 38(a), the wetting front has progressed well into the fill material by this point, and at time B (frame (a) in Fig. 38) it is nearing the slope surface. As this seepage front progressed towards the toe, the slope was slowly creeping (Fig. 37(b)). According to Take et al. (2004), the slope mass is observed to accelerate after time B (points B-C on Fig. 37(b)). Digital images at these two instants in time are presented as Figs 38(a) and (b), respectively. As the toe continues to accelerate horizontally, the
Figure 38. Failure mechanism in the loose fill model (modified from (modified from Take et al. 2004).
Figure 37. (a) Measured PWP at the toe of the model slope (PPTT1) and (b) observed soil displacement (Take et al. 2004).
surface of the model fill accelerates towards the toe (Fig. 38(c)) with the velocity increasing to such a point that it exceeds the shutter speed of the camera (Fig. 38(c) and (d)). When the fill material finally came to rest, it formed a low-angle run-out. It is evident from the images that a large portion of soil remains unsaturated throughout the experiment. The initiation of the slide-to-flow mechanism differs from that of the model slope destabilised by downward seepage starting from the crest of the slope in the experiment for the Housing Department in which the slope was not blinded hydraulically at the toe (see Fig. 33). Although the initiation of the two slides is different, no static liquefaction is observed in these two loose fill slopes. Unlike the static liquefaction mechanism of loose sand fill slopes, the non-liquefied slide triggering mechanism is argued to be independent of soil density (Take et al. 2004). In order to verify this hypothesis, the experiment was therefore repeated with a fill compacted to 95% maximum Proctor density while all other factors remained constant in their second test. As before, seepage water was introduced to the crest of the model fill slope and it was quickly transmitted to the toe of the slope, building up localized transient pore water pressures at an identical rate as in the loose fill model (Fig. 37(a)). Since the slope material was initially dry, the position of the wetting front could be observed (see Fig. 39(a)). This dense slope exhibited a much stiffer response to the build up of pore water pressure, with less than one half of the pre-failure displacements signalling the onset of failure (see Fig. 37(b)). Just before reaching the failure pore water pressure, the brittle fill material cracked and water rapidly entered the fill. As high-pressure water entered the crack, the acceleration of the slide increased. The extent to which this crack injected water into
73
Figure 39. Failure mechanism in the dense fill model (modified from Take et al. 2004).
the fill material at time B is shown in Fig. 37(a). After time B, the slope mass accelerated, although at a slower slide velocity than observed in the loose fill slope (points B-C in Fig. 37(b)). As the toe continued to accelerate horizontally, the surface of the model fill accelerated towards the toe (Fig. 39(b)), with the velocity increasing to such a point that it exceeded the shutter speed of the camera (Fig. 39(c)). Eventually, the slope came to rest (Fig. 39(d)). Similar to the shallow loose fill slope, the landslide event triggered from localized transient pore water pressure formed a low-angle run-out. The densification of the fill slope slightly increased the pore water pressure required to initiate failure (see Fig. 37(a)), but it made the failure more brittle (Take et al. 2004). The observed slide-to-flow mechanism in both loose and dense CDG fill slopes (refer to Figs. 38 and 39) may be readily explained by the triaxial element tests on initially unsaturated CDG specimens sheared at a constant net stress ratio ( p/q) as p and q are kept constant while reducing soil suction during a stress path test (refer to Figs. 21 and 25(a)). In each stress path test, a sudden increase in axial strain (similar to soil displacement in Fig. 37(b)) is induced by a gradual reduction in soil suction. Comparing the four stress path tests, a larger reduction in soil suction is needed to initiate a sudden increase in axial strain at a larger constant p test. These element test results are consistent with the measured larger excess pore water pressures at the initiation of the slide-to-flow mechanism in the dense than that in the loose CDG fill slopes (see Fig. 37(a)). Lee et al. (2008) commented that the condition modelled by Take et al. (2004) may be too extreme and such a highly unfavourable condition may not exist in loose fill slopes in the field. Instead, local heterogeneity with layering of limited extent may be more realistic. Therefore, they carried out a similar centrifuge model test on a loose CDG fill slope at 40 g, where a short preferential drainage layer terminating at the mid-height of the slope is simu-
lated (see Fig. 40). The model condition is more realistic in that the soil layering represents any local heterogeneity that originated from the construction process which involves random dumping of fill materials. The geometry and boundary conditions are similar to those adopted by Take et al. (2004). The results of the initially unsaturated loose CDG model slope show that the presence of a permeable layer underneath the loose fill caused instability due to the development of confined water pressure. A slip surface can be identified from the image taken by the high speed camera as shown in Fig. 41. The slope experienced progressive shear failure initiated from the middle of the slope where a hydraulic gradient developed. No sign of liquefaction was observed. Another similar centrifuge model test was also carried out at 40 g by Lee et al. (2008) to investigate the layering effect in an initially unsaturated dense fill slope, where the CDG fill was compacted at a higher dry density of 95% of the maximum dry density. Global instability failure was induced by the increase in pore water pressure at the middle of the slope, where a slip plane could be identified.
15
Coarse Layer Granite Fill
PPT 1 Wooden Bedrock High Speed Digital Camera
100
Web Camera
40
60 85
Drainage
Measurements are in Millimeters
Figure 40. Model package of an instrumented shallow fill slope (Lee et al. 2008).
Figure 41. Observed failure plane of the layered loose fill slope (Lee et al. 2008).
74
Comparing the results of the two centrifuge models, the loose fill requires less hydraulic pressure to initiate a failure and larger displacement is recorded before the onset of failure. The observed difference in hydraulic pressures to initiate a failure in the dense and loose fill slopes is consistent with the measured reduction in soil suctions to induce a sudden increase in axial strain in the stress path tests at different constant net mean stresses shown in Fig. 25. By using the two-dimensional shaking table in the centrifuge (Ng et al. 2004), one-dimensional and two-dimensional earthquakes were fired on saturated loose CDG fill model slopes by up to 0.3 g peak ground acceleration (prototype). Properly scaled viscous pore fluid was used in the tests. The earthquakes did not liquefy the loose slopes but only induced excessive positive pore water pressure and settlement. Based on the observed failure mechanisms in various static and dynamic centrifuge model tests by different research groups at Cambridge and HKUST independently, it is evident that a non-liquefied slide will occur in CDG fill slopes de-stabilised by rainfall and/or rising ground water table, irrespective of compaction density. Static liquefaction is very unlikely to happen in loose CDG fill slopes. The significant difference between the observed physical test results between the loose sand and CDG fill slopes may be attributed to the difference in fine contents, gradation and particle shapes leading to the substantial difference in the liquefaction potential of the two materials as shown in Figs. 22 and 27. Given that a loose CDG fill slope is very unlikely to liquefy, together with the sufficiently large shear strength available after the peak (see Fig. 22(b)) and the strain-hardening behaviour observed during constant water shearing (see Fig. 22(a)), it is evidently reasonable to use soil nails as a stabilisation measure to nail down any possible non-liquefied slides in CDG fill slopes. A design guideline for the use of soil nails in loose fill slopes was published by the Hong Kong Institution of Engineers in 2003 (HKIE, 2003) and it has been widely used in Hong Kong ever since.
saprolitic hillslope at Tung Chung (TC), Lantau Island in Hong Kong by HKUST and Ove Arup and Partners (HK) Ltd. It aims to establish a sitespecific geological and hydrogeological model to capture the hillslope behaviour and the landslide triggering mechanism under seasonal climatic variations (Leung et al. 2010). To assist in the interpretations of the field monitoring data, an in-situ one-dimensional infiltration test and extensive laboratory tests were carried out to determine the hydraulic characteristics and properties of the saprolites. Key procedures and findings are described and discussed in this paper later. 4.1.1 Features of the active landslide mass Fieldworks including boreholes and trial trenches exploration and electrical resistivity surveys are carried out to identify and characterise the features of the active landslide mass in the natural hillslope. Figure 42 shows the identified features of the active landslide mass. A series of sub-parallel tension cracks contouring the hillside of approximately 45 m width are identified at elevations between 84 to 86 mPD (Principal Datum). A number of lateral tension cracks extend north from the main scarp, running down the slope from the eastern and western landslide flanks. Besides this, a major thrust feature is found at about 64 mPD whereas some smaller thrust features are identified on both landslide flanks at about 76 mPD. 4.1.2 Preliminary geological model As revealed from the ground investigations and geophysical surveys, a preliminary simplified geological profile is established and is shown in Fig. 43 for section A-A and B-B. Colluvial materials are encountered at superficial regions to 3 m in depth. The underlying saprolite is typically described as extremely to moderately weak, completely or highly decomposed coarse ash tuff (CDT/HDT). Slightly or moderately decomposed coarse ash tuff (SDT/MDT) is then encountered beneath the saprolites. Close to medium spaced rock joints are kaolin infills. As shown in Fig. 43(b), a shallow decomposed rock stratum dipping towards the east is identified across the hillslope at about 75 mPD. The base groundwater profile right after instrumentation is generally at about 10 m below ground level (bgl). The profile approximately follows that of the MDT material and is about 1 to 2 m above the decomposed rock head. The inferred rupture surfaces are also shown for reference. The active landslide mass undergoes a retrogressive translational-slide type of failure (Varnes 1978) of limited mobility along some poorly defined rupture surfaces at 3 to 5 m below ground level, which are just beneath the colluvium/saprolite boundary.
SOA: FIELD MONITORING AND TESTING OF A NATURAL SAPROLITIC HILLSLOPE IN HONG KONG Background
4.1
Supported by the Geotechnical Engineering Office of the Civil Engineering and Development Department of the HKSAR, a joint governmentindustry-university full-scale field monitoring programme was recently implemented in a volcanic
75
A Approximate location of the IP test (Ng et al. 2010) N
Lateral tension cracks EPC 2 (A/B) Thrust features TC IPI 1

70 mP
CP 2 (A/B) CP 3 (A/B)
IPI 2
80 m
JFT/TDR RG
PD
B
CM A EPC 1 (A/B)
Tension crack
90
mP
Figure 42. Identified active landslide body and locations of instruments in Tung Chung of Lantau Island (modified from Leung et al. 2010).
4.1.3 Research strategy Prior to investigating the hillslope response under seasonal climatic variations, it is essential to determine and characterise the properties of the geomaterials in assisting the interpretation of the field testing and monitoring data. The hydraulic properties, i.e., stress-dependent soil-water characteristic curve (SDSWCC) (Ng & Pang 2000) and permeability function, k(), of an unsaturated saprolites are the most relevant parameters when dealing with seepage problems in a hillslope. A brief testing programme of this research is summarised below. A series of laboratory tests are first conducted. The SDSWCCs of both compacted and intact specimens are measured using a modified pressure plate extractor. On the other hand, the k() of a compacted specimen is measured using a newly-
developed 1 m high soil column apparatus. As part of the joint government-industry-university project, the Instantaneous Profile (IP) test (Ng et al. 2010) is then carried out near the slope toe (see Fig. 42) to determine the in situ SDSWCCs and k() of the ground. Finally, a full-scale field monitoring with comprehensive instrumentation scheme on the natural saprolitic hillslope is implemented. 4.2 Stress-dependent soil-water characteristics of unsaturated CDT 4.2.1 Test material The tested material is light-grey, dappled light brown completely decomposed coarse ash crystal tuff (CDT), which is sampled from Tung Chung (TC) on Lantau Island. The measured index
76
Elevation (mPD) 120 115 110 105 100 95 90 85 80 75 70 65 60 55 50 120 115 110 105 100 95 90 85 80 75 70 65 60 55 50 0 0
(a)
N
Inferred rupture surface Slope angle = 28 CM EPC1
JFT, TDR CP 3 (A/B) TC IPI1
Thrust feature
BH 6 Initial groundwater table right after instrumentation IPI2 EPC2
CP 2 (A/B) MDT or SDT BH 3 HDT

CDT with corestones
Colluvium
10 (b)
20
30
40
50
60
70
80
90
100
Inferred rupture surface
CP 3 (A/B)
Colluvium
CDT with corestones
Shallow, dipped, weathered rock stratum

MDT or SDT
HDT
Postulated cross-slope groundwater flow
Initial groundwater table right after instrumentation
10
20
30
40
50 Distance (m)
60
70
80
90
100
Figure 43. Preliminary geological model for (a) Section A-A and (b) Section B-B and the arrangement of instruments (modified from Leung et al. 2010).
properties are summarized in Table 4. According to the Unified Soil Classification System (USCS) (ASTM 2000), the CDT may be described as inorganic silty clay of low to medium plasticity (CL). Figure 44 shows the particle-size distribution of the CDT.
To prepare compacted specimens, each soil sample is first oven-dried for 48 hours. After removal from the oven, those particles with one dimension larger than 2 mm are discarded by dry sieving. Water is then added to achieve the water content of 17.3%. After mixing, each soil sample is kept in a plastic
77
Table 4. Index
Index properties of CDT (Leung et al. 2010). CDT 2.68

3
Specific gravity Maximum dry density (kg/m ) Optimum moisture content (%) In situ moisture content (%) Sand content (2 mm, %) Silt content (63 m, %) Clay content (2 m, %) Liquid limit (%) Plastic limit (%) Plasticity index (%)
Humidity probe Fan Salt solution Specimen
1552 17.2 17.3 35.2 42.3 22.5 34.2 20.2 14.0

Stirrer
Electronic balance
Figure 45. General layout of the experiment setup for suction control using RH-controlled technique (Leung & Ng 2009).
100 90 80 Percentage passing (%) 70 60 50 40 30 20 10 0 0.001 0.01 0.1 1 Particle size (mm) 10 100
Colluvium CDT
Figure 44. Particle size distributions of the colluviums and CDTs sampled from Tung Chung (TC) of Lantau Island (Leung et al. 2010).
bag for moisture equalization in a temperature- and moisture-controlled room. Each soil sample is then compacted in layers to the desired dry density of 1552 kg/m3 in a 70 mm diameter by 19 mm high oedometer ring by the static compaction method (Ng & Chiu 2001). Each intact specimen obtained from block samples is carefully prepared using the 70 mm diameter by 19 mm high oedometer ring. The preparation process is carried out in the moisture and humidity controlled room at HKUST to minimize moisture loss (Ng et al. 2004). 4.2.2 Apparatuses Although it is generally recognised that unsaturated soil behaviour and properties are governed by the two constitutive variables (matrix suction and net stress) (Gens 2009), effects of net stress on SWCCs have often been ignored. At HKUST, a pressure plate system has been developed in measuring SDSWCCs under the one-dimensional (1D) (or K0) stress condition (Ng & Pang 2000). The
axis-translation technique (Hilf 1956) is adopted to control matrix suction in an unsaturated specimen. This apparatus is designed to overcome some limitations of existing volumetric pressure plate extractors by allowing vertical net stress to be applied to a soil specimen and its vertical displacement to be measured during a test. The matrix suction controlled by the above apparatus is often less than or equal to 1500 kPa due to the limitation of the AEV of the ceramic plate. In order to measure soil-water characteristics at high suctions (i.e., 1500 kPa), relative humidity (RH)-controlled technique can be used to control soil suction up to 320 MPa (Leung & Ng 2009). Figure 45 shows the general layout of the experimental apparatus. The assembly is placed in the temperature- and moisture-controlled room at HKUST to minimize temperature fluctuation. 4.2.3 Influence of net stress To investigate the influence of net stress, ua ( is total stress) on the soil-water characteristics of the CDT, SDSWCCs under three different 1D stress levels (0, 40 and 80 kPa) are measured. After the specimen preparation, each soil specimen is pre-consolidated to a desired stress level and a drying-wetting cycle is then applied. Figure 46 shows the measured drying and wetting SDSWCCs of each compacted CDT specimen. It can be seen that the water retention capability of the specimen loaded at net stress of 40 kPa is greater than that at zero stress. Also there is a decrease in rates of desorption and adsorption and a reduction in size of hysteresis loop as the net stress is increased from zero to 40 kPa. Nonetheless, the AEV does not seem to have been affected by the stress level. These observations are likely to be caused by the presence of a smaller average pore-size distribution in the soil specimen under
78
0.5 CDT-TC-0 kPa-comp CDT-TC-40 kPa-comp CDT-TC-80 kPa-comp
0.5 Volumetric water content 0.4 0.3 0.2 0.1 0.1 CDT-TC-40 kPa-comp CDT-TC-40 kPa-intact
Volumetric water content
0.4
0.3
0.2
0.1
0.1
10 100 Matrix suction (kPa)
1000
10 100 Matrix suction (kPa)
1000
Figure 46. Measured SDSWCCs of the compacted CDT specimens at 1D stress levels of 0, 40 and 80 kPa.
Figure 47. Comparisons between the measured drying and wetting SDSWCCs between the compacted and the intact CDT specimens loaded at 1D net stress of 40 kPa.
0.05 0.04 Gravimetric water content 0.25 0.20 0.15 0.10 0.05 0.00 0.03 0.02 0.01 0.00 1000 10000 100000 1000000
higher applied vertical stress (Ng & Pang 2000). However, as the applied stress further increases from 40 to 80 kPa, both measured drying and wetting SDSWCCs at these two stress levels nearly overlap with each other. It is evident that the influence of net stress increasing from 0 to 40 kPa and from 40 to 80 kPa on the soil-water characteristics of the compacted CDT specimen is different. This might be attributed to the significant increases in soil stiffness from 0 to 40 kPa. Comparisons between compacted and intact CDT specimens Both drying and wetting SDSWCCs of the intact CDT specimen are measured to study the differences of water retention capability between the compacted and the intact specimen under a given stress level. Figure 47 compares the measured drying and wetting SDSWCCs of the compacted and the intact CDT specimens loaded to the same vertical net stress of 40 kPa. At a given suction, the intact specimen has less water retention capability than that of the natural specimen, where both rates of desorption and adsorption are substantially higher for matrix suctions up to 400 kPa. As expected, the AEV and the size of the hysteresis loop seem to be smaller for the intact specimen. Due to various geological processes, such as leaching, in the field, the intact specimen is believed to have relatively non-uniform pore size distributions (Ng & Pang 2000). As the two specimens have the same initial density, it may be reasonable to postulate that the intact specimen would have some larger pores than that of the compacted specimen, at least statistically. Thus, the intact specimen has a slightly lower AEV and a higher rate of desorption and adsorption. 4.2.5 The measured full suction range SWCC By combining the measured SWCC at low and high suctions, the full suction range drying and 4.2.4
0.30
0.1
10
100
1000
10000
100000 1000000
Soil suction (kPa)
Figure 48. Full suction range drying and wetting SDSWCC at zero stress of the compacted CDT specimen (modified from Leung & Ng 2009).
wetting SWCCs of the compacted CDT specimen at zero stress are obtained and are shown in Fig. 48. The solid and open symbols refer to the drying and wetting SWCC, respectively. Since the data measured at high suctions does not consider volume changes, the water content of the SWCCs is expressed in terms of GWC, instead of VWC. For illustration purposes, lines are added to join the measured SWCC at low and high suctions. It can be seen from the inset that approximately 2% of GWC is retained in the specimen after experiencing a drying cycle at soil suction of 320 MPa. Similar to those observed in the SWCC at low suctions, hysteresis is observed between the drying and wetting SWCC at high suctions. 4.3 Unsaturated permeability function of CDT Direct measurement of k() of an unsaturated soil is often tedious and time-consuming. The stress effects on k() are generally not considered. A new 1 m high stress-controllable soil column is developed to measure SDSWCC and k() of soils using the transient-state method directly (Ng & Leung
79
2010). The influences of the two constitutive variables (matrix suction and net stress) and a dryingwetting cycle can be studied. 4.3.1 Theoretical considerations The Instantaneous Profile Method (IPM) (Watson 1966) is adopted to determine k() of an unsaturated soil at different vertical net stresses. The calculation procedures are briefly summarised below. Figure 49 (right hand side) shows two arbitrary profiles of VWC, w, and hydraulic head at elapsed time t t1 and t2 along a one-dimensional soil column. The values of VWC and PWP can be measured using different types of instruments at zA (Row A), zB (Row B), zC (Row C) and zD (Row D). The measured VWC profiles may be extrapolated to the surface and the bottom of the soil column for the determination of water flow rate. According to the one-dimensional continuity equation, the water flow rate, vzB,tave, at any depth z (zB in this example) for any average elapsed time tave (t1 t2)/2 can be determined by estimating the change of total water volume between the depth under consideration, i.e., zB, and one end of the soil column, i.e., ze: vzB ,tave d zB V w ( z,t ) dz vze,tave dt ze t2 t1 (4)
On the other hand, the hydraulic head gradient, izB,tave, at any depth zB for any average elapsed time tave can be obtained by estimating the slope of a hydraulic head profile, which is the summation of measured PWP head profile and gravitational head profile. It can be expressed mathematically as follows: 1 dhzB ,t1 dhzB ,t 2 izB ,tave 2 dz dz 1 hzA,t1 hzB ,t1 hzB ,t1 hzC ,t1 zB zC 4 zA zB hzA,t 2 hzB ,t 2 hzB ,t 2 hzC ,t 2 zB zC zA zB (5)
where hzi,tj is the hydraulic head at depth zi (i A, B, C and D) for elapsed time tj (j 1, 2). Hence, the unsaturated permeability, kzB,tave, at any depth zB for any average elapsed time tave can be calculated by dividing the water flow rate by the corresponding hydraulic head gradient: kzB ,tave vzB ,tave izB ,tave (6)
where w(z, t) is the VWC profile as a function of depth z at specific time t; V is the shaded area between the w(z, t) at t t1 and t2 as shown in Fig. 49; dt is the time interval between the measurements, i.e., t2 t1; and vze,tave is the boundary water flow rate.
4.3.2 Newly designed apparatus, test material and sample preparation Figure 49 shows the schematic diagram of the newly-developed stress-controllable soil column. The top and the bottom boundary flow conditions can be controlled and measured. At the top boundary, a 150 mm in diameter, 20 mm thick, circular, perforated, stainless-steel plate is placed on the top
Water storage tank
All dimensions are in mm

R5
Perforated metal plate 150

1
Seating for loading ram Loading frame Air supply
To constant head water supply system
Position of water storage tank during saturation stage for bottom-up saturation
1
Load cell Hole Ponding head
500
Pneumatic actuator
10 200 325 Section 1-1

Perforated metal plate
Extrapolated profiles
w
Row A zA
700
25
t = t1
t = t2
t = t1 t = t2
hw
125
500
Tensiometer
160 150
Thetaprobe Soil column Acrylic Cylinder (5 mm thick)
Row B zB
dhz ,t1 dhz ,t 2 dz dz
825
375
V
O ring
Section A-A 125 500
Row C zC Row D zD
A
Stand
Water storage tank Electronic balance Perforated metal plate Valve
ze
Filter paper
Extrapolated profiles
Datum
Figure 49.
Schematic diagram of the stress-controllable soil column (modified from Ng & Leung 2010).
80
of a compacted soil column so that vertical stress and boundary flux can be applied independently. Upon ponding, constant head infiltration can be achieved and the infiltration rate can be measured by a constant head water supply system. At the bottom of the cylinder, another circular perforated plate is placed, which can be completely filled with water to form a water compartment to allow for uniform drainage. A valve is installed to control the bottom drainage condition. Vertical load can be applied using a pneumatic actuator and is recorded by a load cell. As illustrated in Fig. 49, four pairs of miniature-tip tensiometers (Soilmoisture Equipment Corp. 2005) and theta-probe soil moisture probes (Delta-T Devices Ltd. 1999) are used to measure negative PWP and VWC, respectively at four levels, i.e., Row A, Row B, Row C and Row D. The tested material is the CDT taken from Tung Chung in Hong Kong. Figure 44 shows its particle size distribution and Table 4 summarises the measured index properties. For sample preparation, the stainless steel plate is first placed at the bottommost part of the cylinder and is filled with de-aired water. A saturated filter paper is then placed on the top of the plate for uniform drainage. Each soil column is compacted at GWC and dry density of 17.2% and 1550 kg/cm3, respectively, by dynamic compaction method. The compaction is divided into 20 layers, where the compaction effort is identical for each layer. 4.3.3 Test programme and test procedures In total, there are two tests using the soil column (SC) reported in this study. The targeted vertical net stresses for these two tests are 0 kPa (SC0) and 40 kPa (SC40). After sample preparation, each compacted soil column is first subjected to bottom-up saturation (see Fig. 49) where the bottom valve is opened. Desired vertical load is then applied on the top of each compacted soil column. The preconsolidation stage is considered to be completed when (i) all tensiometers record zero readings and; (ii) the change of water outflow rate is less than 15 cm3/day for at least 24 hours, which is equivalent to an average change of GWC of about 0.06%/day (Ng & Leung 2010). After pre-consolidation, the bottom valve is closed to achieve the zero flux boundary condition while the surface of each soil column is allowed to evaporate naturally. The evaporation stage is terminated when the tensiometers at Row A (see Fig. 49) record a PWP of 80 kPa to minimise the effects of cavitation. After the evaporation stage, about 50 mm constant head ponding is subsequently applied on each column surface by the constant head water supply system. The bottom valve of
the soil column is re-opened to drain water. During both evaporation and ponding stages, the variations of PWP and VWC profiles with time are continuously recorded along each soil column. The laboratory measured results of stress-dependent permeability functions of CDT are compared with in-situ measurements later in this paper (Ng & Leung 2010). 4.4 The Instantaneous Profile (IP) test
4.4.1 Description of test site, instrumentation scheme and test procedures The IP test is carried out near the slope toe of the saprolitic hillslope at Tung Chung on Lantau Island (see Fig. 42). Ground investigation reveals that about a 1 m thickness of loose colluvium (Col) deposits on the ground surface. A layer of about 2 m CDT is then successively encountered and some relict joints with silty clay infill are identified. The CDT overlays moderately and slightly decomposed tuff at further depths. The groundwater table is found to be located at about 3 m below ground surface. Figure 50 shows the schematic setup of the IP test and the instrumentation scheme (Ng et al. 2010). A flat test plot of 3.5 m 3.5 m is formed by cutting into the ground at the test location. Ten tensiometers are installed to record PWP while four time-domain reflectometers (TDRs) are installed to measure volumetric water content (VWC or w). To achieve the one-dimensional downward water flow assumption, a 3 m deep, 1.2 m wide trench is excavated at the uphill side of the test plot to install some polythene sheeting, which acts as a cut-off to lateral ground water flow. The trench is backfilled after the installation. Subsequently, a circular steel test ring (3 m in diameter) is installed to the ground surface and embedded 100 mm into the flattened plot to retain water for the IP test. To investigate the influences of wetting-drying cycles on in situ SDSWCCs and k() of the ground, the test programme is divided into four stages, which are summarized in Table 5. For each ponding
500 200
3500
Test ring, diameter = 3m Top Soil Colluvium CDT 50 JFT_2 JFT_3
2900
JFT_1 TDR_1 JFT_4 JFT_5 TDR_2 3000 45 o
2000
Polythene sheeting Compacted Fill
JFT_6 JFT_7 JFT_8 JFT Depth: TDR_3 JFT_1 - 0.36m

JFT_2 - 0.77m JFT_3 - 0.95m JFT_4 - 1.17m JFT_5 - 1.54m JFT_6 - 1.85m JFT_7 - 2.13m JFT_8 - 2.43m JFT_9 - 2.60m JFT_10 - 2.99m
1200
JFT_99 All TDRs and JFTs are located at 150mm and 600mm from the center of TDR_4 the test ring, respecTDR Depth: tively. All dimenTDR_1 - 0.835m sions are in mm TDR_2 - 1.853m unless otherwise TDR_3 - 2.500m state/ JFT_10
TDR_4 - 3.585m
Figure 50. The IP test set up and the instrumentation plan (modified from Ng et al. 2010).
81
Table 5. IP test schedule (modified from Ng et al. 2010). Stage 1 (Day 04) 2 (Day 428) 3 (Day 2832) 4 (Day 3248) Phase 1st Wetting 1st Drying 2nd Wetting 2nd Drying Date 26 Oct 2007 630 Oct 2007 Duration
Depth (m)
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0

Day 0 (initial) Day 4 (1W) Day 10 (1D) CDT Day 28 (2W) Day 32 (2W) Day 35 (2D) Day 48 (2D) Hydrostatic line
Colluvium
4 days 24 days
30 Oct3 Nov 4 days 2007 319 Nov 2007 16 days
(a)
5 0 -5
-10 -15 -20 -25 -30 -35 -40 -45

Pore-water pressure (kPa)
Day 0 Day 4
0.0 0.5 1.0 1.5 2.0 2.5 3.0 3.5 4.0 25
Day 10 Day 28 Day 32
Day 35 Day 48
Colluvium
phase, a water level of about 0.1 m is applied on the ground surface inside the ring and the water level is checked and refilled to the same level every 12 hours. For each drainage phase, the test plot is allowed to dry under natural evaporation. Responses of the PWP and the VWC profiles Figure 51(a) shows the measured PWP profiles on several key days. The hydrostatic line is also included for reference. 1W, 1D, 2W and 2D denote 1st wetting cycle, 1st drying cycle, 2nd wetting cycle and 2nd drying cycle, respectively. At the beginning of the test, the measured PWPs are fairly constant in the top few metres, which may be attributed to the antecedent rainfalls in September. A positive PWP of 4 kPa was registered near the interface between the colluvium and the underlying CDT, which is probably due to the differences in the permeability of the two strata. The measured in situ k() are discussed later. Upon the first wetting (Day 0 to Day 4), the PWP profile remains almost unchanged, as a result of the wet ground due to the antecedent rainfall. When the drying phases commence (Day 4 to Day 28), the PWPs generally decrease at all depths due to the continuous loss of water content. At the end of the drying period (Day 28), the PWP at the colluvial stratum reaches its minimum value of 42 kPa. On the other hand, the measured PWPs remain larger than hydrostatic pressure at depths between 0.77 m and 2.13 m (CDT) while they are hydrostatic at deeper levels (i.e., 2.43 to 2.99 m). Moreover, the observed delayed response between the colluvium and the CDT shows that the influence of wetting drying cycles on the ground decreases with depth. Similar trends are also observed in the second wetting-drying cycle. Figure 51(b) shows the measured w profiles on several key days. The intersection between the w profiles on Days 0 and Day 4 indicates that w increases due to the downward flux in the ground above 2.50 m while the water drains downward below this level because of gravity. As expected, the entire w generally increases and decreases during each wetting and drying period respectively. 4.4.2
Depth (m)
CDT
(b)
30 35 Volumetric water content (%) 40
Figure 51. Variation of measured (a) PWP profiles and (b) VWC profiles during the IP test (Ng et al. 2010).
Between depths of 0.85 and 1.85 m, the reduction in w due to drying is much more significant than that at greater depths. This suggests that the depth of influence due to evaporation is less than 2 m, consistent with the measurements from the tensiometers (see Fig. 51(a)). 4.4.3 In situ measured SDSWCCs By relating the measured w and the measured matrix suction (negative PWP), in situ SDSWCCs of each material at various depths can be obtained (see Fig. 52). Within the colluvial stratum at 0.36 and 0.95 m (see Figs. 52(a) and (b)), the soils appear to start desaturation beyond a matrix suction of 1 kPa, i.e., AEV is about 1 kPa. At a depth of 0.36 m, the reduction of w is negligible when the matrix suction is increased beyond 5.6 kPa in the first drying curve. Noticeable hydraulic hysteretic loops during wetting-drying cycles are observed and the hysteretic loop size decreases with depth. It is likely to have been affected by the heterogeneous nature of the material with possible reduction of the average pore size but greater uniformity of individual pore sizes in deeper soil within the colluvium layer. Similar to those measured in the colluvium, the size of the measured hysteretic loops decreases with depth (see Figs. 52(c) and (d)). Considering the fact that the ground has been subjected to countless wetting and drying cycles in the past, it is not surprising to find that the characteristics of in situ SDSWCCs obtained from the first wetting-drying cycle are comparable to those from the second cycle (Ng & Pang 2000) for both the colluvium and the
82
CDT strata. Below 1.85 m, the change in w with matrix suction and the hysteresis phenomenon of soil-water characteristics are essentially negligible. 4.4.4 In situ measured permeability functions Based on the monitoring results (Fig. 51), the k() of each material is determined using equations (4) to (6). Figure 53 shows the measured k() at depths of 0.84 and 1.85 m for the colluvium and the CDT, respectively. The range of saturated permeability of the colluviums and the decomposed volcanic (DV) soil recorded in the Mid-levels area in HK is also shown for comparison (GCO 1982). In the colluvial stratum, the permeability at suctions lower than 1 kPa falls within the range of saturated permeability. The measurement range of the k() along the wetting path is small, probably due to errors introduced by the limited resolution of the tensiometer at low matrix suction. It appears that the measured k() is very sensitive to suction from 0 to 4 kPa, resulting in a decrease of permeability from 1 104 to 2 107 m/s.
40
40
In the CDT, the measured data are quite scattered, probably due to the wide spacing of the installed instruments. Nevertheless, the measured permeability along the wetting path is higher than that from the drying path at a given suction. The measured k() in the colluvium is generally higher than that in the CDT along the wetting-drying path, indicating the less permeable nature of CDT, which might have led to the formation of a perched water table at the colluvium/CDT interface (see Fig. 51(a)). 4.4.5 Comparisons between the fieldand laboratory-measured permeability functions The laboratory-measured drying and wetting k() of the compacted CDT at vertical net stresses of 0 kPa and 40 kPa (SC) are shown in Fig. 53 for comparisons, where the solid and open symbols refer to drying and wetting k() respectively. The saturated permeability, ks, at each vertical net stress is independently measured in a triaxial apparatus and is also shown in the figure for reference. Based on the PWP and the VWC measurements, unsaturated permeability of each soil column is determined using the equations (4) to (6). At any applied vertical net stress, the laboratory measured drying permeability decreases log-linearly as matrix suction increases. For an example, the reduction of drying permeability at constant zero vertical net stress is up to two orders of magnitude as matrix suction increases from 0 kPa to 80 kPa. At a matrix suction of 6 kPa, the maximum reduction in drying permeability is about one order of magnitude (from 2 108 m/s to 5 109 m/s) when the vertical net stress increases from 0 kPa to 40 kPa. Moreover, at a given change of matrix suction (from 4 to 80 kPa), the decreasing rate of drying permeability reduces with an increase in vertical net stress. The observed stress dependency of the drying k() is attributed to the increase of soil density (or decrease in void ratio) when a vertical net stress is applied. At a given vertical net stress, remarkable hysteresis loops are identified between the drying and the wetting k() from the laboratory experiments. The observed hysteretic behaviour is probably due to the difference in receding and advancing contact angles of the soil-water interface (Hillel 1982) during a drying and wetting cycle in the soil column. Although the material tested in the laboratory (CDT) is different from that measured in situ (colluvium), the comparison in Fig. 53 is reasonable since both materials are described as silty clay in accordance to the USCS. It can be seen that the field-measured k() of the colluvium is significantly higher than the laboratory-measured one by about three orders of magnitude for suctions ranged from 0.1 to 2 kPa. Moreover, the in
35
Volumetric water content (%)
30
1st wetting 1st drying 2nd wetting 2nd drying
35
30
25
(a)
20
0.36 m
25
Collu viu m
20
(b)
40
0.95 m
Collu viu m
40
35
35
30
25
(c)
20
0.1
1.85 m
1 10
1 wetting 1st drying 2nd wetting 2nd drying
st
30
25
CDT
20
(d)
100
0.1
2.99 m
1 10
CDT
100
Matrix suction (kPa)
Matrix suction (kPa)
Figure 52. In situ measured SDSWCCs at (a) 0.36 m, (b) 0.95 m, (c) 1.85 m and (d) 2.99 m (modified from Ng et al. 2010).
110-3 110-4 Permeability (m/s) 110-5 R2 110-6 110
-7
R1
IP-Col-0.84m-1W IP-Col-0.84m-1D IP-CDT-1.85m-1W IP-CDT-1.85m-1D SC-CDT-0kPa-Dry SC-CDT-0kPa-Wet SC-CDT-40kPa-Dry SC-CDT-40kPa-Wet
ks at 0 kPa Decreasing Rate
110-8 110-9 110

-10
ks at 40 kPa Increasing Rate
1 10 100 Matrix suction (kPa) R1: range of saturated permeability of colluvium (2 10-6 to 9 10-4 m/s) in accordance to GCO (1982) R2: range of saturated permeability of DV (3 10-6 to 9 10-6 m/s) in accordance to GCO (1982)
0.1
Figure 53. Field- and laboratory-measured k() of the colluvium and the CDT (modified from Ng et al. 2010 and Ng & Leung 2010).
83
situ permeability appears to reduce rapidly when matrix suction reaches 2 kPa while the laboratory set-up decreases at a much slower rate beyond a matrix suction of 10 kPa. Due to the presence of cracks, fissures and relict joints in the highly heterogeneous colluviums in situ, preferential water flow is therefore likely to occur, resulting in higher unsaturated permeability. Similar to the observations from the colluvium, the field-measured k() of the CDT is also higher than the laboratory set-up by about three orders of magnitude for matrix suction ranged from 0 to 5 kPa, though the stress level between the two measurements is comparable. 4.4.6 Back-analysis of the IP test The presence of geological features like stratification and discontinuities that are likely to be found in the hillslope (Corley et al. 1999) often induce permeability heterogeneity of a soil matrix. Hence, this may hinder the interpretations of fieldmeasured data. Numerical back-analysis of the IP test is therefore carried out to further improve the understanding of the monitoring results (Leung & Ng, 2010). The two parameters kx/ky (principal permeability ratio) and (principal flow orientation), which control the permeability heterogeneity of an unsaturated soil, are examined to investigate their influence on PWP distribution. Figure 54 shows the finite element mesh of the saprolitic hillslope simplified from Fig. 50. To incorporate the stress effect on the hydraulic properties, the CDT is divided into two layers (CDT1 and CDT2) according to their approximate stress levels. The bedrock is assumed to be impermeable and a zero flux bottom boundary is applied on boundary AF. A constant total head of 24 m and a reviewed zero flux (SEEP/W, 1998) is specified on the boundary EF and the boundary AB respectively. To resemble the initial PWP profiles and groundwater table before the IP test, the antecedent rainfalls recorded in September 2007 are simulated
by carrying out a transient seepage analysis. The IP test is then simulated, following the procedures summarised in Table 5. It should be noted that, for simplicity, natural evaporation is not considered in this study. The wetting SDSWCC of intact colluvium (Col) and CDT specimens obtained in the vicinity of the test plot are measured using the one-dimensional stress-controllable pressure-plate extractor (Ng & Pang 2000). The colluvium is tested under net normal stress of 0 kPa (Col) while the CDT is measured under net normal stresses of 40 kPa (CDT1) and 80 kPa (CDT2). The k() of each material is then predicted using the method proposed by Fredlund et al. (1994). For simplicity, the hysteretic behaviour of both the SDSWCC and k() of each material is not considered. Regarding the PS, it is assumed that it does not have water retention capability, i.e., w 0, while permeability to the order of 34 m/s is constant for the entire suction range. Parametric study shows that the overall seepage characteristic in the hillslope can be satisfactorily mimicked using a particular set of kx/ky and for each material (see Fig. 55). The parameters used are shown in Fig. 55. Based on the back analyses, it may be reasonable to postulate that branches of discontinuities dipping at 30 on average exist within the CDT stratum at a 1 to 3 m depth in this saprolitic hillslope. Lateral seepage along these preferential flow paths thus reduces the downward seepage and
Pore-Water Pressure (kPa) -20 -10 0 (a) 0 10 20 -20 -10 0 (b) 1 0 10 20
colluvium kx/ky = 10 and = 0
colluvium kx/ky = 10 and = 0 CDT1 kx/ky = 10 and = 30
CDT1 kx/ky = 10 and = 30
Polythene sheeting, PS C A D
Col 1m 2m CDT1 1m E Col-N-0-D CDT-N-40-D CDT-N-80-D
A CDT2
3 CDT2 CDT2 kx/ky = 1 kx/ky = 1 Day 4-Field Day 0-Field and and Day 4-Computed Day 0-Computed = 0 = 0 Hydrostatic line Hydrostatic line 4 4 -60 -40 -20 0 20 -20 -10 0 10 20 0 0 (c) (d) colluvium kx/ky = 10 colluvium and = 0 k /k = 10 and = 0 1 x y 1 CDT1 kx/ky = 10 and = 30 CDT1 -20 -10 0 0 2 kx/ky = 10 2 and = 30 1 3 3 CDT2 2 kx/ky = 1 Day 28-Field and Day 28-Computed = 0 Hydrostatic line 3 CDT2 kx/ky = 1 Day 48-Field and Day 48-Computed = 0 Hydrostatic line
Y Slope angle = 30
Y X X
Depth (m)
Figure 54. Finite element mesh of the idealised hillslope (Leung & Ng 2010).
Figure 55. Comparisons between the measured and the computed PWP profile on (a) Day 0 (after antecedent rainfall), (b) Day 4 (end of 1 W), (c) Day 28 (end of 1 W) and (d) Day 48 (end of 2 W) (modified from Leung & Ng 2010).
84
100 80
15 10 5 0
JFT_0.5m JFT_1.5m
60 40 20 0
4.5
Full-scale field monitoring of an unsaturated saprolitic hillslope

-5 -10
40
After characterising the in situ hydraulic properties of the ground and having preliminary understanding of the hillslope responses from the IP test, a full-scale, comprehensive field monitoring programme on the saprolitic hillslope is implemented (see Figs. 42 and 43). By investigating the hillslope responses under 2-year seasonal climate variations, a site-specific geological and hydrogeological model may be established to improve the understanding of the landslide triggering mechanism (Leung et al. 2010). 4.5.1 Instrumentation programme The comprehensive instrumentation scheme is designed to measure changes of the two constitutive variables (i.e., matrix suction and net stress) and their influences on the unsaturated ground. The measurements include PWP, VWC, subsurface total horizontal stress and the horizontal displacement. The instruments used consist of jet-filled tensiometers (JFTs), time-domain reflectometers (TDRs), heat dissipation matrix water potential sensors (TCs), vibrating-wire earth pressure cells (EPCs), in-placed inclinometers (IPIs), Casagrande-type piezometers (CPs), vibrating-wire crackmeter (CM) and a tipping-bucket rain gauge (RG). The arrangement of the instruments is schematically shown in Figs. 42 and 43. Infiltration characteristic and groundwater condition Within the monitoring period from November 2007 to January 2009, the highest monthly rain depth of 1391 mm was recorded in June 2008, which was 45% of the annual rainfall in 2008. The peak 4-hour rolling rainfall of about 323.6 mm was recorded on 7 June, where the estimated return period is about 245 years. The hillslope responses between the periods from 5 to 9 June 2008 are selected for assessments. Figures 56(a) and (b) show the measured PWP and VWC, respectively, at depths of 0.5, 1.5 and 2.5 m, while Figure 56(c) illustrates the variations of some selected piezometric head with time. The laboratory-measured saturated VWC of 32% (see Fig. 47) and location of the response zone of each CP is also shown in Figs. 56(b) and (c), respectively, for reference. The initial measured small 4.5.2
(b)
TDR_1.5m
140 120 100 80

Saturated VWC
35
TDR_0.5m
60 30
TDR_2.5m
40 20
25
78 76
0
140
(c)
CP 3A (3.7 to 5m)
120 100 80
Piezometric level (mPD)
74 72
CP 3B (8.7 to 10 m)
70 68 66 64
CP 2A (3.7 -5 m) CP 2B (7 to 8.4 m)
60 40 20 0
5 June
6 June
7 June Date
8 June
9 June
Figure 56. Variation of (a) PWP, (b) VWC and (c) piezometric level with time in 2008 (modified from Leung et al. 2010).
amount of positive PWPs and relatively high VWC on 5 June (see Figs. 56(a) and (b)) is likely to be a result of the antecedent rainfalls that happened on 19 April 2008, where a peak rainfall intensity of 62 mm/hr was recorded. When the peak hourly rainfall intensity of 133.5 mm occurred on 7 June, PWPs increased rapidly and maximum positive PWP of 25 kPa was recorded at about 2.5 m depth by JF_2.5 m (see Fig. 56(a) and responses of CP 2 A and CP 3 A in Fig. 56(c)). On the other hand, the VWCs attained and maintained maximum value of 36% (see Fig. 56(b)). These observations seem to suggest a transient perched groundwater table developed within the bouldery colluvial deposit at the top 3 m of the ground. This is consistent with the observation from the IP test, that positive PWPs are recorded near the colluvium/ CDT interface (see Fig. 51), resulting from the substantial lower unsaturated water permeability of the underlying CDT strata (Ng et al. 2010; see Fig. 53). Besides this, the main groundwater table is found to rise by about 5 to 6 m from its initial position (see Fig. 43(a) and responses of CP 2B
85
Rainfall intensity (mm/hr)
Rainfall intensiry (mm/hr)
hence limits the GWT rise. Nonetheless, the numerical simulations are highly idealised, where actual topography and stratification of the natural hillslope may be oversimplified. Moreover, preferential water flows as a result of the presence of complex crack networks in both soil strata should be incorporated using advanced numerical tools.
30 25 20
140
(a)
JFT_2.5m
120
and CP 3B in Fig. 56(c)), approaching the colluvium/CDT interface. 4.5.3 Response of in situ total horizontal stress Figure 57 shows the measured in situ total horizontal stresses recorded by EPC2 at 2 m depth during the rainfall events from 5 to 9 June. The total horizontal stresses in down-slope and crossslope direction were measured. The PWP measurements recorded by JFT_1.5 m at 1.5 m depth (Fig. 56(a)) are also shown for comparison. The initial total stress difference of 5 kPa between down-slope and cross-slope direction is likely due to the increase of earth pressure resulting from the permanent down-slope ground movement after the heavy rainstorm on 19 April 2008. When the peak rainfall intensity occurred on 7 June, the maximum total horizontal stress recorded in down-slope and cross-slope direction was 28 and 15 kPa, respectively, and the stress difference between down-slope and cross-slope direction became larger (15 kPa) from 7 to 9 June. Moreover, it can be seen that the measured PWP nearly overlaps the total horizontal stress in B direction. This indicates that the total horizontal stress change in cross-slope direction at 2 m depth is likely due to the increase of positive PWP instead of the earth pressure itself. In other words, the observed increase of the total stress difference between down-slope and cross-slope direction (from 5 to 15 kPa) before and after the heaviest rainfalls may be attributed to significant downslope ground movement (see Fig. 58(a)). 4.5.4 Deformation characteristic of the hillslope Figures 58(a) and (b) show the measured horizontal displacement in down-slope and cross-slope direction recorded by IPI2 (including four tilt sensors at ground surface, 1, 3 and 5 m below ground
30 in situ total horizonal stress (kPa) / Pore-water pressure (kPa) 25 20 15
crest
-10 0
Down-slope displacement (mm) 0 10 20 30 40 50

(a)
60 0
-4
Cross-slope displacement (mm) -3 -2 -1 0 1 2 3

(b)
1
Colluvium
1
Colluvium
2
Depth (m)
2
CDT
3 4 5 6 7
3 4
21 March 08 (0mm/hr) 19 April 08 (62mm/hr) 19 May 08 (0mm/hr) 7 June 08 (133.5mm/hr) 11 June 08 (0mm/hr) 12 Jan 09 (0mm/hr) 12 Jan 09 (Manual)
CDT
5 6 7
21 March 08 (0mm/hr) 19 April 08 (62mm/hr) 19 May 08 (0mm/hr) 7 June 08 (133.5mm/hr) 11 June 08 (0mm/hr)
Figure 58. Horizontal ground displacement in (a) downslope and (b) cross-slope direction on some selected key days (modified from Leung et al. 2010).
140
EPC 2A
toe
120
15kPa
100
EPC 2B
EPC2B
80 60
10 5
5kPa
0 -5
PWP measured by JFT_1.5m
40 20 0
-10
5 June
6 June
7 June Date
8 June
9 June
Figure 57. Variation of in situ total horizontal stress with time in 2008 at the landslide body (EPC 2 A/B) (modified from Leung et al. 2010).
EPC2A
level) on some selected key days. Positive displacement in down-slope and cross-slope direction denotes that the ground moves towards the directions of North and East, respectively. The number shown in each bracket in each figure denotes the peak rainfall intensity during the day. In response to the heaviest rainfall on 7 June, significant down-slope ground movements were observed in the landslide body. A nearly irrecoverable lateral displacement of 40 mm was recorded, where the displacement profile is fairly constant to a depth of 5 m (see Fig. 58(a)). Moreover, comparison between the manual inclinometer survey (dotted line) and the measurements made by the IPI2 (open triangle) on 12 Jan 2009 provides evidence that well-defined rupture surfaces situated at 5 m below ground or deeper have possibly developed due to the relative large displacement at 5 and 7 m. These newly-developed rupture surfaces are likely to have been caused by the significant rise of groundwater table (5 to 6 m from its initial position (see responses of CP 2B and CP 3B in Fig. 56(c)), which may substantially destroy the matrix suction at deep region. In other words, except for the shallow, translational-type of failure deduced from the ground investigations (see Fig. 53(a)), the landslide mass may undergo another distinct type of failure after the heaviest rainfall on 7 June. Considering also the deformation characteristic of the ground recorded by IPI1 (Leung et al. 2010), a rotationalslide type of failure may have been triggered. The combined failure mode may be described as multiple, translational- and rotational-slide type (Varnes 1978). By comparing measured data on 7 June 2008, 11 June 2008 and 12 January 2009 shown in Fig. 58(a), it is interesting to note that there are some recoverable displacements recorded after the heavy rainstorm of 133.5 mm/hr on 7 June 2008. Similarly, some partial recoverable displacements are also reported by Ng et al. (2003) from their
86
field monitoring of an 11 m high unsaturated cut slope in expansive clay with medium plasticity in Zaoyang, close to the Middle-route in Hubei, China. In contrast to the down-slope direction, the displacement in cross-slope direction appears to be limited and less sensitive to the heavy rainstorms (see Fig. 58(b)). The ground deforms in the easterly direction, where the maximum displacement of about 3 mm takes place near the colluvium/ CDT interface. The eastern ground movement may possibly be a result of the significant cross-slope groundwater flow on top of the dipped, shallow, decomposed rock head regime at about 75 mPD (see the arrows in Fig. 43(b)). 5 CONCLUDING REMARKS
5.1 General comments on reviewed papers A total of fifty five papers under the Topic Geotechnical problems and case histories have been reviewed. Eight and nine papers under the themes Geotechnical problems and Case histories respectively, are selected for discussion in this SOA report. 5.1.1 Geotechnical problems It is evident that there has been increasing interest in the study of soil-atmosphere interaction through theoretical considerations, laboratory testing and field monitoring (sixteen out of the fifty five papers). Some of them present invaluable and comprehensive sets of meteorological data including rainfall, temperature, solar radiation, relative humidity and wind velocity under various seasonal climatic variations. They are extremely useful and relevant to analyse the mechanisms involved in soil-atmosphere interface and to calibrate existing theoretical and numerical models. It is clear that more complete sets of meteorological data are needed to improve our understanding of soil-atmosphere interaction and the accuracy in predicting the actual evaporation from theoretical and empirical models. Different geotechnical problems are addressed from the perspective of unsaturated soil mechanics in many submitted papers. One of the studies measures the changes of PWPs of an initially unsaturated specimen using a high capacity tensiometer when it is loaded to different confining pressures (0 to 1000 kPa) under undrained conditions to both air and water phases. The test results help to quantify the amount of excess PWP generated when an embankment is compacted to a certain height. It would be ideal if any excess poreair pressure can be measured so that the matrix suction, which is one of the constitutive variables
governing unsaturated soil behaviour, can be quantified. Three out of the fifty five papers describe small-scale model tests under one gravity (i.e., 1 g) investigating the influence of matrix suction (up to 12 kPa) on bearing capacity of model strip and square footings with various sizes in sand and shaft resistances of a small model pile (0.65 m in diameter and 1.5 m in length) at suctions ranging from 0.9 to 16 kPa in sand. The predicted bearing capacity and shaft resistance, using some existing and newly proposed semi-empirical equations, seem to show fairly good agreements with the measured results. However, the use of these semi-empirical equations should be treated with caution if they are extrapolated to predict prototypes. This is because they are calibrated by the tests carried out at relatively low stress levels at which soil dilates un-conservatively as oppose to contractive behaviour at high in situ stresses. Surprisingly there is only one paper that investigates thermal effects on volume change of an expansive bentonite/silt mixture, if there is no similar study grouped in other Topics. Some effort seems to be necessary to research the thermalhydro-mechanical behaviour of an unsaturated soil. It would be useful for some engineering applications like pavement design, where superficial subgrade materials are sensitive to climate variations, i.e., continuous drying-wetting cycles and temperature changes. 5.1.2 Case histories A number of excellent case histories (thirteen out of fifty five papers) are presented in this conference. Most slope monitoring programmes cover long-term measurements (up to five years). Matrix suction (the first constitutive variable) and water content at various depths under the effects of climate are often measured. In order to have a complete assessment of the responses of unsaturated soil slopes and provide data for calibrating advanced constitutive soil models and numerical modelling procedures, measurements of the second constitutive variable, net stress, and ground deformations are essential. In addition to slopes, the performance of other earth structures like embankments, isolation barriers, capillary barrier systems and landfill covers under seasonal climatic variations are studied and reported in this conference. Some important parameters for geoenvironmental designs such as lechate discharge rate and landfill gas emission rate are of increasing concern. They are two essential parameters for the design of embankment and landfill cover, for municipal solid waste treatment, in particular.
87
5.2
SOA review on geotechnical problems: Deformation and failure mechanisms of man-made (loose and dense fill) slopes
Static liquefaction/fluidization in a loose LB sand fill slope due to a rising ground water table was successfully created in the geotechnical centrifuge facility at HKUST. The occurrence of liquefaction in sand is captured by in-flight video cameras and proven by the measured significant and sudden build-up of excessive positive pore water pressures at various locations of the slope. It is found that strain-softening of a material is a necessary but not sufficient condition to cause liquefaction. A trigger such as seepage force or additional loading is needed. On the contrary, only no-liquefied slide, slide-toflow and global instability mechanisms with clear failure planes are observed in thin CDG fill slopes in centrifuge, depending on boundary triggering conditions only but not compaction density. The observed centrifuge test results may be explained by triaxial element tests on initially unsaturated CDG specimens sheared at a constant net stress ratio ( p/q) as p and q were kept constant while reducing soil suction during a stress path test. If a loose CDG fill slope is thick, due to being either subjected to rising ground water table and/ or heavy rainfall in the centrifuge, only excessive settlements are measured. No liquefaction is observed. The significant difference between the observed static liquefaction in loose LB sand fill slope and non-liquefied but excessive settlement in loose CDG fill slopes may be attributed to the significant difference in the liquefaction potential (i.e., qmax qmin) observed during consolidated undrained triaxial element tests on the two soil types. 5.3 SOA review on case histories: The performance of the natural saprolitic hillslope at Tung Chung on Lantau Island in Hong Kong
measured in the laboratory by about three orders of magnitude, which is likely to be due to the presence of preferential flows through cracks, fissures and relict joints in the highly heterogeneous colluvium in situ. Based on the monitoring results recorded from the heavily instrumented natural hillslope, sitespecific infiltration and deformation characteristic of the hillslope in response to heavy precipitations are investigated. When the hillslope was subjected to rainstorms with a peak rainfall intensity of 133.5 mm/hr (estimated return period of 245 years), a transient perched groundwater table is believed to have developed at the bouldery colluvial deposits on the top 3 m, where PWP up to 25 kPa is typically measured. The main groundwater table rose by about 5 to 6 m, approaching the colluviums/CDT interface at about 3 m below ground level. Based on the measured PWP at 1.5 m depth, it can be deduced that the measured increase in total horizontal stress in down-slope direction at 2 m depth are not only originated from the increase of positive PWP, but the increase may also be attributed to significant lateral soil displacement of up to 40 mm at the ground surface in the down-slope direction. Rupture surfaces may have been formed at depths below 5 m resulting in a multiple translational- and rotational-slide type of failure. This case history highlights the importance of monitoring both two constitutive variables (matrix suction (or PWP) and net stress) and their influences on ground deformations in order to better reveal and understand the responses of the hillslope. ACKNOWLEDGEMENTS The authors would like to acknowledge the support by research grant HKUST9/CRF/09 provided by the Research Grants Council of the Hong Kong Special Administrative Region (HKSAR). REFERENCES
ASTM. 2000, Standard practice for classification of soils for engineering purposes (Unified Soil Classification System). ASTM standard D2487. In 2000 Annual Book of ASTM Standards. ASTM International, West Conshohocken, USA. ASTM. 2003. Standard Test method for Measurement of Soil Potential (Suction) Using Filter Paper. ASTM standard D5298. In 2000 Annual Book of ASTM Standards. ASTM International West Conshohocken, USA. BSI. 1990. BS1377: Methods of tests for soils for civil engineering purposes. British Standards Institution, London.
In order to assist in understanding of field testing and monitoring data, the influences of the two constitutive variables (matrix suction and net stress) on soil-water characteristic and unsaturated permeability of the CDT underlying the natural saprolitic hillslope at Tung Chung are investigated in the laboratory. The field observations from the IP test reveal that there is a clear trend of decreasing sizes of hysteretic loops of SDSWCCs with depth. The field-measured permeability ranges from 4 107 to 3 104 m/s in the colluvium, whereas the corresponding measured values vary from 4 107 to 1 104 m/s in the CDT. The field-measured permeability is higher than those
88
Cai, Z.Y. 2001. A comprehensive study of state-dependent dilatancy and its application in shear band formation analysis. PhD thesis, HKUST. Castro, G. 1969. Liquefaction of sands. Harvard Soil Mechanics Series 87, Harvard University, Massachusetts. Chu, J. and Leong, W.K. 2002. Effect of fines on instability behaviour of loose sand. Gotechnique 52(10): 751755. Corley, H.P., Martin, R.L. and Macklin, R.L. (1999). Characterisation of saprolite heterogeneities using innovative techniques. Geotechnical Special Publication No. 92, ASCE, pp. 6476. Delta-T Devices Ltd., 1999. ThetaProbe soil moisture sensor (Type ML2x) User Manual ML2x-UM-1.21, Cambridge, UK. Fredlund, D.G., Xing, A. and Haung, S.Y. 1994. Predicting the permeability function for unsaturated soils using the soil-water characteristics curve. Can. Geotech. J., 31: 533546. Gens, A. 2009. Some issues in constitutive modelling of unsaturated soils. Proc. 4th Asia-Pacific Conf. on Unsaturated Soils, 2325 Nov. Newcastle, Australia. pp. 613626 Geotechnical Control Office (GCO). 1982. Mid-level Studies: Report on geology, hydrology and soil properties. Geotechnical Control Office, Hong Kong, 2 Volumes, 266 p. Geotechnical Engineering Office (GEO). 1988. Guide to rock and soil descriptions, Geoguide 3, Geotechnical Engineering Office, Hong Kong. Geotechnical Engineering Office (GEO). 1996. Methods of tests for soils in Hong Kong for civil engineering purposes. GEO report, 36, Geotechnical Engineering Office, Hong Kong. Hilf, J.W. 1956. An investigation of pore water pressure in compacted cohesive soils. Technical Memo 654. Denver: Bureau of Reclamation. Hillel, D. 1982. Introduction to soil physics. Academic Press, New York, N.Y. Hong Kong Institute of Engineers (HKIE) 2003. Soil nails in loose fill slopes: A preliminary study. Geotechnical Division. Final Report. Japanese Society of Soil Mechanics and Foundation Engineering (JSSMFE), 1982. Soil testing methods, Japanese Society of Soil Mechanics and Foundation Engineering, Tokyo. Kawai, K., Nagareta, H., Hagiwara, M. and Iizuka, A. 2003. Suction changes of compacted soils during static compaction test, Proc. of 2nd Asian Conf. on Unsaturated Soils: 429434. Knill, J.L., Lumb, P., Mackey, S., Mello, V.F.B., Morgenstern, N.R., and Richards, B.G. 1976. Report of the independent review panel on fill slopes, Government of Hong Kong. Kramer, S.L. 1996. Geotechnical earthquake engineering, Prentice-Hall, Inc. New Jersey, USA. Lade, P.V. 1992. Static instability and liquefaction of loose fine sandy slopes. J. Geotech. Eng. ASCE, 118(1): 5171. Lee, Y.S., Cheuk, C.Y. and Bolton, M.D. 2008. Instability caused by a seepage impediment in layered fill slopes. Canadian Geotechnical Journal. 45: 14011425. Leung, A.K. and Ng, C.W.W. (2009). Experimental study of the soil-water characteristics of a decomposed soil
using the relative humidity (RH) control technique. Proc. 4th Asia-Pacific Conf. on Unsaturated Soils, 2325 Nov. Newcastle, Australia. pp. 199204. Leung, A.K. and Ng, C.W.W. (2010). Back-analysis of infiltration characteristic of a saprolitic hillslope by considering permeability heterogeneity. Proc. 5th Int. Conf. on Unsaturated Soils, 69 Sep. Barcelona, Spain. Leung, A.K., Sun, H.W., Millis, S., Pappin, J., Ng, C.W.W. and Wong, H.N. 2010. Field monitoring of an unsaturated saprolitic hillslope. Submitted to Canadian Geotechnical Journal. Maciel, F.J. 2003. Estudo da gerao, percolao e emisso de gases no aterro de resduos slidos da Muribeca/PE. Master Dissertation. Federal University of Pernambuco. Recife-PE. 173 p. Malone, A.W. 1998. Risk management and slope safety in Hong Kong. Proceedings of the Annual Seminar on Slope Engineering in Hong Kong. Geotechnical Division of the Hong Kong Institution of Engineers: 329. Ng, C.W.W. 2007. Keynote paper: Liquefied flow and non-liquefied slide of loose fill slopes. Proc. 13th Asian Regional Conference on Soil Mechanics and Geotechnical Engineering, Kolkata. Vol. 2. Allied Publishers Private Ltd. Ng, C.W.W. 2008. Invited special lecture: Deformation and failure mechanisms of loose and dense fill slopes with and without soil nails. Proc. of 10th Int. Sym. On Landslides and Engineered Slopes. Xian, China. Vol. 1, 159177. Ng, C.W.W. and Chen, R. 2005. Keynote lecture: Advanced suction control techniques for testing unsaturated soils (In Chinese), Proc. of 2nd National Conference on Unsaturated Soils, Hangzhou, China, 2223 April, pp. 144167. Ng, C.W.W. and Chiu, A.C.F. 2001. Behaviour of a loosely compacted unsaturated volcanic soil. J. Geotech. And Geoenviron. Engg., ASCE, 127(12), 10271036. Ng, C.W.W. and Chiu, A.C.F. 2003. Laboratory study of loose saturated and unsaturated decomposed granitic soil. J. Geotech. and Geoenviron. Eng., ASCE, 129, No. 6., pp 550559. Ng, C.W.W and Leung, A.K. 2010. Measurements of drying and wetting permeability function using a new stress-controllable soil column. Submitted to GeotechniqueSymposium In Print. Ng, C.W.W. and Pang, Y.W. 2000. Experimental investigations of the soil-water characteristics of a volcanic soil. Canadian Geotechnical Journal, 37: 12521264. Ng, C.W.W., Leung, E.H.Y. and Lau, C.K. 2004. Inherent anisotropic stiffness of weathered geomaterial and its influence on ground deformations around deep excavations. Canadian Geotechnical Journal 41: 1224. Ng, C.W.W., Zhang, M. and Shi, X.G. 2002b. Keynote (In Chinese). An investigation into the use of soil nails in loose fill slopes, Proc. of the 1st Chinese Symposium on Geoenvironment and Geosynthetics, Hangzhou, Zhejiang, China, 1719 Nov., pp 6180. Ng, C.W.W, Zhang, L.M. and Wang, Y.H. 2006. Proceedings of 6th Int. Conf. on Physical Modelling in Geotechnics. Volumes 1 and 2. Publisher: Taylor & Francis. ISBN: 978-0-415-41587-3 and 978-0-415-41588-0. Ng, C.W.W., Fung, W.T., Cheuk, C.Y. and Zhang, L.M. 2004. Influence of stress ratio and stress path on
89
behaviour of loose decomposed granite. J. Geotech. and Geoenviron. Eng., ASCE, 130, No.1., pp. 3644. Ng, C.W.W., Pun, W.K., Kwok, S.S.K., Cheuk, C.Y. and Lee, D.M. 2007. Centrifuge modelling in engineering practice in Hong Kong. Proc. of the Geotechnical Divisions Annual Seminar, Hong Kong Institution of Engineers, 5568. Ng, C.W.W., Zhan, L.T., Bao, C.G., Fredlund, D.G. and Gong, B.W. 2003. Performance of an Unsaturated Expansive Soil Slope Subjected to Artificial Rainfall Infiltration, Geotechnique, 53(2): 143157. Ng, C.W.W., Zhang, M., Pun, W.K., Shiu, Y.K. and Chang, G.W.K. 2009. Investigation of static liquefaction mechanisms in loose sand fill slopes. Submitted to CGJ. Ng, C.W.W., H.N. Wong, Y.M. Tse, J. Pappin, Sun. H.W., S. Millis and Leung, A.K. 2010. Field study of stress-dependent soil-water characteristic curves and Hydraulic conductivity in a saprolitic slope. Geotechnique (provisionally accepted). Ng, C.W.W., Van Laak, P.A., Zhang, L.M., Tang, W.H., Zong, G.H., Wang, Z.L., Xu, G.M. and Liu, S.H. 2002a. Development of a four-axis robotic manipulator for centrifuge modeling at HKUST. Proc. Int. Conf. on Physical Modelling in Geotechnics, St. Johns, Canada: 7176. Sasitharan, S., Robertson, P.K., Sego, D.C. and Morgensterm, N.R. 1993. Collapse behavior of sand. Canadian Geotechnical Journal, 30, No.3., pp. 569577. SEEP/W (Version 4) for finite element seepage analysis. 1998. Geo-slope International, Canada. Sladen, J.A., DHollander, R.D. & Krahn, J. 1985. The liquefaction of sands, a collapse surface approach. Canadian Geotechnical Journal 22: 564578. Soilmoisture Equipment Corp. 2005. Operation instructions: the model 2100F soilmoisture probe. Soilmoisture Equipment Corp., Goleta, Calif. Take, W.A., Bolton, M.D., Wong, P.C.P. and Yeung, F.J. 2004. Evaluation of landslide triggering mechanisms in model fill slopes. Landslides, Japan, Vol. 1: 173184. Tang, W.H. and Lee, C.F. 2003. Potential use of soil nails in loose fill slope: an overview, Proc. of the International Conference on Slope Engineering, Hong Kong, China, pp. 974997. van Genuchten, M. Th., 1980, A Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils, Soil Science Society of American Journal, 44, pp. 892898. Varnes, D.J. 1978. Slope movement types and processes. In: R.J. Schuster and R.J. Krizek (eds), Landslides, Analysis and Control (Special Report 176, pp. 1133). Transportation Research Board, National Academy of Sciences, Washington, DC. Watson, K.K. 1966. An instantaneous profile method for determining the hydraulic conductivity of unsaturated porous materials. Water Resources Research, 2, 709715. Willson, G.W., Fredlund, D.G. and Barbour, S.L. 1997. The effect of soil suction on evaporative fluxes from soil surfaces. Canadian Geotechnical Journal, 34: 145155. Zhang, M. 2006. Centrifuge modelling of potentially liquefiable loose fill slopes with and without soil nails. PhD Thesis, the Hong Kong University of Science and Technology.
Zhang, M. and Ng, C.W.W. 2003. Interim Factual Testing Report I-SG30 & SR30. Hong Kong University of Science & Technology. Zhang, M., Ng, C.W.W., Take, W.A., Pun, W.K., Shiu, Y.K. and Chang, G.W.K. 2006. The role and mechanism of soil nails in liquefied loose sand fill slopes. Proc. of 6th Int. Conf. Physical Modelling in Geotechnics, Hong Kong. Vol. 1: 391396.
LIST OF THE SEVENTEEN PAPERS DISCUSSED IN THIS REPORT

Arnedo, D., Alonso, E.E., Olivella, S. and Romero, E. Gas migration in sand/bentonite mixtures through preferential paths. Birle, E., Boso, M. and Heyer, D. Investigation of the water flow in an embankment built of organix soil: Experimental results and large scale field test. Blight, G.E. Soil heat and evaporation from soil. Caruso, M. and Jommi, C. A prototype soil column to calibtate numerical models accounting for soilatmosphere interaction. Le, T.M.H., Gallipoli, D., Wheeler, S. and Sanchez, M. Influences of spatially varying porosity on unsaturated flow through earth structures. Leung, A.K. and Ng, C.W.W. Back-analysis of infiltration characteristic of a saprolitic hillslope by considering permeability heterogeneity. Lopes, R.L., Alves, M.C.M. and Juc, J.F.T. Water infiltration and methane emission through three different cover layers of an experimental municipal waste landfill at Muribeca, Recife, Pernambuco, Brazil. Macial, F.J., Lopes, R.L. and Juc, J.F.T. Evaluation of landfill gas emission in experimental cover layers in Brazil. Marinho, F.A.M. and Massad, F. Behaviour of wet fill using residual soil. Montrasio, L., Valentio, R. and Quintavalla, C. Use of satellite observations for estimating the degree of saturation of superficial soils. Nacinovic, M.G.G., Mahler, C.F., Izzo, R.L.S. and Brandt, P.F.S. The use of filter paper method for field suction investigation. Nomura, S., Kawai, K., Kanazawa, S., Iizuka, A., Ohno, S., Thirapong, P. and Tachibana, S. Study of stratification resistance to salt damage. Rahardjo, H., Hua, C.J., Leong, E.C. and Santoso, V.A. Performance of an instrumented slope under a capillary barrier system. Scotto di Santolo, A. and Evanglista, A. Matric suction in unsaturated pyroclastic slopes. Shibata, M., Kawai, K., Kanazawa, S., Iizuka, A., Tachibana, S., Ohno, S. and Honda, M. Simulation of static compaction with unsaturated soil/water coupled F.E. analysis. Ta, A.N., Tang, A.M., Cui, Y.J. and Thiriat, J. An environmental chamber for studying the soil-atmosphere interaction. Vanapalli, S.K., Eigenbrod, K.D., Taylan, Z.N., Catana, C., Oh, W.T. and Garven, E. A technique for estimating the shaft resistance of test piles in unsaturated soils.
90
Constitutive modelling of unsaturated soils: Discussion of fundamental principles

D. Sheng
Centre for Geotechnical and Materials Modelling, the University of Newcastle, NSW, Australia
ABSTRACT: An unsaturated soil is not a special type of soil, rather a state of the soil. All soils can be partially saturated with water. Therefore, constitutive models for soils should ideally represent the soil behaviour over entire ranges of possible pore pressure and stress values and allow arbitrary stress and hydraulic paths within these ranges. This paper attempts to present an overview of constitutive modelling for unsaturated soils. In particular, it focuses on the fundamental principles that govern the volume change, shear strength, yield surfaces, water retention and hydro-mechanical coupling. Alternative forms of these principles are critically examined in terms of their advantages and disadvantages. The paper also presents a short summary of implementing constitutive models into the finite element method. 1 INTRODUCTION can lead to severe damages to foundations and structures. Shear strength of soils can also change dramatically as the degree of saturation changes, and a related engineering problem is slope failures caused by rainfall. Unsaturated soils also have distinctive hydraulic behaviour which has profound implications in designing cover and containment systems for various industrial and municipal wastes. These fundamental issues are indeed the main concerns of unsaturated soil mechanics and its engineering applications (Fredlund & Rahardjo 1993; Houston 2002). Constitutive modelling of unsaturated soils generally involves the generalisation of constitutive models for saturated states to unsaturated states, by incorporating the fundamental issues mentioned above. The research was pioneered by Alonso et al. (1990) and it has since attracted extensive interest. A large number of constitutive models can be found in the literature. There are several state-ofthe-art reports and review papers over the last 15 years or so, e.g., Gens (1996), Wheeler & Karube (1996), Kohgo (2003), Gens et al. (2006), Wheeler (2006), Gens (2008), Sheng & Fredlund (2008), Sheng et al. (2008c), Gens (2009), Cui & Sun (2009) and Gens (2010). These papers may serve as good references for studying the topic. They usually contain (1) a thorough discussion of stress state or constitutive variables used to establish various models, (2) an in-depth analysis of specific constitutive models and their advantages and disadvantages, and (3) latest developments in the area of unsaturated soil modelling. The papers by Gens (2009, 2010) also provide interesting discussions on the physical significance of different suction components and their roles in constitutive modelling.
Soils that are partially saturated with water are often referred to as unsaturated soils. It should be stressed that all soils can be partially saturated with water. An unsaturated soil is just a state of the soil, not a special type of soil, as pointed out by Gens et al. (2006). Some soils may exhibit distinctive volume, strength and hydraulic properties when become unsaturated. For these soils, a change in the degree of saturation can cause significant changes in the volume, shear strength and hydraulic properties. Nevertheless, the distinctive volume, strength and hydraulic behaviour for unsaturated states only represent a form of material non-linearity and should therefore be handled within a continuous and coherent framework. In other words, a constitutive model for a soil should represent the soil behaviour over entire ranges of possible pore pressure and stress values and should allow arbitrary stress and hydraulic paths within these ranges. Soil mechanics principles are more established for soils at saturated states. Generalisation of these principles to unsaturated soils requires careful consideration of the following fundamental issues: (1) volume change behaviour associated with suction or saturation changes, (2) shear strength behaviour associated with suction or saturation changes, and (3) hydraulic behaviour associated with suction or saturation changes. Soils can experience significant volume changes upon changes of the degree of saturation or suction. Some soils expand upon wetting, some collapse and some do both depending on the stress level. The large volume changes associated with saturation change
91
Instead of a comprehensive review of constitutive models for unsaturated soils, this paper devotes its main attention to a number of specific issues: (1) volume change behaviour, (2) variation of yield stress and shear strength with suction, (3) water retention behaviour and hydro-mechanical coupling. These issues represent the most fundamental components of constitutive models. Alternative methods for tackling these fundamental issues are scrutinised, particularly against the principle that partial saturation is a state of soil. The issue of stress state variables, a seemingly unavoidable topic in unsaturated soil mechanics, is not specifically discussed in this paper. In other words, constitutive models are not judged based on the stress state variables they use, rather on their qualitative predictions of observed soil behaviour. In addition, a number of advanced topics are excluded: 1. Thermodynamics of unsaturated soils. This topic often leads to interesting discussion of thermodynamically consistent stress-state variables and models. Readers may refer to Houlsby (1997), Hutter et al. (1999), Gray & Schrefler (2001), Li (2007), Samat et al. (2008), Coussy et al. (2010) and Zhao et al. (2010). 2. Micromechanical modelling of unsaturated soils. This approach can often lead to insights into soil behaviour and sometimes also validation of macroscopic (continuum) constitutive equations. Some good examples of this approach include the work by Gili & Alonso (2002), Jiang et al. (2004), Katti et al. (2007) and Scholts et al. (2009). A closely related approach is the micro-macro double-structure models by Gens & Alonso (1992), Alonso et al. (1999), Snchez et al. (2006) and Cardoso & Alonso (2009). 3. Thermo-hydro-chemo-mechanical modelling of unsaturated soils. In this approach, additional environmental variables such as temperature and chemical concentrations are introduced to study soil behaviour. Some recent work in this area refers to Loret et al. (2002), Guimares et al. (2007), Cleall et al. (2007), Kimoto et al. (2007) and Gens (2006, 2008, 2010). 4. Miscellaneous topics such as non-isothermal behaviour, anisotropy, rate-dependent behaviour and liquefaction of unsaturated soils. These topics are usually related to specific soils or specific problems. Some representative work on these topics are Cui & Delage (1996), Stropeit et al. (2008) and DOnza et al. (2010) for anisotropy; Modaressi & Modaressi (1995) and Cui et al. (2000) for non-isothermal behaviour; Cardoso & Alonso (2009) for degradation modelling; Arson & Gatmiri (2008) and Yang et al. (2008) for damage modelling; Oldecop & Alonso (2007) and Pereira & de Gennaro (2010)
for rate-dependent behaviour; Unno et al. (2008) and Bian & Shahrour (2009) for liquefaction. The paper is organised as follows. It first presents a brief discussion of two basic concepts used in constitutive modelling, i.e. the net stress and suction. The fundamental issues on volume change, yield stress, shear strength, water retention and hydromechanical coupling are then discussed. The paper finally outlines the challenges and possible solution strategies for implementing unsaturated soil models into the finite element method. 2 NET STRESS AND SUCTION
Net stress is commonly used in interpreting unsaturated soil behaviour and in constitutive modelling. It is defined as
ij ij ij ua
(1)
where ij is the net stress tensor, ij the total stress tensor, ij the Kronecker delta, and ua the pore air pressure. The net stress is often used to analyse laboratory data, particularly those based on the axis-translation technique where the air pressure is not zero. It is sometimes perceived to recover the effective stress when soils become saturated. Such a perception should however be avoided. Under natural ground conditions where the air pressure is atmospheric, the net stress is equivalent to the total stress. Indeed, we never use the net stress concept to describe the behaviour of dry sand. In other words, the atmospheric pore air pressure should be considered zero. Net stress is different from total stress only when the air pressure is not atmospheric. The concept of net stress can be useful in interpreting experimental data based on axis-translation technique, if the technique is indeed valid for applying suction (see discussions on its validity in e.g. Ng et al. 2007; Baker & Frydman 2009). In this case, the air pressure is used as a reference value for stress measures and the net stress is the total stress in excess of the pore air pressure. The soil suction in the literature of unsaturated soil modelling usually refers to the matric suction and is usually expressed as: s ua uw (2)
where s is called the matric suction in soil physics terminology and also called the matrix suction (Baker & Frydman 2009), and uw the pore water pressure. The matric suction is used interchangeably with the matric potential in soil-water potential. The latter is a measurement of energy and consists of two parts: capillary and adsorptive potentials.
92
When pore water exists as capillary water at relatively high degrees of saturation, the capillary potential is dominant in the matric potential and the definition by (2) is then considered to be valid. However, when pore water exists as adsorbed water films at low degrees of saturation, the adsorptive potential (a) becomes dominant in the matric potential. Consequently, questions have been raised regarding whether equation (2) is still valid for the matric potential (Baker & Frydman 2009). In the case when water exists as adsorbed films to solid particles, the true water pressure is not well defined. It is not unique at one material point and is dependent on the proximity to the solid particle surface (Marshall et al. 1996). However an apparent water pressure can be introduced: uw ua a, i.e. the apparent water pressure represents the negative adsorptive potential measured in excess of air pressure. Such an apparent water pressure is then unique at one material point. With such a definition of uw, the matric potential can be expressed by equation (2) over the full range of saturation. Nevertheless, the matric suction should be differentiated from capillary phenomena. Its two components may not easily be separable in a soil with double-porosity. As pointed out in Gens (2010), it is more appropriate to think of matric suction as a variable that expresses quantitatively the degree of attachment of water to solid particles that results from the general solid/water/interface interaction. In constitutive modelling, the matric suction is often treated as an additional stress state variable in a stress space for establishing constitutive laws. This approach was pioneered in the Barcelona Basic Model (BBM) by Alonso et al. (1990) and is then followed in most existing models, with a few exceptions where the suction is treated as an internal variable (e.g. Bolzon et al. 1996; Loret & Khalili 2002). Again, these approaches are not differentiated in this paper, even though there are arguments against treating suction as an internal variable (e.g. Laloui & Nuth 2009). Since suction can vary independently of stress, it is treated as an independent axis in the stress space. In addition, the matric suction is considered to coincide with negative pore water pressure for fully saturated states and thus the suction axis goes from negative infinity to positive infinity in the stress space. 3 VOLUME CHANGE BEHAVIOUR
and shear strengthsuction relationships (Sheng et al. 2008c). It is indeed the only absolutely necessary component that is needed to add in order to extend a saturated soil model to unsaturated states. The model that defines the volume change caused by stress and suction changes should again be applicable to the entire range of possible pore pressure or suction values. The discussions below are limited to isotropic stress states. The volume change associated with changes of deviator stress has to be considered in a three-dimensional constitutive framework, which depends on the specific model used for saturated soils, and is not discussed in this paper. For saturated soils, a common starting point is the linear relationship between the specific volume (v) and the logarithmic effective mean stress (ln p) for normally consolidated clays: v N ln p N ln(p uw) (3) where is the slope of the v ln p line, and N is the intercept on the v axis when ln p 0. Equation (3) is only valid for positive increments of the effective stress. For unloading and reloading, the volume change depends on the specific plasticity framework adopted in the constitutive model. For example, hypoplasticity and bounding surface plasticity adopts different volume change mechanisms than classical elastoplasticity. However, for normally consolidated soils subject to positive stress increments, equation (3) is usually used independently of the theoretical framework. It should be noted that equation (3) represents a straight line in the v ln p space only if the pore water pressure is zero. If the pore water pressure were kept at a negative value (suction), equation (3) would predict a smooth curve in the v ln p space, as shown by Figure 1. The air entry suction for the soil in Figure 1 is assumed to be larger than 100 kPa, so that the soil remains saturated. Indeed, these compression lines look very much like those for overconsolidated soils. However, the curvature of the normal compression lines is purely due to the nature of the logarithmic function and the translation from the effective stress space (v ln p) to the total stress space (v ln p), not due to overconsolidation. Equation (3) can also be written in an incremental form as follows: dv d( uw ) dp p uw p uw (4)
The volume change behaviour is one of the most fundamental properties of soils. For unsaturated soils, the large volume changes associated with suction change can cause severe damages to foundations and structures. The volume change equation also underpins the yield stresssuction
It is clear that a negative increment in pore water pressure has exactly the same effect on the volume of a saturated soil as an equal positive increment in mean stress.
93
3.0
uw =0
2.5
uw =-10 kPa uw =-100 kPa
2.0
1.5 1 10 100 1000
p, kPa
Figure 1. Normal compression lines for saturated clay under constant pore water pressures ( 0.2, N 3).
In the literature, equation (3) is extended to unsaturated states in one of the three approaches: Approach A: Separate stress and suction (the net stress and suction) approach. Approach B: Combined stress-suction approach (the effective stress approach). Approach C: SFG approach (middle ground between Approach A and B). These approaches are discussed below in detail. 3.1 Separate stress and suction approach
In Approach A, the volume change due to stress change is separated from that due to suction change. A typical example of the volume change equations in this approach is: s uat v N vp ln p vs ln uat (5)
but is also supported by experimental data. Toll (1990) and Toll & Ong (2003) showed that the two compressibilities vp and vs can be totally different (Figure 6). It is usually true that the suction shrinkability (vs) decreases with decreasing degree of saturation. On the other hand, the stress compressibility (vp) can increase with decreasing degree of saturation, particularly for compacted soils where highly compressible macropores (inter-aggregate pores) are formed (Romero et al. 1999; Gallipoli et al. 2003a). However, there are also a few disadvantages about equation (5). First, equation (5) does not recover equation (3) when the soil becomes saturated. It represents a linear v ln p relationship for constant suctions, unless vp is assumed to be a function of stress (as in Georgiadis et al. 2005). This is not consistent with the saturated soil model (Figure 1). As such, the volume change becomes undefined at the transition suction between saturated and unsaturated states. Second, the volume change caused by suction changes is independent of stress. This contradicts with experimental observation shown in Figure 3. In addition, the atmospheric pressure (uat) in (5) makes the suction change insignificant when s uat, which is not consistent with the drying of a slurry soil. The first point, i.e. the discontinuity at the transition between saturated and unsaturated states, was also one of the reasons that some researchers turned to the effective stress approach (e.g. Sheng et al. 2003a). A simple numerical example will illustrate this problem. Let a soil be compressed at the transition suction (ssa) from mean stress 1 kPa to 100 kPa. Let the air pressure remains atmospheric. In the saturated zone, the volume changes according to equation (3): v s- vp ln
sa
where N is the specific volume when ln p 0 and s 0, vp is the slope of an assumed v ln p line or the compressibility due to stress change, vs the slope of an assumed v ln s line or the shrinkability due to suction change, and uat the atmospheric pressure and is added to avoid the singularity when s 0. Again, equation (5) is only used for increasing mean stress or increasing suction. Indeed, vs is usually replaced by the elastic compression index vs in most applications, unless the suction increases above the suction-increase yield surface (see section below on yield stress). Equation (5) has been used in many models like Alonso et al (1990), Wheeler & Sivakumar (1995), Cui & Delage (1996), Chiu & Ng (2003), Georgiadis et al. (2005) and Thu et al. (2007a). The main advantage of equation (5) is that the compressibility due to stress and suction changes can be dealt with separately. This does not only provide extra flexibility for modelling soil behaviour,
100 ssa 1 ssa
In the unsaturated zone, the volume changes according to equation (5): v s+ vp ln100
sa
These two volume changes can be quite different, dependent on the value of the transition suction. The transition suction is either the airentry or the air-expulsion value, dependent on the hydraulic path. 3.2 Combined stress-suction approach
In Approach B, the matric suction and the net mean stress are combined into one single variable,
94
i.e., an effective stress, to define their effects on soil volume. A general form of the effective stress is p p f ( s ) (6)
where f is either a function of suction or a function of suction and degree of saturation. Obviously such a definition of effective stress is very general and covers most existing definitions in the literature. With such an effective stress, equation (3) is assumed to be valid for unsaturated states: v N ln p N ( s ) ln p f ( s ) (7)
Figure 2. Variation of compressibilities with degree of saturation (after Toll 1990).
where N is the specific volume when ln p 0. If the effective stress is indeed effective in controlling soil volume, v should remain constant under constant p. As such, parameters N and should be independent of suction. However, this is seldom the case in reality. In the literature, is usually assumed to be function of s, while N is treated either as a constant or a variable. We first discuss the case where N is constant and then explore the possibilities with a varying N. Equation (7) is widely used in the literature and in fact most models based on the effective stress approach adopt it as the volume change equation (Kohgo et al. 1993; Bolzon et al. 1996; Jommi 2000; Loret & Khalili 2002; Sheng et al. 2003a; Sheng et al. 2004; Pereira et al. 2005; Santagiuliana & Schrefler 2006; Sun et al. 2007a, c; Kohler & Hofstetter 2008; Nuth & Laloui 2008; Buscarnera & Nova 2009). Equation (7) generally recovers equation (3) when the soil becomes saturated. This is one of the greatest advantages of using the effective stress. However, there are also some disadvantages with equation (7). The obvious issue is the difficulty in addressing the different compressbilities due to stress and suction changes, as shown in Figure 2. The second issue is related to a constraint on the compressibility . Let a saturated slurry soil be dried from zero suction to an arbitrary suction under constant mean stress of 1 kPa, i.e. the stress path AB in Figure 4. As will be shown in the section on yield stress (Figure 8), this drying path is elastoplastic, not purely elastic. The volume of the soil then changes according to: vB N (s) ln (1 f (s)) (8)
Figure 3. Variation of shrinkability with stress (after Delage and Graham 1996).
p f (s) vC vB ( s ) ln r 1 f (s)
(9)
Along path AC, the volume changes according to: vC N (0 ) ln pr f (0 ) N (0 ) ln pr (10)
Replacing (8) into (9) and then equating (9) with (10) lead to:
(s) ln( pr ) 1 (0 ) ln( pr f ( s ))
(11)
Now compress the soil under constant suction, i.e. stress path BC in Figure 4. The compression line will be curved in the v ln p space, due to the f (s) term. If the suction at point B is above the air entry value, this compression line is expected to intersect with the initial compression line for saturated states. Let the intersection be point C (at net mean stress of pr ). The volume at C is then:
We usually anticipate the effective stress increases with increasing suction, at least at low suction values. Therefore, we have: (s) (0), meaning that the slope of the compression line decreases with increasing suction. Such a constraint on is however not supported by experimental data. In the data by Jennings & Burland (1962) for air-dry soils, the slope of the compression lines is more or less constant. In the data by Sivakumar & Wheeler (2000) for compacted soils, the slope of the compression lines increases with increasing suction.
95
v N A
(a) Drying path constant p
v N A
(b)
Drying path under constant p ? B C
B C
s>0, ds=0 s=0 1 pr s=0

ln p
s>0, ds=0
ln p
Figure 4. Normal compression lines according to the effective stress approachconstraint on (s) with a constant N: (a) net mean stress space, (b) effective mean stress space.
In addition, experimental data on wetting-induced collapse (e.g. Sun et al. 2007d) do not support an ever increasing collapse volume with increasing mean stress. In some constitutive models that use the combined stress-suction approach, parameter N is assumed to vary with suction (Kohgo et al. 1993; Gallipoli et al. 2003a; Kikumoto et al. 2010). If N decreases with increasing suction, the same constraint on , i.e. equation (11) would apply. To avoid this constraint, N has to increase with increasing suction. Gallipoli et al. (2003a) proposed the following volume change equation: v (N ln p) (1 a(1 exp(b)) (12)
where N and are the two parameters of the normal compression line for saturated states, a and b fitting parameters, and a positive variable representing the bonding effects of suction. The bonding variable () is a function of both s and Sr. Equation (12) is hence equivalent to equation (7) with N and being functions of s and Sr. Gallipoli et al. (2003a, 2008) showed that equation (12) is able to predict the volume change at both normal compression and critical states for a variety of compacted soils. One challenge in using equation (12) is that the yield stress is likely to be functions of both suction and degree of saturation. Because the s Sr relationship is usually not unique due to hydraulic hysteresis, the resulting loading-collapse yield surface may not be well defined. However, Gallipoli (personal communication) suggests that it is sufficient to define the bonding variable () in terms of Sr only (e.g. 1-Sr), which would then resolve the non-uniqueness problem and lead to a unique loading-collapse surface in the Sr p space. To avoid the constraint defined by equation (11), another possible augmentation to equation (7)
is to assume that the compressibility is a function of degree of saturation, i.e. (Sr), while keeping N constant. This is similar to the approach by Gallipoli et al. (2003a) with being function of Sr only. Because Sr changes with soil volume even if the suction is kept constant, the slope of the compression line will then change and will most likely increase with increasing mean stress. Therefore, it is possible to have (Sr) decreasing with decreasing Sr, but the slope of the compression line for constant suction increases with increasing stress (Figure 5b). Al-Badran & Schanz (2009) used an approach similar to Figure 5b, but in the net stress space. The same issue of non-uniqueness of the loading-collapse yield surface in the s p space may arise, due to hydraulic hysteresis. Instead, the yield surface may have to be defined in the Sr p space. Clearly further research is required if equation (7) is adopted for the volume change, particularly in terms of consistent explanation of the suctioncaused volume change for reconstituted soils. A worthwhile endeavour in this direction is perhaps to explore the possibilities of using (Sr) in equation (7) while keeping N constant. 3.3 SFG approach Sheng et al. (2008a) proposed a third way to model the volume change for unsaturated soils under isotropic stress states. The new model, referred to as the SFG model, represents a middle ground between Approach A and Approach B and is expressed in an incremental form as follows: dv vp dp ds vs p f (s) p f (s) (13)
Equation (13) is in the same form as equation (4). Similar to Approach A, equation (13) is defined
96
v N(s)
(a)
N
(b) Drying path under constant p ? B
Drying path under constant p?

N(0)
Drying path under constant p ?
(Sr)
s, Sr
s>0, ds=0 s=0

ln p
s>0, ds=0 s=0

ln p
Figure 5. Normal compression lines according to the effective stress approachpossible augmentations: (a) N increases with increasing suction. (b) is a function of degree of saturation (Sr).
in terms of net stress and suction and separates the compressibilities due to the two variables, i.e. vp and vs. Similar to Approach B, it combines the suction with the net mean stress in the denominator, i.e. the term p f ( s ) , and recovers equation (4) for saturated states. The term p f ( s ) represents the interaction between stress and suction and makes the normal compression lines for non-zero suction curved in the v ln p space. However, there is no constraint on parameter vp. As a first approximation, vp can be assumed to be independent of suction, as indicated by the data of Jennings & Burland (1962) for air-dry soils. More realistically it should depend on suction. For example, the data of Sivakumar & Wheeler (2000) shows that vp increases with increasing suction for compacted soils. Parameter vs must equal vp when the soil is fully saturated, because of equation (4). It generally decreases with increasing suction and approaches zero. Sheng et al. (2008a) suggested the following simple function for vs vp , vs s 1 , vp sa s 1 s ssa s ssa (14)
The difference between equation (14) and (15) is minimal, but equation (15) is preferred. Equation (15) can be applied as long as the transition suction is not absolutely zero. Kurucuk et al. (2009) used equation (13), but a different function for vs than (15). Again, both vp and vs can vary with stress path and take different values on a loading and unloading path respectively. The function f(s) in equation (13) can also take different forms. Sheng et al. (2008a) initially used the following function: f (s) s (16)
where ssa is the transition suction and was also called the saturation suction in Sheng et al. (2008a). It is the unique transition suction between saturated and unsaturated states in the SFG model. We note that the number 1 in equation (14) is used to avoid the singularity when ssa 0 and is not truly needed if ssa is not absolutely zero. A better expression would be: vp , vs s vp sa , s s ssa s ssa (15)
This is perhaps the simplest form possible for f (s) and yet guarantees the continuity between saturated and unsaturated states. Even with this simplest form, Zhou & Sheng (2009) showed that the SFG model is able to predict a good set of experimental data on volume change and shear strength, both for reconstituted soils prepared from slurry states and for compacted soils. Due to the f(s) term in equation (13), the normal compression lines will be curved in the v ln p space (Figure 6). In Figure 6a, the compression curves for s 400, 650 1000 kPa are all normal compression lines that do not involve any unloading or reloading. The SFG predictions for these curves were obtained with one single vp value. Figure 6b shows the difference in the estimated yield stress for s 300kPa by equations (5) and (13), respectively. The yield stress is indicated by the meeting points of the unloadingreloading line and the normal compression lines. The parameter vp was allowed to change with suction in Figure 6b. One shortcoming of equation (16) is that the soil compressibility approaches zero as suction increases to infinite. There is also a theoretical discontinuity between unsaturated states and completely dry state (Sr 0). To avoid these prob-
97
1.54 Reloading line 1.50
Measured Predicted by eq. (13) s = 0 kPa s = 400 kPa s = 650 kPa s =1000 kPa
2.25 Reloading NCL fitted by eq. (5)
2.15
Reloading
NCL 1.46 NCLs 1.42 NCL
v
2.05 s=0 s = 300 kPa
NCL fitted by eq. (13) 100 1000
1.38
10
100
1000
10000
1.95 10
p (kPa)
p (kPa)
(b) Compacted Kaolin
(a) Air-dry silty clay
Figure 6. Predicted isotropic compression curves for (a) air-dry silty clay (data by Cunningham et al. 2003), (b) compacted kaolin (data by Sivakumar and Wheeler 2000).
lems, an alternative form of f(s) could be used, for example: f (s) Sr s (17)
This equation will not only guarantee the continuity between saturated and unsaturated states, but also the continuity between unsaturated and the completely dry state (Sr 0). More interestingly, both equations (16) and (17) will lead to the same shear strengthsuction relationship. The degree of saturation in (17) can also be replaced by the effective degree of saturation (Sre ), as suggested by Pereira & Alonso (2009) when discussing Bishops parameter. However, the performance of equation (17) is yet to be validated against experimental data, and the loading-collapse yield surface may become non-unique due to hydraulic hysteresis. The SFG approach seems to be able to overcome some disadvantages of Approach A and Approach B. The main disadvantage of the SFG approach is that it exists only in an incremental form and its integration depends on stress path (Zhang & Lytton 2008; Sheng et al. 2008b). The stress path dependency requires special treatment in the stress integration of the constitutive model (Sheng et al. 2008d). 4 YIELD STRESS VERSUS SUCTION
In the literature of unsaturated soil mechanics, the yield stress of an unsaturated soil is usually assumed to be a function of soil suction. The concept of yield stress in classical elastoplasticity theory refers to the stress level that causes plastic deformation. In other plasticity frameworks such as bounding surface plasticity (Dafalias 1986; Russell & Khalili 2006; Morvan et al. 2010),
hypoplasticity (Kolymbas 1991; Man & Khalili 2008) and generalised plasticity (Pastor et al. 1990; Snchez et al. 2006; Manzanal et al. 2009), plastic deformation occurs along all loading paths including reloading. In this case, a loading function or a bounding surface is used to differentiate unloading from loading. In the discussion below, the yield surface concept is based on the classical elastoplasticity theory, but can be adopted to the loading or bounding surface concepts. Under isotropic stress states, the yield stress is also called the preconsolidation stress. For unsaturated soils, the yield net mean stress, denoted here by pc, is conventionally determined from isotropic compression curves obtained at constant suctions. These compression curves are usually plotted in the space of void ratio versus logarithmic net mean stress. The initial portion of the curve is typically flatter than the ending portion of the curve, if the suction is larger than zero. Such a curve is then approximated by two straight lines, one representing the elastic unloading-reloading line and the other the elastoplastic normal compression line. The intersection point of the two lines is used to define the preconsolidation stress or yield stress (Figure 7a). The yield stress so determined is generally found to increase with increasing suction, irrespective of samples prepared from slurry states or from compacted soils, leading to the socalled loading-collapse yield surface that is widely used in constitutive models for unsaturated soils (Figure 7b). The procedure outlined above for determining the yield stress is based on the assumption that the e ln p relationship for normally consolidated soils at s 0 is linear and may not be consistent with the definition of yield stress. To demonstrate this inconsistency, we should first realise that the isotropic compression curves shown in Figure 7
98
e
s=0
0 < s1 < s2< s3

pc1 < pc2 < pc3
s s3
LC yield surface
s2 s1
s1
pc1 pc2 pc3
s2
s3
ln p
pc1
pc2
pc3
(a) Isotropic compression curves
(b) Variation of yield stress with suction
Figure 7.
Isotropic compression curves under constant suction (s) and derived yield stresses.
are typical of unsaturated soils reconstituted from slurry (e.g. Jennings & Burland 1962) as well as of compacted soils. Because it is difficult to define the yield stress for a compacted soil and is relatively easy to understand the preconsolidation stress for a slurry soil, we use a slurry soil as an example here. Let us assume that the slurry soil has not been consolidated (with a zero preconsolidation stress). The initial yield stress for the soil is then zero (Point A in Figure 8a). We also note that the effective stress for saturated states is constant along the 135o line in the p s space. Drying the slurry soil to suction B under zero mean stress is similar to consolidating the soil to stress E under zero suction (Figure 8a). Indeed, if the air entry value of the soil is larger than suction at B, the length AB is exactly the same as AE, due to the effective stress principle for saturated soils. However if the air entry value is lower than suction at B, the length AB should generally be larger than AE, because a suction increment is generally less effective than an equal stress increment in terms of consolidating the unsaturated soil. Once the soil becomes unsaturated, the yield stress does not necessarily change with suction along the 135o line (denoted by the dashed line). However, the new yield surface always go through the current stress point, as shown in Figure 8b. The elastic zone expands as the suction increases from A to B. The stress points (e.g. B and C) are on the current yield surfaces. Let now the soil be isotropically compressed under the constant suction at Point C (i.e., stress path CD in Figure 8a). The isotropic compression path (CD) is again outside the current elastic zone and the soil at point C is normally consolidated. Therefore, the isotropic compression path (CD) as well as the drying path from B to C is elastoplastic and does not involve a purely elastic portion as Figure 7 indicates, suggesting that the method for determining the yield stress in Figure 7 be conceptually inconsistent. Indeed,
the suction-induced apparent consolidation effect should refer to the increase of the preconsolidation stress at zero suction ( pc (0 ) moves from E to F as suction increases from B to C in Figure 8a), not the preconsolidation stress at the current suction. The same analysis can be done in the effective stresssuction space (Figure 8b). In the effective stress space, the initial yield stress for a slurry soil follows the vertical line that goes through point A. This is also the zero shear strength line, which is commonly assumed to be vertical. However, the stress path for suction increase under constant net mean stress is initially inclined to horizontal by 45o for saturated states. Therefore, the stress path will cross the current yield surface and the drying path AB for a slurry soil is elastoplastic, not purely elastic. Once the soil becomes unsaturated, the stress path will drift away from the 45o line and the yield surface will also drift away from the vertical line (the dashed lines in Figure 8b). However, stress points A, B, C, and D stay on the current yield surfaces and the stress path ABCD causes elastoplastic volume change, in consistency with Figure 8a. In the literature, the yield stress variation with suction is a rather confusing point. Many models adopt three yield surfaces in the net stresssuction space, namely the loading-collapse yield surface, the suction-increase yield surface and the apparent tensile strength surface. Figure 9 shows some examples of these models. The loading-collapse yield surface is used to model the volume collapse when an unsaturated soil is first loaded under constant suction and then wetted under constant stress. The suctionincrease yield surface is used to capture the plastic volume change when an unsaturated soil is dried to a historically high suction. The apparent tensile strength surface defines the zero shear strength or the apparent tensile strength due to suction increase. These yield surfaces are usually defined separately. For example, setting the preconsolidation stress
99
D Initial Yield Stress
B 45o A E F 135o
Current yield surfaces
Current yield surfaces F G

p
45o A E (b) Effective stress suction space
(a) Net stress suction space
Figure 8.
Evolution of the elastic zone during drying and compression of a slurry soil.
Thu et al. (2007a) Alonso et al. (1990) Delage and Graham (1995)
( pc0 ) to zero in the loading-collapse yield function does not recover the apparent tensile strength function. The suction-increase yield surface is usually horizontal or gently sloped (Figure 9) and is not related to the loading-collapse surface or the apparent tensile strength surface. In an alternative approach, Sheng et al. (2008a) showed the relationship between the loadingcollapse surface, the apparent tensile strength surface and the suction-increase yield surface. In the SFG approach proposed by them, the yield stresssuction relationship, the apparent tensile strengthsuction relationship and the shear strengthsuction relationship are all derived from the volume change equation, i.e. equation (13). In this model, the yield stress for a slurry soil that has never been consolidated or dried varies with suction in a unique function. This function also defines the apparent tensile strength surface or the zero shear strength surface in the stresssuction space (the curve through point A in Figure 10). The curve approaches the 45o line as the suction becomes zero or negative (positive pore water pressure). Drying this slurry soil under zero stress (stress path ABC) is somewhat equivalent to consolidating the soil, leading to the yield surface expansion to point C in Figure 10. Therefore, the suction-increase yielding is already included in the yield function and there is no need to define a separate function. If the unsaturated soil at point C is then compressed under constant suction (stress path CD in Figure 10), the yield surface will evolve to the curve that passes through point D in Figure 10. The yield surface in the stress space represents the contours of the hardening parameter, which is usually the plastic volumetric strain. The stress path CD will change the initial shape of the yield surface, because the plastic volumetric strain caused by the isotropic compression depends on the current suction level. The loading-collapse yield function recovers the apparent tensile strength function when the preconsolidation stress at zero suction ( pc0) is set to zero. All these yield surfaces are continuous and smooth in the stresssuction space.
Suction-increase yield surfaces Apparent tensile strength surface
Loading collapse yield surfaces
pc0
Figure 9. Loading-collapse, suction-increase and apparent tensile strength surfaces in various models.
s
Apparent tensile strength B Suction increase yield surface Loading collapse yield surface
45o
45o
45o
pc0
pcn0
Figure 10. Yield stress variation with suction in the SFG model (Sheng et al. 2008a).
As pointed out by Wheeler & Karube (1996) and shown by Sheng et al. (2008a) and Zhang & Lytton (2009b), the apparent tensile strength function, the suction-increase yield function and the loading-collapse yield function are all related to the
100
volumetric model that defines the elastic and elastoplastic volume changes caused by stress and suction changes. If Approach A, i.e. equation (5), is adopted to describe the volume change, the loading-collapse yield surface will take the following form: s ssa pc0 s, ( s ) vp pc p s ( 0 ) vp c0 sa pr , s ssa pr
The function defines the suction-increase surface and is independent of the specific function f(s) used in (13). The apparent tensile strength function is also defined in (23) by setting pc0 0. The loading-collapse surface takes a very similar form: s ssa pcn0 s, pc pcn0 s pc0 f ( s ) ssa ssa ln f ( s ), s ssa p s c0 sa (24) where pcn0 is the new yield stress at zero suction (Figure 10). This function is however dependent on the specific forms of f(s) used in (13). The functions defined in (18) to (24) are all continuous over the entire ranges of possible suction or pore pressure values. However, functions (18) and (21) may not be smooth, dependent on functions vp(s) and (s), respectively. On the other hand, functions (23) and (24) are smooth. All these functions can be incorporated into constitutive models for saturated states. For example, if the modified Cam clay model is used for saturated soil behaviour, the yield function can be generalised to unsaturated states along the suction axis: f q 2 M 2 ( p p0 )( pc p ) 0 (25)
(18)
where pc0 is the yield stress at zero suction, pr a reference stress, and the elastic compression index. The specific shape of this function depends on the variation of the compressibility with suction, i.e. the function vp(s). The suction-increase yield surface can also be derived from equation (5). If the shrinkability (vs) is assumed to be independent of stress, the suction-increase yield surface is simply: s s0 (19)
where s0 is the yield suction. Equation (19) represents a horizontal line in the stresssuction space. Because equation (5) is not defined at zero suction and zero mean stress, the apparent tensile strength surface can not be derived from (18). A separate function is usually introduced: p0 k s (20)
where k was assumed to be a constant in Alonso et al. (1990), but as a function of suction in Georgiadis et al. (2005). If Approach B, i.e. equation (7), is adopted to describe the volume change, the loading-collapse surface can then be written as: s ssa , pc0 ( s ) pc p ( 0 ) , s ssa pr c0 pr
where f is the yield function in the stress space, q is the deviator stress, and M is the slope of the critical state line in q p space. Again, equation (25) is valid for all pore pressure and suction values. 5 SHEAR STRENGTH
(21)
The suction-increase surface is the same as equation (19). The apparent tensile strength surface is recovered from (21) by setting pc0 0 : p0 0 (22)
which represents the vertical line going through the origin of the stress space. If the SFG model, i.e. equations (13) and (15), is adopted to describe the volume change, the yield stress takes the following form: s ssa pc0 s, pc s pc0 ssa ssa ln s , s ssa sa
(23)
The change of shear strength with suction or saturation is one of the main reasons behind rainfallinduced landslides. It is related to the volume change equation (Sheng et al. 2008c). However, this relationship has been overlooked in most existing models for unsaturated soils. If the slope of the critical state line is assumed to be independent of suction, such as supported by experimental data of Toll & Ong (2003), Ng & Chiu (2001) and Thu et al. (2007b), the shear strengthsuction relationship can indeed be derived from the volume change equation. If the slope of the critical state line depends on suction, as supported by data of Toll (1990) and Merchn et al. (2008), two equations are needed to define the shear strengthsuction relationship, namely the volume change equation, and the M(s) function, with M being the slope of
101
the critical state line in the deviatormean stress space. Bishop & Blight (1963) first proposed an effective stress definition to interpret the shear strength of unsaturated soils:
c n tan c n s tan
c n tan
(26)
where is the shear strength, c is the effective cohesion for saturated states and is usually assumed to be zero, n and n are respectively the effective and net normal stress on the failure plane, is the effective friction angle of the soil, is the wellknown Bishops effective stress parameter, and c is the apparent cohesion which includes the friction term due to suction, i.e. c c s tan . Fredlund et al. (1978) proposed the following relationship which conveniently separates the shear strength due to stress from that due to suction:
b c n ua tan ua uw tan
c n ua tan
(27)
where is the shear strength, c the effective cohesion for saturated states and is usually assumed to be zero, n the normal stress on the failure plane, the effective friction angle of the soil, and b the frictional angle due to suction. Obviously, if b is set to in equation (27), the Coulomb friction criterion in terms of effective stress for saturated soils is recovered. The shear strength of an unsaturated soil is fully defined if the apparent cohesion ( c ) or the friction angle due to suction (b), and the friction angle due to stress in equation (27) are known. A number of alternative equations have been used
600
Test data (after Thu et al. 2007b) Predictions
5
in the literature to define c or b (e.g., Fredlund et al. 1996; berg & Sllfors 1997; Vanapalli et al. 1996; Khalili & Khabbaz 1998; Toll & Ong 2003; Miao et al. 2007). Most of these equations are empirically based and are defined independently from the volume change equation, which, if incorporated into a constitutive model, may lead to inconsistence with the yield surface as discussed in the Section above. Rojas (2008) presented an analytical expression for Bishops parameter in terms of the saturated fraction and Sr of the unsaturated fraction of the soil and used it to interpret shear strength. One issue with this approach is that can not be easily determined either experimentally or theoretically. A similar approach was proposed by Pereira and Alonso (2009) where is expressed in terms of micro- and macro-degrees of saturation. Sheng et al. (2010) recently compared various shear strength equations for unsaturated soils against a large number of data sets. The equations include those empirically based and those embedded in constitutive models. The equations studied by Sheng et al. (2010) are listed in Table 1. Figure 11 presents one example of such comparisons. The parameters used in the shear strength equations are given in Table 1. The comparative study by Sheng et al. (2010) reveals that: 1. All shear strength equations in the literature can be formulated either in form of equation (26) or (27). Unlike the volume change equations, there is little difference in formulating shear strength in one single stress variable or in two independent stress variables. 2. The performance of the shear strength equations in predicting experimental data depends on the careful determination of material parameters and also on the specific data set. A shear strength equation may predict one data set better than other data sets.
600
Test data (after Thu et al. 2007b) Predictions
5
Deviator Stress, kPa
500
7 8 6 3 2
500
7 1 1 8 6 3 4 4 2
400
400
300 Confining pressure: 100 kPa 200
300 Confining pressure: 200 kPa 200
100
200
300
400
100
200
300
400
Suction, kPa
Suction, kPa
Figure 11. Predictions of triaxial test data on compacted kaolin clay by various shear strength equations (data by Thu et al. 2007b).
102
Table 1. Shear strength equations and parameters used for compacted kaolin clay. Equations tanb/ Parameters SWCC 1.3, s 0.659, SWCC s 0.659, SWCC k 1.2, SWCC 1. berg & Sr Sllfors (1997) 2. Fredlund et al. (Sr) (1996) 3. Vanapalli et al. r (1996) s r 4. Toll & Ong (2003) 5. Alonso et al. (1990) 6. Sun et al. (2000) 7. Khalili & Khabbaz (1998) 8. Sheng et al. (2008a)
Sr Sr 2 Sr1 Sr 2
a sa (sae/s)r
0.4 110 kPa sae 60 kPa, r 0.55
ssa ssa s ssa 25 kPa ln s s ssa
of coupled models soon followed (e.g. Wheeler et al. 2003; Sheng et al. 2004). With respect to hydraulic behaviour of unsaturated soils, many models (van Genuchten 1980; Fredlund & Xing 1994) take advantage of the fact that the influence of suction on degree of saturation is more significant than the influence of deformation. The dependency of degree of saturation on suction is described by a soil-water characteristic curve (also called soil-water retention curve, SWRC). Only until recently, the effects of deformation on SWCCs have been considered (e.g., Gallipoli et al. 2003b; Wheeler et al. 2003; Sun et al. 2007b; Miller et al. 2008; Zhou 2009). As pointed out by Wheeler et al. (2003), the mechanical behaviour of an unsaturated soil depends on degree of saturation even if the suction, net stress and specific volume are kept the same for the soil. Separate treatment of mechanical and hydraulic components in modelling unsaturated soil behaviour has certain limitations in reproducing experimental observation. It would be difficult to consider the saturation dependency in a mechanical model that is independent of the hydraulic behaviour. Similarly, a hydraulic model that is independent of mechanical behaviour can not easily take into account the effects of soil density on the SWCC. Experimental data generally demonstrate the following points: 1. A SWCC obtained under a higher net mean stress tends to shift towards the higher suction (Matyas & Radhakrishna 1968; Ng & Pang 2000; Gallipoli et al. 2003b; Lee et al. 2005; Tarantino & Tombolato 2005). This means that the incremental relationship between degree of saturation (Sr) and suction (s) depends on net mean stress ( p ) or soil density. 2. When the suction is kept constant, isotropic loading or unloading can also change the degree of saturation of an unsaturated soil (Wheeler et al. 2003). This implies that the degree of saturation is related to stress or soil density when the suction is kept constant. One of the early models that fully couple the hydraulic and mechanical components of unsaturated soil behaviour is that by Wheeler et al. (2003). Models that appeared before or soon after Wheeler et al. (2003) tend to accentuate the influences of the hydraulic component on the mechanical component, not vice versa (e.g., Vaunat et al. 2000; Sheng et al. 2004; Nuth & Laloui 2008b). The interaction between the mechanical and hydraulic components in the model by Wheeler et al. (2003) was realised through the use of the average soil skeleton stress (effective stress), the modified suction and the coupling between the loading-collapse and suctionincrease and suction-decrease surfaces. The average * ) is an amalgam of stress, soil skeleton stress ( ij
3. The shear strength equations that incorporate the soil-water characteristic curve (SWCC) generally require more parameters. This group of equations seem to provide reasonable prediction of shear strength for unsaturated soils (e.g. Fredlund et al. 1996; Vanapalli et al. 1996; Toll & Ong 2003). However, some equations are sensitive to the residual suction (Vanapalli et al. 1996; Toll & Ong 2003), which can be difficult to determine accurately from the SWCC. 4. Simpler shear strength equations that are embedded in constitutive models appear to provide reasonable predictions of shear strength (Khalili & Khabbaz 1999; Sun et al. 2000; Sheng et al. 2008a). These equations contain typically one or two parameters. However, these equations usually do not predict any peak value of the shear strength attained at an intermediate suction. 6 WATER RETENTION BEHAVIOUR AND HYDRO-MECHANICAL COUPLING
The issue of interaction between the mechanical and hydraulic behaviour was perhaps first raised by Wheeler (1996) and then by Dangla et al. (1997). The first complete model that deals with coupled hydro-mechanical behaviour of unsaturated soils was perhaps due to Vaunat et al. (2000). A number
103
suction and degree of saturation. The modified suction (s*) is a combination of suction and porosity. Therefore, the influence of hydraulic behaviour on the stress-strain relationship is considered via the definition of the average stress. The influence of porosity on the saturation-suction relationship is considered via the definition of the modified suction. The model by Wheeler et al. (2003) is one of the few models that are qualitatively tenable in terms of coupling mechanical behaviour with hydraulic behaviour for unsaturated soils. However, the use of the modified suction and the soil skeleton stress, which is one of the advantages that makes the model rigorously consistent in thermodynamics, can become a disadvantage as well, particularly in terms of quantitative prediction and the application of the model. For example, one of the difficulties in using this model is to quantify the synchronised movement between the loadingcollapse (LC) surface and the suction-increase (SI) and suction-decrease (SD) surfaces. This synchronicity can not easily be calibrated by laboratory experiments (Raveendiraraj 2009) or defined theoretically. In more recent models, the influences of mechanical properties on the hydraulic behaviour are usually modelled via the dependency of the SWCC on soil volume (Gallipoli et al. 2003b; Tarantino 2009), soil density (Sun et al. 2007b; Man 2010), or volumetric strain (Nuth & Laloui 2008a). Gallipoli et al. (2003b) suggested including a function of specific volume (v) in the SWCC equation of van Genuchten (1980). Tarantino (2009) showed there is a unique relationship between the water ratio (product of Sr and e) and the matric suction and used this relationship to modify van Genuchtens equation. The modified van Genuchtens equation takes a similar form as Gallipoli et al. (2003b). It is also common to express the SWCC equation in incremental forms. For example, Sun et al. (2007b) proposed a hydraulic model in the following form: dSr se de ss ds/s (28)
soil and deformation of the soil can shift this intrinsic SWCC along the suction axis. The shift is governed by an air entry value that depends on the volumetric strain. The models by Sun et al. (2007b), Nuth & Laloui (2008a), Man (2010) and many others (e.g. Sheng et al. 2004, 2008a; Nuth & Laloui 2008b; Khalili et al. 2008; Zhou 2009) essentially all adopt a water retention equation in the following form: dSr () ds () dv (29)
This equation is not wrong, but the embedded Sr s relationship is for constant volume (dv 0). Therefore, it does not recover the conventional SWCC equations, which are obtained under constant stress. The volume change along a conventional SWCC can be significant. It is not common to obtain an Sr s relationship for constant soil volume. The vast available data on SWCCs that were obtained under constant stresses would be of limited use in these models. In addition, neglecting the volume change along SWCC can lead to inconsistent prediction of the degree of saturation during undrained compression, an issue raised by Zhang & Lytton (2008). Sheng & Zhou (2010) proposed a new method for coupling hydraulic with mechanical behaviour. This new method is based on the fact that SWCCs are obtained under constant stress. In Sheng & Zhou (2010), the volume change behaviour and the water retention behaviour under isotropic stress states are represented by the following incremental equations, respectively: d v A dp B ds Sr (1 Sr )m A dp n (30)
dSr E ds
S S E B r (1 Sr )m ds r (1 Sr )m de n e
(31)
where ss is the slope of main drying or wetting curve, and se the slope of degree of saturation versus void ratio curve under constant suction. In theory, ss in the equation above can only be determined from constant-volume tests (de 0). However, such tests are not common. Man (2010) used a similar equation as (28). In his model both air entry value (sae) and the slope of main drying curve (ss) vary with void ratio. Nuth & Laloui (2008a) provided an alternative approach of modelling SWCC for a deforming soil. They assume there is an intrinsic SWCC for a non-deforming
where parameters A and B are related to the specific volume change equation as discussed in Section 3, parameter E is the gradient of the conventional SWCC, e is the void ratio, n is the porosity, and m is a fitting parameter. If the SFG model is used to describe the volume change, parameters A and B would then take the form: A
vp
p f (s)
B
vs p f (s)
(32)
104
The Sr s relationship is defined for constant stress (dp 0) in equation (31) and hence parameter E refers to the gradient of the SWCC: E d SrSWCC ( s ) ds
(33)
where SrSWCC represents the conventional SWCC equation. The void ratio in equation (31) refers to the value at the current stress and suction. It is clear from equation (31) that the Sr s relationship for constant volume (de 0) is more complex than the conventional SWCC equation (dSr Eds). Equation (31) was proposed based on experimental observation as well as the intrinsic phase relationship for undrained condition: dSr Sr d v , dw 0 n (34)
where w is the gravimetric water content. Equation (34) actually imposes a constraint on suction change under undrained compression. This constraint is obtained by substituting (34) into (31):
S S 1 S A dp nE B ds,
r r r m
dw 0 (35)
Zhang & Lytton (2008) recently noted that some models fall short in predicting undrained behaviour of unsaturated soils. The specific issue raised is illustrated in Figure 12. Assume the initial state of a soil is inside the elastic zone, but on the main wetting curve, i.e. point A in Figure 12a. Compressing the soil under undrained condition will lead to some suction decrease (Sun et al. 2008; Tang et al. 2008), say to point B. Assume B is still inside the elastic zone. Unloading from B to A will recover the initial volume of the soil and hence the initial degree of saturation should be recovered as well. However, if the volume change
s SI surface SD surface A D C B
p
along the SWCC is neglected, the change of Sr along path ACB would follow the main wetting curve and is hence elastoplastic, whereas the change of Sr along path BDA would follow the scanning curve and is hence elastic, leading to inconsistent change of Sr (Figure 12b). It would then seem unlikely that a model where the irreversible volume change is not synchronised with the irreversible saturation change could lead to a consistent prediction of saturation change over the closed path ABA (Zhang & Lytton 2009a). This inconsistency is actually due to the assumption that the main wetting curve is defined for constant volume. If equation (31) is used, the inconsistence in Figure 12b will be avoided. Because of equation (34), the model will predict no change of Sr as long as v 0, irrespective of evolution of the suction-decrease surface. Equation (34) is satisfied as long as the suction changes according to (35). It is easy to understand that the loading path ACB is not on the initial main wetting curve and the unloading path BDA is not on the scanning curve, because the mean stress is not constant along those stress paths. The main wetting curve at point A also shifts to that at point B, as the mean stress changes (Figure 12c). The synchronised evolution of the LC, SI and SD surfaces (as in Wheeler et al. 2003) is not necessary for the consistent prediction in Figure 12c. Indeed, the suction path can be elastoplastic albeit the elastic stress path. In the case of constant stress, the following equation can be derived from equation (31): Sr S (1 Sr )m r , dp 0 e0 e0 (36)
The void ratio (e0) in equation (36) refers to the initial void ratio at the current stress and at Sr 1. It can also be interpreted as the initial void ratio at the start of the SWCC tests. Equation (36) shows that
s
LC yield surface
Main drying curve A C A D Scanning curve A
Main drying curve C D B
B Main wetting curve Sr (b) inconsistent change of Sr
Main wetting curve Sr (c) consistent change of Sr
(a) stress path in net stress-suction space
Figure 12.
Qualitative analysis of isotropic compression under undrained condition.
105
Sr
the SWCC for a soil shifts with its initial void ratio. This is similar to the approach by Gallipoli et al. (2003b) where the van Genuchten equation was modified to include the initial void ratio. Indeed, their SWCC equation can be re-written as: Sr mn e0
1/ m Sr (1 Sr )
1.1 1 0.9 0.8
e0 =0.444 e0 =0.474 e0 =0.514 e0 =0.517
e0
(37)
0.7 0.6
Predictions
0.1 1 10
where m and n are two fitting parameters in the original van Genuchten equation, and is another parameter introduced by Gallipoli et al. (2003b). If the product (mn) was set to 1, equation (37) would be equivalent to (36). Zhou et al. (2010) showed that the intrinsic phase relationship requires: 1 Sr Sr Sr e e e (38)
s (kPa)
100
1000
10000
Figure 14. SWCCs for specimens compacted at optimum water content (data after Vanapalli et al. 1999).
The above constraint is satisfied if mn 1 in equation (37). It is also interesting to note that all the numerical examples in Gallipoli et al. (2003b) used a value of 1.1 for mn. Equation (36) can be integrated either analytically for certain m values or numerically in more general cases. Because equation (36) is in an incremental form, integration of the equation requires one specific SWCC equation that corresponds to a reference initial void ratio. In other words, the conventional SWCC equation is only used for the reference initial void ratio and the new SWCC for a new initial void ratio is obtained by integration of (36). The model by Sheng & Zhou (2010) is validated against a variety of data sets. In Figure 13, the predicted SWCCs are compared with the data set by Tarantino (2009) on reconstituted Barcelona silt. In the prediction, the SWCC for e0 0.62 was fitted by van Genuchten equation. The other SWCCs were predicted with m 0.2. In Figure 14, the data
1.10
by Vanapalli et al. (1999) on a compacted till were used. The SWCC for e0 0.517 was fitted by the van Genuchten equation, while the other SWCCs are predicted by equation (36) with m 0.03. Figure 14 shows that both the slope and the air entry value of the SWCCs change with the initial void ratio. In the two cases studied, the model by Sheng & Zhou (2010) seems to be able to capture the effect of initial void ratio on the soil water retention behaviour. 7 FINITE ELEMENT IMPLEMENTATION
One of the ultimate goals of constitutive modelling is to implement the model in a numerical method to solve boundary value problems. A constitutive model can generally be formulated in the following incremental form (Sheng et al. 2004, 2008c): d Dep d T W ep d ds H (39)
1.00
0.90
0.80
0.70
e0=0.62 e0=0.54 e0=0.50
0.60
Predictions
0.50 10 100
s (kPa)
1000
10000
Figure 13. SWRCs of reconstituted Barcelona silt with different initial voids ratios (data after Tarantino 2009).
where is the stress vector, is the vector of soil skeleton strain, is the volumetric water content, Dep, Wep, T and H are constitutive matrices (Sheng et al. 2008c). In the displacement finite element method, the nodal displacements and pore pressures are first solved from the equilibrium and continuity equations, based on the current stress states and the current volumetric water content. The strain and suction increments are then derived from the displacements and pore pressures. For given strain and suction increments, the current stress vector, the volumetric water content and the internal variables must be updated according to equation (39). This update is generally carried out by numerical integration.
Sr
106
One of the main challenges in integrating equation (39) arises from the non-convexity of the yield surface around the transition between saturated and unsaturated states (Wheeler et al. 2002, Sheng et al. 2003b; Sheng 2003). The non-convexity seems to persist irrespective of the stress variables and is demonstrated in Figure 15 (Sheng et al. 2008d). Both implicit and explicit schemes have been used to integrate unsaturated soils models. In implicit schemes, all gradients and functions are estimated at advanced unknown stress states and the solution is achieved by iteration. This group of stress integration schemes include Vaunat et al. (2000), Borja (2004), Hoyos & Arduino (2008), Zhang & Zhou (2008), and Tamagnini & De Gennaro (2008). They usually do not deal with the non-convex problem by avoiding the transition between saturated and unsaturated states. Borja (2004) was perhaps the only one in this group who noticed the problem. He suggested keeping the step size sufficiently small to avoid overshooting the plastic zone by the elastic trial stress (Figure 15). Otherwise, there is currently no general method used to tackle the non-convexity problem in implicit schemes. On the other hand, explicit schemes estimate the gradients and functions at the current known stress states and proceed in an incremental fashion. These methods are theoretically more appropriate for non-convex models (Pedroso et al. 2008). Gonzlez & Gens (2010) compared both an implicit and an explicit scheme for integrating the BBM and found that the latter is more robust and efficient. However, explicit schemes usually require to determine the intersection between the current yield surface and the elastic trial stress path and some substepping methods to control the integration error (Sheng et al. 2003a, 2003b; Sheng et al. 2008c, 2008d; Snchez et al. 2008; Soowski & Gallipoli 2009a, 2009b). A key issue in integrating the incremental stressstrain relationships using an explicit scheme is thus to find the intersection between the elastic trial stress path and the current yield surface. The most complex situation occurs when the yield surs
Elastic zone unsaturated 45 saturated
o
face is crossed more than once, such as shown in Figure 15a. However, it is not possible to know a priori how many times the yield surface is crossed. Therefore, for non-convex yield surfaces, the key task is to find the very first intersection for any possible stress and hydraulic path. Finding the intersection between the elastic trial stress increment and the current yield surface can be cast as a problem of finding the multiple roots of a nonlinear equation: f ( ) f ( , s , z ) 0 (40)
where 0 1, f ( , s, z ) is the yield function, z is a set of internal variables, and subscript indicates the quantity is estimated at strain increment and suction increment s. Pedroso et al. (2008) proposed a novel method to bracket the roots (). The method is illustrated in Figure 16. For a given increment ( 1), the number of root of f() is first computed. If there are more than one roots, the increment is divided into two equal sub-increments. The number of roots of each sub-increment is then computed. If the first sub-increment contains more than one root, it is further divided into two sub-increments. This process is repeated until the first sub-increment contains at most one root (Figure 16). Once the roots are bracketed, the solution of the first root can be found by using numerical methods such as the Pegasus method (Sloan et al. 2001). Sheng et al. (2008d) applied the method by Pedroso et al. (2008) to integrate the SFG model and found that the method can indeed provide an accurate solution of the intersection problem. However, this method is found to be computationally expensive. The formula used to find the number of roots of a nonlinear function requires both the first and second orders of gradients of the function and requires numerical integration. Furthermore, the root-finding procedure must be applied for all suction increments near the nonconvexity. Further research is required to improve its efficiency.
s
LC surface
s
Elastic zone unsaturated saturated
Elastic zone unsaturated 45o
Elastic trial stress
p
p
saturated
(a) net stress (Alonso et al., 1990)
(b) effective stress (Sheng et al., 2004)
(c) net stress (Sheng et al., 2008a)
Figure 15.
Non-convexity of yield surfaces in unsaturated soil models in the suction-stress space.
107
1/8
1/4
1/2

1
Figure 16. Bracketing the roots for nonlinear function according to Pedroso et al. (2008).
istic curves. Neglecting this volume change can lead to inconsistent prediction of volume and saturation changes. 8. Unsaturated soil models are characterised by non-convex yield surfaces at the transition between saturated and unsaturated states. This non-convexity, if handled rigorously, can significantly complicate the implementation of these models into finite element codes. In this case, an explicit stress integration scheme incorporating an efficient root search algorithm is preferred. ACKNOWLEDGEMENT The author would like to thank: EE Alonso & A Gens for their invitation to write this general report; YJ Cui, DG Fredlund, D Gallipoli, JM Pereira, DA Sun & X Zhang for their valuable comments on the draft report; DG Fredlund, A Gens, DM Pedroso, DA Sun & AN Zhou for their contributions to the work presented in this report; the Australian Research Council for its financial support. REFERENCES
Al-Badran Y & Schanz T. 2009. Yielding surface model of volume change characteristics for unsaturated fine grained soils. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 863871. Alonso EE, Gens A & Josa A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40: 405430. Alonso EE, Vaunat J & Gens A. 1999. Modelling the mechanical behaviour of expansive clays. Engng Geol., 54: 173183. Arson C and Gatmiri B. 2008. On damage modelling in unsaturated clay rocks. Physics Chemistry Earth, 33: 635640. Baker R & Frydman S. 2009. Unsaturated soil mechanics: Critical review of physical foundations. Engng Geol., 106: 2639 Bian H & Shahrour I. 2009. Numerical model for unsaturated sandy soils under cyclic loading: Application to liquefaction. Soil Dynamics Earthquake Engng, 29: 237244. Bishop AW & Blight GE. 1963. Some aspects of effective stress in saturated and partly saturated soils. Gotechnique, 13: 177197. Bolzon G, Schrefler BA & Zienkiewicz OC. 1996. Elastoplastic soil constitutive laws generalised to partially saturated states. Gotechnique, 46: 279289. Borja RI. 2004. Cam clay plasticity, Part V: A mathematical framework for three phase deformation and strain localization analysis of partially saturated porous
CONCLUSIONS
Some conclusions can be drawn from this review of constitutive modelling of unsaturated soils: 1. Partial saturation is only a state of soil. Constitutive models for soils should represent the soil behaviour over entire ranges of possible stress, pore pressure and suction values. This requires a coherent merge of fundamental soil mechanics principles for saturated and unsaturated states. 2. The volume change behaviour is one of the most fundamental properties of soils. For soils in unsaturated states, the volume change equation also underpins the yield stresssuction and shear strengthsuction relationships. It also affects the soil water retention behaviour. 3. Three groups of volumetric models are compared and it is shown that each has advantages and disadvantages. There is certainly room for improvement of all these models. One observation here is that it seems difficult to describe the volume change of unsaturated soils in terms of one single stress variable. 4. The methods used to define the loading-collapse, suction-increase and apparent tensile strength surfaces should be consistent with the volume change equation. In addition, these surfaces and the shear strength function are all related to each other and hence should be defined consistently with each other. 5. Unlike the volume change equations, there is little difference in formulating shear strength equations in one single stress variable or in two independent stress variables. All shear strength equations in the literature can be formulated in one form or the other. 6. The performance of the shear strength equations in predicting experimental data depends on the careful determination of material parameters and also on the specific data set. A shear strength equation may predict one data set better than other data sets. 7. When coupling the hydraulic component with the mechanical component in a constitutive model, it is recommended to take into account the volume change along soil-water character-
108
media. Comp. Methods Applied Mechanics Engng, 193: 530138. Buscarnera G & Nova R. 2009. An elastoplastic strainhardening model for soil allowing for hydraulic bondingdebonding effects. Int. J. Numer. Anal. Methods Geomechanics, 33: 10551086. Cardoso R & Alonso EE. 2009. Degradation of compacted marls: a microstructural investigation. Soils & Foundations, 49: 315328. Cleall PJ, Seetharam SC & Thomas HR. 2007. Inclusion of some aspects of chemical behaviour of unsaturated soil in thermo/hydro/chemical/mechanical models. I: Model development. ASCE J. of Engng Mechanics, 133: 338347. Cui YJ & Delage P. 1996. Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique, 46(2): 291311. Cui YJ, Sultan N & Delage P. 2000. A thermomechanical model for clays. Can. Geotech. J., 37: 607620. Cui YJ & Sun DA. 2009. Constitutive modelling: From isothermal to non-isothermal behaviour of unsaturated soils. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 493506. Coussy O, Pereira JM & Vaunat J. 2010. Revisiting the thermodynamics of hardening plasticity for unsaturated soils. Comput. Geotech., 37, 207215. Cunningham MR, Ridley AM, Dineen K & Burland JB. 2003. The mechanical behaviour of a reconstituted unsaturated silty clay. Gotechnique, 53: 183194. Dafalias YF. 1986. Bounding surface plasticity. I: mathematical foundation and hypoelasticity. ASCE J. Engng Mechanics, 112: 966987. Dangla P, Malinsky L & Coussy O. 1997. Plasticity and imbibition-drainge curves for unsaturated soils: a unified approach. In: Numerical Models in Geomechanics (NUMOG VI, Pietruszczak S & Pande GN, eds), Balkema Rotterdam, pp. 141146. Delage P & Graham J. 1996. Understanding the behaviour of unsaturated soils requires reliable conceptual models: state of the art report. In: Unsaturated Soils (Alonso EE & Delage P, eds), Balkema, Rotterdam. vol. 3, pp. 12231256. DOnza F, Gallipoli D & Wheeler SJ. 2010. Effect of anisotropy on the prediction of unsaturated soil response under triaxial and oedometric conditions. Proc. 5th Int. Conf. Unsaturated Soils. Taylor & Francis (In press). Fityus SG & Buzzi O. 2009. The place of expansive clays in the framework of unsaturated soil mechanics. Applied Clay Science, 42: 150155. Fredlund DG, Morgenstern NR & Widger A. 1978. Shear strength of unsaturated soils. Can. Geotech. J., 15: 313321. Fredlund DG & Rahardjo H 1993, Soil Mechanics for Unsaturated Soils, John Wiley & Sons, New York. Fredlund DG, Xing A, Fredlund MD & Barbour SL. 1996. Relationship of the unsaturated soil shear strength to the soil-water characteristic curve. Can. Geotech. J., 33: 440448. Fredlund DG & Xing A. 1994. Equations for the soilwater characteristic curve. Can. Geotech. J., 31: 521532. Gallipoli D, Gens A, Chen G & DOnza F. 2008. Modelling unsaturated soil behaviour during normal con-
solidation and at critical state. Comput. Geotech., 35: 825834. Gallipoli D, Gens A, Sharma R & Vaunat J. 2003a. An elastoplastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Gotechnique, 53: 123135. Gallipoli D, Wheeler SJ & Karstunen M. 2003b. Modelling of variation of degree of saturation in a deformable unsaturated soil. Gotechnique, 53: 105112. Gens A. 1996. Constitutive modelling: Application to compacted soils. In: Unsaturated Soils (Alonso EE & Delage P, eds), Balkema, Rotterdam, vol. 3, pp. 11791200. Gens A. 2009. Some issues in constitutive modelling of unsaturated soils. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 613626. Gens A. 2010. Soil-environmental interactions in geotechnical engineering. Gotechnique, 60: 374. Gens A & Alonso EE. 1992. A framework for the behaviour of unsaturated expansive clays. Can. Geotech. J., 29: 10131032 Gens A, Snchez M & Sheng D. 2006. On constitutive modelling of unsaturated soils, Acta Geotechnica, 1: 137147. Gens A, Guimars L do N, Snchez M & Sheng D. 2008. Developments in modelling the generalised behaviour of unsaturated soils. In Unsaturated Soils: Advances in Geo-Engineering (Toll DG, Augarde CE, Gallipoli D & Wheeler SJ, eds), CRC Press, pp. 5362. Gili JA & Alonso EE. 2002. Microstructural deformation mechanisms of unsaturated granular soils. Int. J. Numer. Anal. Methods Geomechanics, 26: 433468. Gray WG & Schrefler BA. 2001. Thermodynamic approach to effective stress in partially saturated porous media. European Journal of Mechanics A/Solids, 20: 521538. Georgiadis K, Potts DM & Zdravkovic L. 2005. Threedimensional constitutive model for partially and fully saturated soils. Int. J. Geomechanics, 5: 244255. Gonzlez NA & Gens A. 2010. Evaluation of a constitutive model for unsaturated soils: stress variable and numerical implementation. Proc. 5th Int. Conf. Unsaturated Soils. Taylor & Francis (In press). Guimars L do N, Gens A & Olivella S. 2007. Coupled termo-hydro-mechanical and chemical analysis expansive clay subjected to heating and hydration. Transport in Porous Media, 66: 341372. Houston SL. 2002. Applied Unsaturated Soil Mechanics, State of the Art Report, In: Unsaturated Soils (Juca JFT, de Campos TMP, Marinho FAM, eds), Balkema, Lisse, Vol. 3, pp. 11271134. Hoyos LR, Arduino P. 2008. Implicit algorithms in modeling unsaturated soil response in three-invariant stress space. Int. J. Geomechanics. 8: 266273. Hutter K, Laloui L & Vulliet L. 1999. Thermodynamically based mixture models for saturated and unsaturated soils. Mechanics of Cohesive Frictional Materials, 4: 295338. Jennings JEB & Burland JB. 1962. Limitations to the use of effective stresses in partly saturated soils. Gotechnique, 12: 125144.
109
Jiang MJ, Leroueil S & Konrad JM. 2004. Insight into shear strength functions of unsaturated granulates by DEM analyses. Comput. Geotech., 31: 473489. Jommi C. 2000. Remarks on the constitutive modelling of unsaturated soils. In: Experimental Evidence & Theoretical Approaches in Unsaturated Soils (Tarantino A & Mancuso C, eds), Balkema, Rotterdam, 139153. Katti DR, Schmidt SR, Ghosh P & Katti KS. 2007. Molecular modelling of the mechanical behaviour and interactions in dry and slightly hydrated sodium montmorillonite interlayer. Can. Geotech. J., 44: 425435. Khalili N, Habte MA & Zargarbashi S. 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Comput. Geotech., 35: 872889. Khalili N & Khabbaz MH. 1998. A unique relationship for the determination of the shear strength of unsaturated soils. Gotechnique, 48: 681687. Kikumoto M, Kyokawa H, Nakai T & Shahin HM. 2010. A simple elastoplastic model for unsaturated soils and interpretations of collapse and compaction behaviours. Proc. 5th Int. Conf. Unsaturated Soils. Taylor & Francis (In press). Kimoto S, Oka F, Fushita T & Fujiwaki M. 2007. A chemothermo-mechanically coupled numerical simulation of the subsurface ground deformations due to methane hydrate dissociation. Comput. Geotech., 34: 216228. Kohgo Y, Nakano M & Miyazaki T. 1993. Theoretical aspects of constitutive modelling for unsaturated soils. Soils & Foundations, 33: 4963. Kohgo Y. 2003. Review of constitutive models for unsaturated soils and initial-boundary value analyses. In Unsaturated SoilsGeotechnical and Geoenvironmental Issues (Proc. 2 Asian Conf, Karube D, Iizuka A, Kato S, Kawai K, Tateyama K, eds), pp. 2140. Kohler R & Hofstetter G. 2008. A cap model for partially saturated soils. Int. J. Numer. Anal. Methods Geomechanics. 32: 9811004. Kolymbas D. 1991. An outline of hypoplasticity, Archive Applied Mechanics, 61: 143151. Kurucuk N, Kodikara J & Fredlund DG. 2009. Evolution of the compaction process: Experimental study preliminary results. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 887893. Laloui L & Nuth M, 2008. On the use of the generalised effective stress in the constitutive modelling of unsaturated soils. Comput. Geotech., 36: 2023. Lee IM, Sung SG & Cho GC. 2005. Effect of stress state on the unsaturated shear strength of a weathered granite. Can. Geotech. J., 42: 624631. Li JK, Zhang LM and Fredlund DG. 2009. Permeability tensor and REV saturated cracked soil. Can. Geotech. J., 46: 928942. Li XS. 2007. Thermodynamics-based constitutive framework for unsaturated soils, 1: Theory. Gotechnique, 57: 411422. Lloret M, Snchez M & Wheeler SJ. 2009. Generalised elastoplastic stress-strain and modified suction-degree of saturation relations of fully coupled model. In: Unsaturated SoilsTheoretical & Numerical Advances
in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 667672. Loret B, Hueckel T & Gajo A. 2002. Chemo-mechanical coupling in saturated porous media: elastic-plastic behaviour of homoionic expansive clays. Int. J. Solids & Structures, 39: 27732806. Loret B & Khalili N. 2002. An effective stress elastic plastic model for unsaturated porous media. Mechanics of Materials, 34: 97116. Manzanal D, Pastor M & Fernndez Merodo JA. 2009. A unified constitutive model for unsaturated soils based on generalised plasticity theory. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 655660. Marshall TJ, Holmes JW & Rose CW. 1996. Soil Physics. Cambridge University Press, New York. Man D. 2010. Predicting the dependency of a degree of saturation on void ratio and suction using effective stress principle for unsaturated soils. Int. J. Numer. Anal. Methods Geomechanics, 34: 7390. Man D & Khalili N. 2008. A hypoplastic model for mechanical response of unsaturated soils. Int. J. Numer. Anal. Methods Geomechanics, 32: 19031926. Matyas EL & Radhakrishna HS. 1968. Volume change characteristics of partly saturated soils. Gotechnique, 18: 432448. Merchn V, Vaunat J, Romero E and Meca T. Experimental study of the influence of suction on the residual friction angle of clays. In: Unsaturated Soils: Advances in Geo-Engineering (Toll DG, Augarde CE, Gallipoli D & Wheeler SJ, eds), CRC Press, pp. 423428. Miao L, Houston SL, Cui Y & Yuan J. 2007. Relationship between soil structure mechanical behaviour for an expansive unsaturated clay. Can. Geotech. J., 44: 126137. Miller GA, Khoury CN, Muraleetharan KK, Liu C & Kibbey TCG. 2008. Effects of soil skeleton deformations on hysteretic soil water characteristic curves: Experiments and simulations. Water Resources Research, 44: W00C06. Modaressi A & Modaressi H. 1995. Thermoplastic constitutive model for unsaturated soils: A prospective approach. In: Numerical Models in Geomechanics (Pande GN & Pietruszczak S, eds), Balkema, Rotterdam, pp. 4550. Morvan M, Wong H & Branque D. 2010. An unsaturated soil model with minimal number of parameters based on bounding surface plasticity. Int. J. Numer. Anal. Methods Geomechanics, 34 (in press, doi: 10.1002/ nag.871). Ng CWW & Chiu ACF. 2001. Behaviour of loosely compacted unsaturated volcanic soil. J. Geotech. Geoenviron. Engng, 127: 10271036. Ng CWW, Cui YJ, Chen R & Delage P. 2007. The axistranslation and osmotic techniques in shear testing of unsaturated soils. Soils & Foundations, 47: 675684. Ng CWW & Pang YW. 2000. Influence of stress state on soil-water characteristics and slope stability. J. Geotech. Geoenviron. Engng, 126: 157166. Nuth M & Laloui L. 2008a. Advances in modelling hysteretic water retention curve in deformable soils. Comput. Geotech., 35: 835844.
110
Nuth M & Laloui L. 2008b. Effective stress concept in unsaturated soils: clarification and validation of a unified framework. Int. J. Numer. Anal. Methods Geomechanics, 32: 771801. berg A & Sllfors G. 1997. Determination of shear strength parameters of unsaturated silts and sands based on the water retention curve. Geotech. Testing Journal, 20:4048. Oldecop LA and Alonso EE. 2007. Theoretical investigation of the time-dependent behaviour of rockfill. Gotechnique, 57: 289301. Pastor M, Zienkiewicz O & Chan A. 1990. Generalized plasticity and the modelling of soil behaviour. Int. J. Numer. Anal. Methods Geomechanics, 14: 151190. Pereira JM & Alonso EE. 2009. Insights into the links between microstructure and Bishops parameter. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 685690. Pereira JM & de Gennaro V. 2010. On the time-dependent behaviour of unsaturated geomaterials. Proc. 5th Int. Conf. on Unsaturated Soils. Taylor & Francis (In press). Pereira JM, Wong H, Dubujet P & Dangla P. 2005. Adaptation of existing behaviour models to unsaturated states: Application to CJS model. Int. J. Numer. Anal. Methods Geomechanics, 29: 11271155. Pedroso DM, Sheng D & Sloan SW. 2008. Stress update algorithm for elastoplastic models with non-convex yield surfaces. Int. J. Numerical Methods in Engng, 76: 20292062. Raveendiraraj A. 2009. Coupling of mechanical behaviour and water retention behaviour in unsaturated soils. PhD Thesis. University of Glasgow, UK. Rojas E. 2008. Equivalent stress equation for unsaturated soils, I: Equivalent stress. Int. J. Geomechanics, 8: 285290. Romero E, Gens A & Lloret A. 1999. Water permeability, water retention and microstructure of unsaturated compacted Boom clay. Engng Geol., 54: 117127. Romero E & Jommi C. 2008. An insight into the role of hydraulic history on the volume changes of anisotropic clayey soil. Water Resources Research, 44, W12412. Russell AR & Khalili N. 2006. A unified bounding surface plasticity model for unsaturated soils. Int. J. Numer. Anal. Methods Geomechanics, 30: 181212. Samat S Vaunat J & Gens A. 2008. A thermomechanical framework for modelling the response of unsaturated soils. In: Unsaturated Soils: Advances in Geo-Engineering (Toll DG, Augarde CE, Gallipoli D & Wheeler SJ, eds), CRC Press, pp. 547552. Snchez M, Gens A, Guimares L do N & Olivella S. 2005. A double structure generalized plasticity model for expansive materials. Int. J. Numer. Anal. Methods Geomechanics, 29: 751787. Snchez M, Gens A, Guimaraes L do N, Olivella S. 2008. Implementation algorihm of a generalised plasticity model for swelling clay. Comput. Geotech., 35: 860871. Santagiuliana R & Schrefler BA. 2006. Enhancing the Bolzon-Schrefler-Zienkiewicz constitutive model for partially saturated soil. Transport Porous Media, 65: 130.
Scholts L, Hicher PY, Nicot F, Chareye B & Darve F. 2009. On the capillary stress tensor in wet granular materials. Int. J. Numer. Anal. Methods Geomechanics, 33: 12891313. Sheng D. 2003. Non-convexity of the Barcelona Basic ModelComment on SJ Wheeler, D Gallipoli & M Karstunen (2002; 26:15611571), Int. J. Numer. Anal. Methods Geomechanics, 27: 879881. Sheng D, Sloan SW, Gens A, Smith DW. 2003a. Finite element formulation and algorithms for unsaturated soils. Part I: Theory. Int. J. Numer. Anal. Methods Geomechanics, 27: 745765. Sheng D, Smith DW, Sloan SW, Gens A. 2003b. Finite element formulation and algorithms for unsaturated soils. Part II: Verification and application. Int. J. Numer. Anal. Methods Geomechanics, 27: 767790. Sheng D, Sloan SW & Gens A. 2004. A constitutive model for unsaturated soils: thermomechanical and computational aspects. Computational Mechanics, 33: 453465. Sheng D & Fredlund DG. 2008. Elastoplastic modelling of unsaturated soils: an overview. Keynote Lecture, 12th Int. Conf. Int. Assoc. Computer Methods & Advances in Geomechanics (IACMAG), 16 October, 2008, Goa, India, Indian Institute of Technology, Bombay, pp. 20842105. Sheng D, Fredlund DG & Gens A. 2008a. A new modelling approach for unsaturated soils using independent stress variables. Can. Geotech. J., 45: 511534. Sheng D, Fredlund DG & Gens A. 2008b. Reply to the discussion by Zhang and Lytton on A new modelling approach for unsaturated soils using independent stress variables. Can. Geotech. J., 45: 17881794. Sheng D, Gens A, Fredlund DG & Sloan SW. 2008c. Unsaturated soils: from constitutive modelling to numerical algorithms. Comput. Geotech., 35: 810824. Sheng D, Pedroso DM & Abbo AJ. 2008d. Stress path dependency and non-convexity of unsaturated soil models. Computational Mechanics, 42: 685695. Sheng D, Zhou AN & Fredlund DG. 2010. Shear strength criteria for unsaturated soils, Geotechnical and Geological Engineering, 28, 2010 (in press, published online 7102009, doi: 10.1007/s10706009927-x). Sheng D & Zhou AN. 2010. Coupling hydraulic with mechanical models for unsaturated soils, Can. Geotech. J. (submitted 1 Mar 2010, manuscript No: 1049). Sloan SW, Abbo AJ, Sheng D. 2001. Refined explicit integration of elastoplastic models with automatic error control. Engineering Computations, 18: 12154. Erratum: Engineering Computations, 19: 594594 (2002). Sivakumar V & Wheeler SJ. 2000. Influence of compaction procedure on the mechanical behaviour of an unsaturated compacted clay. Part 1: Wetting and isotropic compression. Gotechnique, 50: 359368. Soowski WT & Gallipoli D. 2009a. Explicit stress integration with error control for the Barcelona Basic Model. Part I: Algorithms formulation. Comput. Geotech., 37: 5967. Soowski WT & Gallipoli D. 2009b. Explicit stress integration with error control for the Barcelona Basic Model. Part II: Algorithms efficiency and accuracy. Comput. Geotech., 37: 6881.
111
Stropeit K, Wheeler SJ & Cui YJ. 2008. An anisotropic elasto-plastic model for unsaturated soils. In Unsaturated Soils: Advances in Geo-Engineering (Toll DG, Augarde CE, Gallipoli D & Wheeler SJ, eds), CRC Press, pp: 625632. Sun DA, Matsuoka H, Yao YP & Ichihara W. 2000. An elastoplastic model for unsaturated soil in threedimensional stresses. Soils & Foundations, 40: 1728. Sun DA, Cui HB, Matsuoka H & Sheng D. 2007a. A three-dimensional elastoplastic model for unsaturated compacted soils with hydraulic hysteresis. Soils & Foundations, 47: 253264. Sun DA, Sheng D, Cui HB & Sloan SW. 2007b. A densitydependent elastoplastic hydro-mechanical model for unsaturated compacted soils Int. J. Numer. Anal. Methods Geomechanics, 31: 12571279. Sun DA, Sheng D & Sloan SW. 2007c. Elastoplastic modelling of hydraulic and stress-strain behaviour of unsaturated soils. Mechanics of Materials, 39: 212221. Sun DA, Sheng D & Xu XF. 2007d. Collapse behaviour of unsaturated compacted soil. Can. Geotech. J., 44: 673686. Sun DA, Sheng D, Xiang L & Sloan SW. 2008. Elastoplastic prediction of hydro-mechanical behaviour of unsaturated soils under undrained conditions. Comput. Geotech., 35: 845852. Tamagnini R & De Gennaro V. 2008. Implicit integration of extended Cam-clay model for unsaturated soils. In Unsaturated Soils: Advances in Geo-Engineering (Toll DG, Augarde CE, Gallipoli D & Wheeler SJ, eds), CRC Press, pp. 713719. Tang AM, Cui YJ & Barnel N. 2008. Compressioninduced suction change in a compacted expansive clay. In Unsaturated Soils: Advances in Geo-Engineering (Toll DG, Augarde CE, Gallipoli D & Wheeler SJ, eds), CRC Press, pp. 369374. Tarantino A & Tombolato S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique, 55: 307317. Tarantino A. 2009. A water retention model for deformable soils. Gotechnique, 59: 751762. Thu TM, Rahardjo H & Leong EC. 2007a. Elastoplastic model for unsaturated soil with incorporation of the soilwater characteristic curve. Can. Geotech. J., 44: 6777. Thu TM, Rahardjo H & Leong EC. 2007b. Critical state behaviour of a compacted silt specimen. Soils & Foundations, 47: 749755. Toll DG. 1990. A framework for unsaturated soil behaviour. Gotechnique, 40: 3144. Toll DG & Ong BH. 2003. Critical state parameters for an unsaturated residual sandy clay. Gotechnique, 53: 93103. Unno T, Kazama M, Uzuoka R & Sento N. 2008. Liquefaction of unsaturated sand considering the pore air pressure and volume compressibility of the soil particle skeleton. Soils & Foundations, 48: 87100. van Genuchten MT. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal, 44: 892898. Vanapalli SK, Fredlund DG, Pufahl DE & Clifton AW. 1996. Model for the prediction of shear strength with respect to soil suction. Can. Geotech. J., 33: 379392. Vanapalli SK, Fredlund DG, & Pufahl DE. 1999. The influence of soil structure and stress history on soil-
water characteristics of a compacted till. Gotechnique, 49: 143159. Vaunat J, Cante JC, Ledesma A, Gens A. 2000. A stress point algorithm for an elastoplastic model in unsaturated soils. Int. J. Plasticity, 16: 121141. Vaunat J, Romero E & Jommi C. 2000. An elastoplastic hydromechanical model for unsaturated soils. In: Experimental Evidence & Theoretical Approaches in Unsaturated Soils (Tarantino A & Mancuso C, eds), Balkema, Rotterdam, pp. 121138. Wheeler SJ. 1996. Inclusion of specific water volume within an elasto-plastic model for unsaturated soil. Can. Geotech. J., 33: 4257. Wheeler SJ. 2006. Constitutive modelling of unsaturated soils. Keynote Lecture presented at 4th International Conference on Unsaturated Soil (unpublished PDF presentation), Arizona, USA, 2006. Wheeler SJ, Gallipoli D & Karstunen M. 2002. Comments on use of the Barcelona Basic Model for unsaturated soils. Int. J. Numer. Anal. Methods Geomechanics, 26: 15611571. Wheeler SJ & Karube D. 1996. Constitutive modelling. In: Unsaturated Soils (Alonso EE & Delage P, eds), Balkema, Rotterdam, vol.3, pp. 13231356. Wheeler SJ, Sharma RS & Buisson MSR. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Gotechnique, 53: 4154. Wheeler SJ & Sivakumar V. 1995. An elasto-plastic critical state framework for unsaturated soil. Gotechnique, 45: 3553. Yang C, Cui YJ, Pereira JM & Huang MS. 2009. A constitutive model for unsaturated cemented soils under cyclic loading. Comput. Geotech., 35: 853859. Zhang HW & Zhou L. 2008. Implicit integration of chemo-plastic constitutive mode for partially saturated soils. Int. J. Numer. Anal. Methods Geomechanics, 32: 17151735. Zhang X & Lytton RL. 2008. Discussion on A new modelling approach for unsaturated soils using independent stress variables. Can. Geotech. J., 45: 17841787. Zhang X & Lytton RL. 2009a. Modified state-surface approach to unsaturated soil behaviour study. Part I: Basic concept. Can. Geotech. J., 46: 536552. Zhang X & Lytton RL. 2009b. Modified state-surface approach to unsaturated soil behaviour study. Part II: General formulation. Can. Geotech. J., 46: 553570. Zhao CG, Liu Y & Gao FP. 2010. Work and energy equations and the principle of generalised effective stress for unsaturated soils. Int. J. Numer. Anal. Methods Geomechanics, In press with DOI: 10.1002/ nag.839. Zhou AN & Sheng D. 2009a. Yield stress, volume change and shear strength behaviour of unsaturated soils: Validation of the SFG model. Can. Geotech. J., 46: 10341045. Zhou AN, Sheng D & Carter JP. 2010. Modelling the effect of density on hydraulic properties of unsaturated soils. Water Resources Research (Submitted 23 March 2010, Article No.: 2010WR009338). Zhou C. 2009. Modelling of suction effect on fabric yielding and kinematic hardening of reconstituted soils. In: Unsaturated SoilsTheoretical & Numerical Advances in Unsaturated Soil Mechanics (Buzzi O, Fityus SG & Sheng D, eds), CRC Press, pp. 629634.
112
Unsaturated soils: Compacted versus reconstituted states

Alessandro Tarantino
Department of Civil Engineering, University of Strathclyde, Glasgow, UK Formerly Dipartimento di Ingegneria Meccanica e Strutturale, Universit degli Studi di Trento, Italy
ABSTRACT: The paper presents a comparison between compacted and reconstituted soils in terms of microstructure, and hydraulic and mechanical response. It is commonly assumed that reconstituted and compacted soils exhibit a fundamentally different behaviour due to different microstructures. However, the variety of pore size distributions observed in both compacted and reconstituted/natural soils suggests that the boundary between compacted and reconstituted states is more blurred. In the paper, an attempt is made to recognize similarities and differences between compacted and reconstitutes states based on a number of recent experimental studies where the microstructure and the hydraulic and mechanical behaviour of unsaturated soils in compacted and reconstituted states have been investigated. This exercise will also offer the opportunity to gain a better insight into the microstructure of compacted soils. 1 INTRODUCTION mono-modal pore size distribution may evolve to a bi-modal pore size distribution upon drying (Gens et al. 1995, Simms & Yanful 2001). Conversely, soils compacted on the dry-side of optimum and having an initial bi-modal pore size distribution may evolve to a mono-modal pore size distribution upon wetting (Monroy et al. 2010). Similarly, initial bi-modal PSDs in soils compacted on the dry side of optimum may evolve to a mono-modal PSD due to mechanical effects as observed in volumetric collapse upon wetting under loading (Barrera 2002, Gomez et al. 2009) or loading at constant suction (Barrera 2002). On the other hand, reconstituted normally consolidated soils may exhibit a bi-modal pore size distribution at low stresses (Griffiths & Joshi 1989) and a bi-modal pore-size distribution may appear in saturated natural clays in an overconsolidated state (Ninjgarav et al. 2007). It is commonly assumed that reconstituted and compacted soils exhibit a fundamentally different behaviour due to different microstructures. However, if one looks at the variety of pore size distributions in both compacted and reconstituted/natural soils, the boundary between these states is perhaps more blurred. It may therefore be worth comparing the microstructure as well as the macroscopic behaviour of soils in reconstituted and compacted states to identify similarities and differences. A number of experimental studies of hydraulic and mechanical behaviour of soils in both compacted and reconstituted states have been presented in the literature (Marinho 1994; Jotisinkasa 2005; Monroy 2006; Barrera 2002, Boso 2005, and Buenfil 2007; Tarantino &
Geotechnical research on unsaturated soils has largely focused on compacted double-porosity geomaterials. When an air-dried soil is moistened, aggregates of clay/silt particles are formed and are generally not destroyed as the soil is statically or dynamically compacted. This results in at least two populations of pores, intra-aggregate pores (within the aggregates) and inter-aggregate pores (between the aggregates) as evidenced by Mercury Intrusion Porosimetry (MIP) tests. In geotechnical engineering, aggregated double-porosity soils generally result from compaction on the dry side of optimum (Ahmed et al. 1974, Delage et al. 1996). Aggregation also occurs in agricultural soils due to abiotic (freeze-thaw and drying-wetting cycles) and biotic factors (fecal pellets and casts by soil fauna) (Carter 2004). On the other hand, silty and clayey soils from natural deposits often exhibit a mono-modal Pore Size Distribution (PSD) (e.g. Tanaka et al. 2003), at least in the pore size range investigated by the MIP. The microfabric of geomaterials deposited in water and later exposed either by lowering of the water level or by uplifting of the land can be replicated in the laboratory by reconstituting soils from slurry. The double-porosity microstructure of soils compacted on the dry side of optimum and the single-porosity microstructure of soils reconstituted from slurry represent two extremes. However, the distinction between single- and double porosity soils is not clear-cut. Soils compacted on the wet side of optimum and having an initial
113
Tombolato 2005; Tarantino 2009; Koliji 2008) and these will serve as a basis to discuss similarities and differences between compacted and reconstituted states. This comparison will also give the opportunity to gain a better insight into the microstructure of compacted soils. There is indeed another motivation behind the comparison of compacted and reconstituted states. Many of the concepts developed for unsaturated soils have been established for compacted doubleporosity soils. The first generation of unsaturated soil constitutive models formulated in terms of net stress and suction (Alonso et al. 1990, Wheeler & Sivakumar 1995, Cui & Delage 1996) and the second generation of constitutive models that have included the degree of saturation in generalised stress variables (Wheeler et al. 2003, Gallipoli et al. 2003, Tamagnini 2004), have been formulated for and validated against data on double-porosity compacted soils. Similarly, mechanical effects on water retention behaviour have mainly been investigated for compacted double-porosity soils (Vanapalli et al. 1997, Romero et al. 1999, Romero & Vaunat 2000, Gallipoli et al. 2003). It may therefore be interesting to investigate the aspects of compacted soil behaviour that can be extrapolated to reconstituted/natural unsaturated soils. 2 MICROSTRUCTURE OF COMPACTED SOILS
Figure 1. Aggregates forming after spraying Speshwite Kaolin powder with water to different water contents w as observed using an optical microscope (ticks are 1 mm spaced) (Tarantino & De Col 2009).
To prepare compacted samples, air-dried or ovendried soil powder is sprayed with water to target water contents. Upon spraying, clay particles tend to form aggregates, the size of the aggregates depending on the amount of water sprayed. This is clearly illustrated in Figure 1 which shows aggregates formed after spraying air-dried powder of Speswhite Kaolin to different water contents and then hand-mixing the moistened powder. At high water contents (w 0.30) very large aggregates formed and, when sieving the moistened powder on the sieve having 1 mm aperture size, about 2030% of the material was retained on the sieve. The sieved material still had large aggregates as shown in Figure 1a. As water content was decreased to w 0.24, the size of the aggregates forming upon spraying decreased as shown in Figure 1b. At water contents lower than 0.220.24, almost all the moistened powder passed through the 1 mm aperture size sieve. Eventually, at very low water content (w 0.10), there was little aggregation as shown in Figure 1c. It is interesting to note that aggregation at very low water contents (w 0.10), as observed by visual inspection, does not seem to significantly differ from that occurring
to the kaolin powder under hygroscopic water content wh as shown in Figure 1d. The hygroscopic water content is here defined as the water content of the soil in equilibrium with laboratory air at a relative humidity of RH 0.400.5. At this relative humidity, water is likely to be adsorbed on clay particles. The thickness of the adsorbed water film surrounding the clay particle is related to the hygroscopic gravimetric water content wh and the specific surface Sa by the following relationship:
wh Sa w
(1)
where w is the density of water. On the other hand, the thickness can be related to the total suction (chemical potential) of adsorbed water through the following relationship according to Iwamatsu & Horii (1996):
3
Avsl 6
(2)
where is the total suction and Avsl is the Hamaker constant, which can be assumed to be equal to 6 1020 J according to Tuller & Or (2005). This relationship only takes into account van der Waals forces on planar surfaces and neglects contribution of capillary condensation, which occurs when adsorbed films on opposing surfaces becomes unstable, grow without bounds and coalesce (Christenson 1994, Tuller et al. 1999). Total suction can be inferred from the measurement of relative humidity of the air in equilibrium
114
with the adsorbed film through the psychrometric law (Tarantino 2010):
pv RT RT ln ln RH vw pv 0 T vw
(3)
where T is the absolute temperature, R is the universal gas constant, vw is the molar volume of liquid water, pv is the pressure of the vapour in equilibrium with the liquid, pv0 is the saturated vapour pressure at the same temperature, and RH is the relative humidity. For a relative humidity of RH 0.5, the thickness of the adsorbed film calculated using Eq. (2) and (3) is 3.2 corresponding to the size of a water molecule. Assuming Sa 14 m2/g for Speswhite Kaolin according to the technical specification of the supplier, hygroscopic water content wh 0.0046 is obtained using Eq. (1), which is in excellent agreement with the measured value of wh 0.005. Banin & Amiel (1970) and Dirksen & Dasberg (1993) also found that hygroscopic water contents in equilibrium with laboratory relative humidity are associated with a monomolecular layer of adsorbed water. Figure 1d therefore shows the level of aggregation when a monolayer of water molecules envelopes the clay particles. Increasing the water content to w 0.10 (Figure 1c) does not apparently change the level of aggregation. As discussed in the next sections, it is speculated that adding water from hygroscopic water content to w 0.10 does not change the level of aggregation and water only fills the unsaturated intra-aggregate pore space. When the water content is increased beyond w 0.24 (Figure 1a,b), the added water directly control the size of the aggregates, which are now saturated with water (this point will also be discussed in the next sections). As the compaction water content increases, the size of the aggregates increases. Similar observations have been made by Jotisinkasa (2005) by inspecting thin sections of compacted silty clay. As the aggregates are placed in the mould for compaction, the inter-aggregates voids are filled with air and water only occupies the intra-aggregate pore space. Pore-water pressure in the aggregates is negative and controlled by the curvature of the menisci at the aggregate boundary. As the soil is loaded (dynamically or statically), the saturated aggregates are subjected to undrained loading, which causes the pore-water pressure in the aggregates to increase (as a first approximation, the saturated aggregates can be assumed to be normally consolidated). If the pore-water pressure remains at values lower than zero (i.e. lower than the pressure of the surrounding air), the aggregates may undergo some distortion but their volume will
remain constant. However, if the pore-water pressure tends to increase to values greater than zero, water will be expelled from the aggregates and will invade the inter-aggregate pore-space. This qualitative representation of the compaction process will serve as basis to discuss the effect of water content, compactive effort and compaction method on microfabric as observed in Mercury Intrusion Porosimetry (MIP) and, to a lesser extent, in the Scanning Electron Microscope (SEM). The MIP investigation is here prioritised as it provides a quantitative assessment of soil microfabric. Since the pioneering work of Diamond (1970, 1971), MIP has been widely used to assess hydraulic and mechanical effects on microstructure in natural and compacted soils (a review of this technique can be found in Romero & Simms 2008). The technique used for soil dehydration before MIP testing is of great importance to preserve soil microfabric. Standard air-drying has proven to change significantly the pore-size distribution (Ahmed et al. 1974) and the freeze-drying technique (Delage & Pellerin 1984) has therefore become a standard in soil specimen preparation for MIP. This technique consists in rapid freezing the soil in liquid nitrogen or intermediate cooling liquid and subsequent sublimation under vacuum so as to eliminate the formation of air-water menisci that will cause shrinkage on drying. Cooling can be achieved using intermediate cooling liquid like freon (Delage et al. 1982; Griffiths & Joshi 1989) or isopentan (Tarantino & De Col 2008). However, cooling of nitrogen down to its freezing point (210C) by applying a vacuum appears to be the simplest and most efficient method (Delage et al. 2006). In this paper, only MIP results on freezedried samples will be discussed. 2.1 Effect of compaction water content
Ahmed et al. (1974) investigated the effect of compaction water content on the pore-size distribution of Grundite silty clay (index properties summarised in Table 1). Two samples were compacted on the dry side and wet side of optimum at nearly the same void ratio, e 0.80 and e 0.79 respectively. A third sample was compacted at the optimum water content and a lower void ratio was obtained (e 0.73). The results of MIP tests are presented in Figure 2 in terms of cumulative and frequency distribution (original data were processed to derive the frequency distribution shown in Figure 2b). Data are presented in terms of intrusion volume ratio eMIP defined as the intruded volume Vintruded per unit volume of solids, Vs: eMIP Vintruded Vs (4)
115
Table 1. Index properties of soils discussed in the paper. Soil Grundite clay (Ahmed et al. 1974) Jossigny silt (Delage et al. 1996) Silt/kaolin (90/10) (Garcia-Bengochea et al. 1979) MX80 Bentonite (Tang & Cui 2005) Boom Clay (Romero et al. 1999) Kaolinite (Shridaran et al. 1971) Speswhite kaolin (Tarantino & Tombolato 2005) Kaolinite, montmorollonite (Griffiths & Joshi 1989) Silty sand (Huang et al. 1994) BCN silt (Barrera 2002, Boso 2005, Buenfil 2007) Sandy loam (Cuisinier & Laloui 2004) Clayey silt (Jotisinkasa et al. 2009) London clay (Monroy et al. 2010) Glacial till (Simms & Yanful 2001) Champlain clay (Delage & Lefebvre 1984) Kaolinite clay (Hattab et al. 2010) Clay Silt (%) (%) 65 34 8 60 50 72 80 35 56 90 40 47 28 20 3 wl Sand (%) wp (%) (%) 10 2 52 19 N/A 46 29 26 32 22 37 N/A 520 56 62 64
0.8 0.6 eMIP 0.4 0.2 0
e=0.80
e=0.79 e=0.73
ew
(a)
Optimum (w=0.214, e= 0.73) Wet of optimum (w=0.261, e= 0.79) Dry of optimum (w=0.156, e= 0.80)
eMIP/(logd)
(b)
0 0.01 0.1 1 10 Entrance pore size, d: m 100 1000
N/A N/A 10 37 37 45 53 18
30.4 17 16 18 18 29 18 25 20
100 22 32 30 28 83 27 60 40
Figure 2. Pore size distribution of Grundite clay compacted at various water contents (a) Cumulative intrusion volume ratio. (b) Intrusion volume ratio frequency (after Ahmed et al. 1974).
N/A N/A N/A 27 58 8 80 59 55 40 14 20 41 18 2 18
In this way, the value eMIP at the end of the intrusion stage can directly be compared with the void ratio e determined in a traditional way by measuring the volume and water content of the sample from which the specimen for MIP was taken. As shown in Figure 2a, there is a large volume of pores that were not intruded. In this particular case, the maximum pressure applied to mercury was relatively small and a significant amount of intra-aggregate porosity was not intruded. More in general, when
the pressure applied to mercury attains the classical 200 MPa, the pores that are not intruded are generally the pores within elementary particle arrangements (inter-particle pores) or intra-particle pores. A schematic representation of microfabric of compacted soils is shown in Figure 3 according to Collins & McGown (1974) and Alonso et al. (1987). The basic unit in natural and compacted clayey soils is not the single clay platelet but domains constituted by groups of clay platelets. Intra-domain pores are referred to as intra-elemental pores in Figure 3. Domains then group together to form aggregates and pores within the aggregates are referred to as intra-aggregate pores. Finally, pores exist between the aggregates and are referred to as inter-aggregate pores. In addition to these classes of pores, interlayer pores exist between the elementary layers inside smectite particles (Delage et al. 2006). MIP can only detect inter-aggregate and intra-aggregate pores, rarely intra-domain pores, never intra-particle pores. Figure 2 shows that samples compacted on the dry side of optimum exhibit a bi-modal pore-size distribution whereas samples compacted on the wet side of optimum and at optimum water content essentially exhibit a mono-modal pore-size distribution. Similar observations were made by Delage et al. (1996) and Prapaharan et al. (1991) for two other clayey soils. For the sample compacted on the dry of optimum, the larger modal pore size (30 m) represents inter-aggregates pores whereas the smaller
116
Figure 3. Inter-aggregate, intra-aggregate and intradomain (intra-element) pores in clayey soils. (a) compacted on the wet side of optimum. (b) compacted on the dry side of optimum. (d) elementary particle arrangement or domain (after Alonso et al. 1987).
Figure 4. SEM micrographs of Jossigny silt compacted (a) on the dry side of optimum and (b) on the wet side of optimum (Cui 1993).
modal size (0.06 m) represents the intra-aggregate pores (Figure 2b). On the other hand, the sample compacted on the wet side of optimum exhibits a single mode at around 0.3 m representing intra-aggregate pores. The different fabric of soils compacted on the dry and wet side of optimum is shown in Figure 4 for the Jossigny silt (Cui 1993). On the dry side of optimum aggregates are clearly visible as well as inter-aggregate pores being in the order of tenths of microns. On the other hand, there are no detectable aggregates in the soil compacted on the wet side of optimum and the space between silt particles are filled with the clay fraction that form a relatively uniform matrix. It is then interesting to visualise the pore space occupied by pore-water. If one assumes that water saturates the smallest pores, the pore entrance size separating the pores filled with water from those filled with air can be can be obtained by subtracting the water ratio ew to the void ratio e as shown in Figure 2a (water ratio ew is the volume of water per volume of solids, ew wGs). If the pore-water is reported on the frequency pore size distribution, it clearly appears that water has saturated the aggregates and that the inter-aggregates pores are filled with air. It is interesting to observe that as the
compaction water content increases, the modal size of the intra-aggregate pores and their frequency also increases. For example, the modal intra-aggregate pore size increases from 0.05 to 0.08 to 0.3 m when compaction water content w is increased from 0.156 to 0.214 to 0.261. As shown later in this section, the increase in frequency of the modal intra-aggregate pore size can be explained by considering that the size of the aggregates increases as the compaction water content increases (as shown in Figure 1). The mono-modal pore size distribution of soils compacted on the wet side of optimum is characteristic of fine-grained soils (as shown in Table 1, Jossigny silt and Grundite clay have a significant clay fraction). This may not be the case of soils having a reduced clay fraction. Garcia-Bengochea et al. (1979) tested a mixture consisting of 90% silt and 10% clay and observed a bi-modal pore size distribution even in samples compacted on the wet side of optimum (Figure 5). In this case, aggregations of clay particles are likely to coat the silt particles or to form bridges between them. The soil skeleton is made of silt grains, which do not expand and fill the large
117
pores as the water content is increased. The large pores observed in the samples compacted on the wet side of optimum and at the optimum water content are perhaps inter-grain pores rather than inter-aggregate pores. Bi-modal pore size distributions of soils compacted on the wet side of optimum were also observed by Simms and Yanful (2001) testing a glacial till, which again included a relatively small clay fraction (Table 1). Inspection of Figure 5 reveals that pore-water fills the aggregates in the sample compacted on the dry side of optimum (similarly to the Grundite clay in Figure 2). In contrast, water also fills the inter-grains pores when the soil is compacted at optimum water content and on the wet side of optimum. The effect of compaction water content on the PSD of clays compacted on the dry side of optimum was investigated in detail by Tarantino & De Col (2008). Samples with water content ranging between 0.086 and 0.299 were statically compacted to 1200 kPa vertical stress and subsequently tested in the porosimeter (encircled samples in Figure 6). The frequency distribution is shown in Figure 7. Specimens with the higher water content (w 0.299, 0.259, and 0.215) show two modal pore sizes at about 0.10.2 and 0.60.7 m. As water content increases, the modal sizes remain unchanged but the intra-aggregate and inter-aggregate porosity redistribute. In particular, inter-aggregate porosity decreases and intra-aggregate porosity
0.6
e=0.51 e=0.48 e=0.48
1.44 1.36
v=1200 kPa v= 900 kPa v= 600 kPa v= 300 kPa
w=0.299 w=0.259 w=0.215 w=0.141 Sr=0.9
Dry density, d : Mg/m3
1.28 1.2 1.12 1.04 0.96 0
w=0.126 w=0.086
Sr=0.7
Sr=0.2 0.08
Sr=0.3 0.32
Sr=0.5 0.4
0.16 0.24 Water content, w
Figure 6. Compaction plane for Speswhite kaolin (after Tarantino and De Col 2008).
Intrusion volume frequency, eMIP/(logd) 2.5 w=0.299 w=0.259 w=0.215 w=0.141 w=0.086
1.5
0.5
0 0.001 0.01 0.1 1 Entrance pore size, d: m 10 100
Figure 7. Pore size distribution Speswhite kaolin compacted at various water contents on the dry side of optimum (after Tarantino and De Col 2008).
0.4
eMIP
ew 0.2
(a)
0 Optimum (w=0.119, e=0.48) Wet of optimum (w=0.138, e=0.45) Dry of optimum (w=0.075, e=0.51)
eMIP/(logd)
0.1
(b)
0 0.01 0.1 1 10 Entrance pore size, d: m 100 1000
Figure 5. Pore size distribution of Silt/Kaolin (90/10) mixture compacted at various water contents (a) Cumulative intrusion volume. (b) Intrusion volume frequency (after Garcia-Bengochea et al. 1979).
progressively increases as water content increases. Such an evolution of the PSD with water content is very similar to the one observed in Grundite clay (Figure 2). At much lower compaction water contents (w 0.141 and 0.086), the intra-aggregate modal size seems to disappear with a significant increase in the frequency associated with the interaggregate modal pore size of 0.7 m. The evolution of the PSD with compaction water content presented in Figure 7 is directly related to the size of the aggregates, which increases as water content increases (Figure 1). This effect can be explained by considering a very simple microstructural model. Let us assume that aggregates are formed by domains of clay platelets, represented by circles in Figure 8, and that saturated aggregates are formed when the dry powder is sprayed with water. If the compaction water content is low, aggregates need to be small for the aggregates to be saturated (Figure 8a). As water content is increased, larger aggregates may form still remaining fully saturated (Figure 8b, c).
118
Intra-aggregate pore (Inter-domain pore) Domain
Inter-aggregate pore Aggregate
2.2 Effect of compactive effort The compactive effort controls the dry density of the compacted soils and, hence, its pore size distribution. Delage (2009) reinterpreted PSD data from Shridaran et al. (1971) relative to a kaolinite compacted to different void ratios. He showed that soil compression occurs at the expenses of the macropores, the intra-aggregate porosity remaining essentially unaffected by the overall change in void ratio (Figure 9). Similarly Lloret et al. (2003) showed that compaction of bentonite only reduced inter-aggregate porosity whereas intra-aggregate pores remained unaffected. Slightly different results have been obtained by Romero et al. (1999) on Boom clay compacted on the dry side of optimum as shown in Figure 10. Similarly to Shridaran et al. (1971), the increase in dry density occurs at the expense of interaggregate macropores (inter-aggregate modal pore-size decreases from 2000 nm to 600 nm when dry density is increased from 13.7 to 16.7 kN/m3). However, the intra-aggregate modal pore size also decreases (from 60 to 20 nm) suggesting that compression also affects the arrangement of the aggregates. It is interesting to observe that pore-water (shaded areas in Figure 10b) completely fills the intra-aggregate pore space and seems to slightly invade the macrospores. It can be inferred that compression of the saturated aggregates have increased pore-water pressure within the aggregate to values greater than zero causing water to be expelled form the aggregates with the consequent reduction of the aggregate size. The fact that water only slightly invaded the macropores implies that suction was not affected by the change in the macropores (inter-aggregate pores) as confirmed by the vertical contours of equal suction in Figure 11 (crosses in Figure 10 at
c 15
Density fucntion
12 8
7 3 0
a, low w b, medium w c, high w d
Intraaggregate
Interaggregate
Pore size
Figure 8. Schematic representation of the relationship between compaction water content and volume of intra- and inter-aggregate pores (black regions represent water).
The density function of such a schematic packing is presented in Figure 8d. As the compaction water content increases, the number of intraaggregate pores (proportional to the amount of compaction water content if the aggregates are assumed to be saturated) increases from 8 to 12 to 15 whereas the number of inter-aggregate pores decreases from 7 to 3 to 0. This evolution of the PSD of this elementary packing is essentially the one observed in Figure 2 and Figure 7. According to the mechanism presented in Figure 8, the decrease in the frequency associated with the intra-aggregate pore size (0.2 m) in the samples compacted at the lower water contents should not be interpreted as the disappearance of aggregates. Indeed, it is associated with a reduction of the aggregate size due to a deficiency in compaction water content. The mechanism illustrated in Figure 8 would suggest that microfabric of compacted soils in terms of size and number of aggregates is essentially controlled by the compaction water content. To a lesser extent, size and number of aggregates may also be controlled by the procedure used to moisten the soil powder. Experimentalists know well how the process of compaction is operatordependent.
Figure 9. Cumulative pore size distributions of Kaolinite compacted to various densities at same water content (Delage 2009 after Shridaran et al. 1971).
119
w 15% represents the as-compacted states investigated). It can be speculated that further compression would have caused more water to be expelled from the aggregates and the suction would eventually be affected by compression of the macropores. The contours of equal suction would have therefore been expected to deviate on the left side as shown by Li (1995, quoted by Delage & Graham 1995). The effect of compaction effort on pore size distribution is also shown in Figure 12 for compacted Speswhite Kaolin (specimens tested in the MIP are dashed-encircled in Figure 6). At relatively high water content (w 0.259), compression occurs again at the expenses of inter-aggregate pores (modal size 0.7 m) with a very slight compression of the intra-aggregate pores). It is worth observing that pore-water (grey shaded area) fills the intra-aggregate pores and partly invades the macropores. It is therefore not surprising that irreversible compression of the macropores associated with the increase in vertical stress from 600 to 1200 kPa causes a change in suction in the post-compacted samples. This is illustrated in Figure 13 where the relationship between suction and degree of saturation during the compaction process of samples having different water contents is presented (Tarantino & De Col 2008). Samples were statically compacted
Figure 10. Pore size distribution of Boom Clay compacted at w 15% at different dry densities (a) Cumulative intrusion volume. (b) Intrusion volume frequency (Romero et al. 1999).
Intrusion volume frequency, eMIP /(log d )
2 w=0.259 1.6
v=600 kPa v=1200 kPa
1.2
0.8
(a)
0.4
0.01
0.1 1 Entrance pore size, d: m

v=300 kPa v=900 kPa v=1200 kPa
10
100
Intrusion volume frequency, eMIP /(log d )
w=0.126
1.6
1.2
0.8
(b)
0.4
0.01
0.1 1 Entrance pore size, d: m
10
100
Figure 11. Compaction plane and contours of isosuction (post-compaction) of Boom clay (Romero et al. 1999).
Figure 12. Pore size distribution Speswhite kaolin compacted to different vertical stresses at constant water content (Tarantino, in prep.,a)
120
Figure 13. Degree of saturationsuction paths at different compaction water contents. Dotted lines join postcompaction suction (Tarantino and De Col 2008).
at constant water content under one-dimensional conditions by applying loading-unloading cycles to 300, 600, 900, and 1200 kPa. The dashed lines join the unloaded states often referred to as-compacted states. For the sample having water content w 0.259, the suction after unloading from 600 kPa (open circle) is different from the suction recorded after unloading from 1200 kPa (solid circle), confirming that an irreversible change in the interaggregate porosity causes an irreversible change in suction when pore-water occupies the inter-aggregate pore-space. The fact that post-compaction suction increases with the degree of saturation, which is not intuitive at first sight, has been demonstrated to be due to the effect of void ratio on the water retention curve (Tarantino & De Col 2008). The evolution of PSD of a sample compacted to vertical stresses of 300, 900, and 1200 kPa at a significantly lower water content (w 0.126) is shown in Figure 12b. It is again observed that volume reduction occurs at the expenses of large inter-aggregate macro-pores. However, two aspects that have not been observed in any of the PSDs previously shown are worth noticing. The intra-aggregate pore space now appears to be unsaturated (the amount of water present in the micropores is represented by the grey shaded area). It will be shown later on in the paper that the partial saturation of the aggregates can be associated with the partial saturation of the same soil in a reconstituted state at the same water content. This demonstrates that the common assumption that aggregates in compacted soils are saturated, although often correct, does not always hold.
Since pore-water has withdrawn within the aggregates, irreversible changes in the macropores (inter-aggregates pores) should not cause any change in suction. This is indeed the case for the sample compacted at the water content of w 0.126 as shown by the vertical contours of equal postcompaction suction in Fig. 10 of the paper by Tarantino & De Col (2008). This is also the case of the sample compacted at the water content of w 0.215 in Figure 13 as the dashed line joining the post-compaction states appears to be vertical. Nonetheless, for this sample, suction decreases during loading presumably because of the elastic water-undrained compression of the aggregates. The second interesting aspect is the modal size appearing in the sample at low water content in the range 0.010.02 m. In particular, this modal size shifts from 0.017 m at v 300 kPa to 0.01 m at v 1200 kPa. These pores may be inter-particle (intra-domain) pores whereas the modal size in the range 0.10.2 mm characterising the intraaggregate size (Figure 12a) may be associated to inter-domain pores (this point will be discussed further on). 2.3 Effects of compaction method A question that is worth addressing is the difference between static, dynamic and kneading compaction. Static compaction is often used in the laboratory to control the stress history of samples subsequently subjected to mechanical testing. However, kneading and dynamic compaction are more representative of field compaction. Ahmed et al. (1974) presented PSDs of Grundite clay compacted on the dry and wet side of optimum and at optimum water content using dynamic, kneading and static compaction. Figure 14 shows the PSDs of samples compacted on the dry side of optimum at the same dry density and using different compaction methods. No significant differences were observed and similar results were obtained for samples compacted on the wet side of optimum and at the optimum water content. Prapaharan et al. (1991) compacted medium plastic clay (wl 38%, wp 16%) using dynamic (impact) and kneading compaction. Some minor differences where only observed for the samples compacted on the dry side of optimum in the range of inter-aggregate porosity. The compaction method does not appear to have any effect on the PSD of compacted soils. As suggested in Figure 8, the size and number of the aggregates would mainly depend on the compaction water content. Data shown in Figure 14 would suggest that the process of compaction does not break the aggregates and, hence, different compaction methods do not alter the microstructure.
121
The comparison is shown in Figure 15 and it can be observed that differences, at least in terms of PSD, essentially occur in the range of inter-aggregate porosity whereas the intra-aggregate porosity is similar as also inferred from the similar non-constricted porosity as detected by mercury extrusion (Jommi & Sciotti 2003). This is remarkable as the soil in the field is compacted in peds, which are likely to retain the same structure of the natural soil. On the other hand, the soils compacted in the laboratory was crushed, sieved, and wetted again.
Figure 14. Pore size distributions for compaction on dry side by different compaction methods (Ahmed et al. 1974).
MICROSTRUCTURE OF RECONSTITUTED SOILS
There are few data on PSD of reconstituted soils in unsaturated state. Speswhite Kaolin reconstituted from slurry and subsequently air-dried was tested at the University of Trento. The drying curves in terms of void ratio e, water ratio ew, and degree of saturation Sr are reported in Figure 16. For suctions lower than around 800 kPa, the slurry remained saturated and the soil followed the saturated normal consolidation line. As the soil desaturated, the void ratio rapidly attained a nearly constant value whereas the water ratio and
1.6 Void ratio, eand watert ratio ew
e, matric suction (tensiometer) e, total suction (psychrometer) ew, matric suction (tensiometer) ew, total suction (psychrometer)
1.2
MIP
Figure 15. Log-log representation of cumulative intrusion curve of field and laboratory compacted samples (Jommi & Sciotti 2003).
0.8
0.4
Normal consolidation line
The negligible effect of the compaction method on the soil microstructure seems to be corroborated by the results obtained by Wheeler & Sivakumar (2000) who tested kaolin samples statically and dynamically compacted to the same dry density and found no appreciable differences in terms of stress-strain response in the triaxial apparatus. It should be noted that the comparison presented in Figure 14 concerns samples compacted in the laboratory. Differences arise when comparing PSD of samples compacted in the field and in the laboratory. Jommi & Sciotti (2003) compared the PSD of a soil taken from a compacted embankment and the one obtained in the laboratory after the borrow soil was initially dried, passed through the sieve No. 200, mixed to the target water content with tap water, cured and dynamically compacted to similar dry density as in the field.
Degree of saturation, Sr
0.8
0.6
0.4
0.2
Matric suction (tensiometer) Total suction (psychrometer)
0 10 100 1000 10000 100000 Matric suction, sm and total suction, st (kPa)
Figure 16. Free-drying behaviour of Speshwite kaolin prepared from slurry at 1.5 wl (Tarantino, in prep., b).
122
the degree of saturation kept decreasing. Two saturated and two unsaturated samples (encircled in Figure 16) were tested using MIP as shown in Figure 17. The PSD distribution is mono-modal with modal size around 0.2 m. This modal size is essentially the same as the intra-aggregate modal pore size of the same soil in compacted state (Figure 7). It is interesting to notice that the modal pore size decrease when the water ratio decreases from 0.934 to 0.901 to 0.720 and then shifts back to a higher value at the lower water ratio (ew 0.495). This is associated to the slight increase in void ratio as the soil desaturates (Figure 16). Figure 18 shows the PSD of two reconstituted samples of Bioley silt (Koliji et al. 2010). One sample (PR2) was loaded under saturated conditions and the other one (PR1) was dried to a suction of 500 kPa under a total vertical stress of 500 kPa. Again, the pore size distribution remains mono-modal under the combined effect of loading and suction. It may be instructive to compare the PSD of these unsaturated reconstituted soils with the PSD of other reconstituted saturated clays. Delage & Lefebvre (1984) measured the PSD of a remoulded saturated Champlain clay (this clay shows similar
5 ew=0.934 4 ew=0.901 ew=0.720 ew=0.495 3 (a) 2
100 PR2 (s=0 kPa, =125 kPa) Pore size density function 80 PR1 (s=500 kPa, =500 kPa)
60
40
20
0 0.001 0.01 0.1 1 10 100 Entrance pore radius (m)
Figure 18. PSD of remoulded Bioley silt at zero suction (PR2) and and after imposing a suction of 500 kPa and a vertical stress of 500 kPa (PR1) (after Koliji et al. 2010).
0 0.1 Entrance pore size ( m) 1
Cumulative intrusion volume ratio, eMIP
ew=0.934 0.8 ew=0.901 ew=0.720 ew=0.495 (b) 0.4
0.6
0.2
0 0.01 0.1 1 10 100 Entrance pore size (m)
Figure 17. PSDs of Speshwite kaolin prepared from slurry at 1.5 wl and desaturated by air-drying (Tarantino, in prep., b).
grain size distribution and consistency limits as indicated in Table 1). They observed that the PSD is mono-modal even if the microstructure consists of aggregates dividing porosity into inter- and intraaggregate pores as shown by scanning electron microscopy. Similarly, Hattab et al. (2010) showed that reconstituted kaolinite essentially exhibits a mono-modal pore-size distribution (Figure 19) even if aggregates can clearly be distinguished using the scanning electron microscope. It is interesting to note that the modal size of 0.182 m of the reconstituted kaolinite consolidated to 120 kPa vertical stress in Figure 19 is classified as interaggregate porosity by Hattab et al. (2010). This modal size is very close to the modal size of reconstituted Speswhite Kaolin as shown in Figure 17, which is in turn very similar to the intra-aggregate size of compacted Speswhite Kaolin as shown in Figure 7. In other words, the same modal size is classified as inter-aggregate in the reconstituted soil and as intra-aggregate in the compacted soil. This leads to the conclusions that aggregates in compacted clays are made of sub-aggregates (elementary particle arrangements according), which were referred to as aggregates in the reconstituted soil investigated by Hattab et al. (2010). A question that might be asked is why the pore size distribution of reconstituted soil appears to be mono-modal despite the existence of aggregates whereas the existence of (macro) aggregates in compacted soils is clearly evidenced by a bi-modal pore size distribution. A possible explanation can be provided by inspecting the PSD measured by Griffiths & Joshi (1989) on KaoliniteMontmorillonite prepared at the liquid limit and subsequently loaded to different vertical stresses. At the liquid limit, two inflection points can be clearly distinguished, which should correspond
Intrusion volume frequency, eMIP/(logd)
123
1 Density function 0.8 0.6 0.4 0.2 0 Density function 0.8 0.6 0.4 0.2 0 Density function 0.8 0.6 0.4 0.2 0 Density function 0.8 0.6 0.4 0.2 0 Density function 0.8 0.6 0.4 0.2 0 1 10 100 1000 10000 100000 1000000
s=0 kPa, Sr=0.91 s=40 kPa, Sr=0.82 s=150 kPa, Sr=0.74 s=470 kPa, Sr=0.70 s=996 kPa, Sr=0.632
Free-porosity
(a)
Water
(b)
Figure 19. Pore size distribution of reconstituted kaolinite consolidated to 120 kPa vertical stress (Hattab et al. 2010).
(c)
(d)
(e)
Figure 20. Cumulative pore size distributions of Kaolinite-Montmorillonite prepared at wl and loaded under saturated conditions (after Griffiths & Joshi 1989).
Entrance pore diameter (nm)
to intra- and inter-aggregate porosity (Figure 20). However, the inter-aggregate porosity tends to collapse as the soil is loaded. As intra-aggregate and inter-aggregate porosity are of the same order of magnitude, it becomes difficult to distinguish these two classes of pores and the PSD appears to have a single inflection. 4 4.1 EFFECTS OF WATER CONTENT CHANGE ON MICROSTRUCTURE Changes in microstructure upon saturation
Figure 21. Evolution of pore size distributions of compacted London clay upon wetting (after Monroy 2006, Monroy et al. 2010).
The pore size distribution generated by the compaction process does not remain unchanged as the soil is subjected to drying or wetting. Monroy et al. (2010) investigated the evolution of the PSD of London clay compacted on the dry side of optimum and then progressively wetted until saturation in a suction-controlled osmotic oedometer. The evolution of the PSD is shown in Figure 21 and shows that intra-aggregate porosity
progressively increases in terms of both modal size and frequency until the inter-aggregate porosity is almost entirely erased (Figure 21e). It is interesting to note that there are similarities between the evolution of the PSDs of compacted London clay as it is progressively wetted and the evolution of PSDs of Speshwite Kaolin compacted at different water contents (Figure 7). In both cases, inter-aggregate porosity tends to disappear as water content increases. The only difference lies in the swelling of London clay aggregates probably associated with the smectite fraction of the clay. Figure 21 also shows that the aggregated fabric is not completely erased by hydration and this was confirmed by comparing ESEM photographs of compacted and subsequently hydrated samples and of reconstituted saturated samples (Monroy 2006). Similar conclusions were drawn by Thom et al. (2007) on the basis of PSDs determined before and after saturation although some concerns arise from
124
their data as porosimeter data shows a decrease in mercury intruded pore volume upon saturation whereas an overall swelling was observed at macroscopic level upon saturation. 4.2 Changes in microstructure upon drying
Cuisinier & Laloui (2004) investigated the evolution of PSD upon drying of a sandy loam compacted on the dry side of optimum, saturated and finally desaturated in the pressure plate. Figure 22 shows the drying curve in terms of void ratio and water ratio (void ratio was calculated from MIP data and corrected for the fraction of pores not intruded by mercury calculated at saturation). As suction is increased from 0 to 200 kPa, the inter-aggregate porosity significantly decreases and the intraaggregate porosity progressively increases even if the total porosity remains essentially unchanged (Figure 23). When suction is increased from 200 to 400 kPa the inter-aggregate porosity seems to increase again but no higher suctions were applied and it is therefore difficult to understand whether this is simply an experimental error.
The evolution of PSD upon drying was also investigated by Simms & Yanful (2001) by testing a glacial till compacted on the wet side of optimum (the drying curves in terms of void ratio and water ratio are shown in Figure 24). Similarly to Cuisinier & Laloui (2004), the drying process tends to erase the macropores (capillary suction causes the larger pores to shrink) and eventually the initial bi-modal pore-size distribution transformed into a monomodal pore-size distribution (Figure 25). However, when the soil was air-dried after applying a suction of 2500 kPa (Figure 26), inter-aggregates appeared again, probably due to the elastic rebound of pores that desaturate d (desaturation causes a release of the capillary stress). Again, the significant change in pore size distribution upon air-drying shown in Figure 26 occurred at nearly constant overall void ratio. These MIP data shows that there is not always a direct relationship between microscopic and macroscopic behaviour (e.g. PSD may change significantly at constant void ratio) and that it is difficult to classify a soil as a double-porosity or doublestructure geomaterials as the PSD evolves during hydraulic and mechanical loading.
Void ratio, e and water ratio ew
1.2
Void ratio, e and water ratio ew
I 1 0.8 0.6 0.4 0 100 200 300 400 D1 D2 D3 D4
e ew
s=840 kPa 0.4

Air-dried s=1400 kPa s=2500 kPa
esat
Oven-dried
0.2
e ew
500
Matric suction (kPa)
1000
10000
100000
1000000
Figure 22. Evolution of water ratio and void ratio (estimated from MIP data) of compacted sandy loam upon drying (after Cuisinier & Laloui 2004).
Figure 24. Evolution of water ratio and void ratio (estimated from MIP data) of compacted glacial till upon drying (after Simms & Yanful 2001).
Figure 23. Evolution of pore size distributions of compacted sandy loam upon drying (Cuisinier & Laloui 2004).
Figure 25. Evolution of pore size distributions of compacted glacial till upon drying (Simms & Yanful 2001).
125
4 Volume frequency, eMIP/(logd)
Reconstituted (w=0.359, e=0.934, Sr=1.00, s=511 kPa) Compacted (w=0.299, e=0.965, Sr=0.81, s=550 kPa)
d=0.20 m
Reconstituted (w=0.276, e=0.863, Sr= 0.83, s=1240 kPa)

Volume frequency, eMIP/(logd)
Compacted
3
(w=0.215, e=1.049, Sr=0.53, s=1140 kPa)
Figure 26. Evolution of pore size distributions of compacted glacial till upon air- and oven-drying (Simms & Yanful 2001).
SOIL MICROSTRUCTURE: COMPACTED VERSUS RECONSTITUTED STATES
d=0.16 m
The variety of pore size distributions in both compacted and reconstituted soils has shown that the boundary between compacted and reconstituted states is blurred. At this point, it may therefore be worth comparing the microstructure of reconstituted and compacted states in terms of pore size distribution. Figure 27 compares the PSDs of Speswhite Kaolin (i) statically compacted to 1200 kPa and (ii) reconstituted from slurry and subsequently air-dried. The comparison is made by considering similar suction levels. It clearly appears that the reconstituted soil exhibits the same modal size as the intra-aggregate modal size of the compacted soil. This leads to the simple conclusion that aggregates are made of reconstituted soil and this appears to be the link between reconstituted and compacted soils. A similar conclusion can be drawn by comparing the PSD of Barcelona red clay in compacted and reconstituted states. Figure 28 shows the PSD of two samples reconstituted from slurry at 1.5 wl and consolidated to 100 kPa vertical stress. The first sample was tested immediately after removal from the consolidometer (s 0 kPa) and the second one was air-dried to a suction s 120 kPa (Boso 2005). The other two samples were compacted to different void ratios (e 0.82 and e 0.55) at the same water content of 12% and had a suction of 270 kPa as measured by a high-capacity tensiometer (Buenfil 2007). The reconstituted samples show mono-modal PSD with modal size that decreases when suction is increased from 0 to 120 kPa and void ratio is decreased from 0.648 to 0.453. No MIP tests were performed on reconstituted samples at higher suctions but it is reasonable to expect that the modal size would further reduce upon drying.
0 0.001 0.01 0.1 1 Entrance pore size, d: m 10 100
Figure 27. Comparison of PSD of compacted and reconstituted samples of Speswhite kaolin at similar suction (Tarantino, in prep., b).
Intrusion volume frequency, eMIP/(logd)
REC (ew=0.648, e=0.648, s=10 kPa) REC (ew=0.453, e=0.510, s=120 kPa) COM (ew=0.320, e=0.82, s=270 kPa) COM (ew=0.320, e=0.55, s=270 kPa)
1.2
0.8
Reconstituted Compacted
0.4
0 0.001 0.01 0.1 1 10 Entrance pore size, d: m 100 1000
Figure 28. Comparison of PSD of compacted and reconstituted Barcelona red clayey silt (after Boso 2005 and Buenfil 2007).
On the other hand, the two compacted samples exhibit a bi-modal pore size distribution with only the inter-aggregate pores affected by compaction. The modal size characterising the intra-aggregate porosity (0.4 m) is only slightly smaller than the modal size of the reconstituted sample at s 120 kPa and this difference is likely to be due to the different suction level (270 kPa for the compacted sample against 120 kPa for the reconstituted sample). Again, it would seem that aggregates are made of reconstituted soil.
126
A similar conclusion can also be drawn from MIP data on compacted and reconstituted samples of Bioley silt (Koliji et al. 2010). Compacted samples were prepared by compacting the soil on the dry side of optimum, which was then gently crushed and sieved. Aggregates were oven-dried and then wetted to w 13% using the procedure described by Koliji (2008). Reconstituted samples were prepared by mixing the soil powder to a water content slightly lower than the liquid limit. Figure 29 compares the PSD of compacted (PS) and reconstituted (PR) samples of Bioley silt at zero suction (PR2 and PS3) and at a suction of 500 kPa and a vertical stress of 500 kPa (PR1 and PS1b). Again, the modal size of the mono-modal pore-size distribution of the reconstituted sample appears to be very similar to the intra-aggregate modal size of the compacted sample.
In the light of these findings, it is not surprising that the field- and laboratory-compacted samples in Figure 15 exhibited the same intra-aggregate porosity. Aggregates forming upon water spraying in the laboratory have a reconstituted structure which is likely to be very similar to natural clay used as construction material. 6 RESPONSE UNDER SATURATED CONDITIONS
Figure 29. Comparison of PSD of compacted (PS) and reconstituted (PR) samples of Bioley silt at zero suction (PR2 and PS3) and a suction of 500 kPa and a vertical stress of 500 kPa (PR1 and PS1b) (after Koliji et al. 2010).
A first comparison between soils in compacted and reconstituted states can be made by considering the response under saturated conditions. Pore size distribution data of samples compacted and then fully hydrated seem to suggest that the aggregated structure of compacted soils is not destroyed by hydration. It would therefore be expected that some differences occur in the mechanical response of compacted (aggregated) soils and reconstituted soils, where aggregation occurs only at the level of elementary particle assemblages. Tarantino & Tombolato (2005) compare the direct shear strength of compacted and reconstituted samples of Kaolin Speswhite. Six specimens were statically compacted at different values of vertical stress and water content and then sheared in the direct shear box. Reconstituted specimens were first consolidated to 100 kPa vertical stress in the consolidometer. Specimens were then cut and trimmed from the cake and sheared in the direct shear box at different vertical stresses. The results from these tests are shown in Figure 30, where the ultimate shear stress and the ultimate void ratio e are plotted against the effective vertical stress . The ultimate shear strength of the compacted specimens appears to be independent of the compaction conditions, as all data lies on the same curve. Ultimate shear strength data from reconstituted specimens appear to fall on the same envelope defined by the compacted specimens. In addition, the relationship between the void ratio e and the vertical stress at failure also appear to be the same regardless of whether the soil was reconstituted or compacted (the void ratio was determined by taking a mass of soil from the shear surface at the end of the test). The process of saturation has probably caused aggregates to swell with a reduction of the interaggregate space in the compacted samples. This has been probably further reduced by shearing (all samples exhibited contractile behaviour). The microstructure of the compacted soil made by a closed packing of reconstituted aggregates would therefore not be very different from the microstructure of the reconstituted soil. This would explain the same response of compacted and reconstituted soils upon shearing.
127
400
Shear stress : kPa
300 A, B, C 200 D, E, F 100

Compacted samples (ultimate) Reconstituted samples (NC)
Ultimate
0 0 400 800 Effective vertical stress 'v : kPa 1200
1.3 e = ew = 1.63 - 0.088 ln('v) D Void ratio e 1.1 E F e = ew = 1.61 - 0.088 ln('v) A C B
1.2
0.9
0.8 100 200 500 1000 Effective vertical stress 'v: kPa 2000
Figure 30. Direct shear data on saturated compacted samples and reconstituted normally consolidated samples of Speswhite kaolin (after Tarantino & Tombolato 2005).
Figure 31. Triaxial data on saturated compacted samples and reconstituted samples of a clayey silt (Jotisinkasa 2005, Jotisinkasa et al. 2009).
Jotisinkasa et al. (2009) tested a mixture of 70% silt, 20% kaolin and 10% London clay in compacted and reconstituted states. Again, in terms of ultimate shear strength there seems to be no difference between reconstituted and compacted samples, TR and TC respectively in Figure 31. Differences appear in the volumetric plane as the critical state line for the compacted soil lies below the one relative to the reconstituted states (Figure 31b). Differently from the Speswhite Kaolin, the mixture tested by Jotisinkasa et al. (2009) contains a significant silt fraction which appears to control the shear strength of the mixture (the critical state friction angle of 32.7 is clearly characteristic of a silt and not of a clay). The fabric of the reconstituted soil observed from thin sections appears to be a uniform matrix of silt and clay particles. On the other hand, the compacted soil exhibits a microstructure where aggregates interact with silt grains (Jotisinkasa 2005). Upon shearing, this microstructure cannot reverse to the uniform matrix of the reconstituted soil and this is perhaps the reason why the critical state lines of the compacted and reconstituted soils appear to be different.
In terms of compression behaviour, Figure 32 compares the virgin compression of reconstituted and compacted Barcelona red clayey silt. In particular, isotopic compression data of compacted samples obtained by Buenfil (2007) are compared with oedometer virgin compression data on reconstituted samples obtained by Boso (2005) (it was assumed k0 1 sin to convert the vertical stress to mean effective stress). There is good agreement between the virgin compression data of compacted and reconstituted samples. Virgin compression behaviour of compacted and reconstituted kaolin under saturated conditions is shown Figure 32. The two virgin compression lines appear to have the same slope but the line relative to the reconstituted state lies above the line associated with the compacted state. It is not clear if this is a fundamental difference or it is simply associated with the error in the estimate of the absolute void ratio (tests were carried out in different oedometers and the reference void ratio was estimated in different ways). Monroy (2005) also compared the virgin compression behaviour of London clay in reconstituted (icl) and compacted samples (c25) under saturated conditions. The curve relative to the compacted
128
0.7
Reconstituted from slurry Compacted (wc=0.12, ec=0.82))
0.6 Void ratio, e
0.5
0.4
specimen seems to converge at high stresses to the reconstituted curve. It is however unclear whether the compacted curve tends to become asymptotic with the reconstituted curve at higher stresses or tends to cross it. From the limited dataset presented in this section, it may be concluded that compacted and reconstituted soils in saturated conditions tend to exhibit the same shear strength. In terms of volumetric behaviour (both virgin compression and critical state), data are not consistent and it is difficult to draw any conclusion in lack of additional data. 7 WATER RETENTION BEHAVIOUR
0.1
10 100 Mean effecvtive stress, p': kPa
1000
Figure 32. Virgin compression data for reconstituted and compacted Barcelona red clayey silt (after Boso 2005 and Buenfil 2007).
1.5 1.4 1.3 1.2 Void ratio, e 1.1 1.0 0.9 0.8 0.7 0.6 1 10 100 1000 Vertical effective stress, 'v: kPa 10000
Reconstituted from slurry Compacted (wc=0.303) Compacted (wc=0.294) Compacted (wc=0.231)
Figure 33. Virgin compression data for reconstituted and compacted Speshwite kaolin (after Tarantino & De Col 2008 and Galvani 2003).
Figure 34. Virgin compression data for reconstituted (icl) and compacted fully hydrated (c25) London clay (Monroy 2005).
The relationship between suction, s, and degree of saturation, Sr, or the water content w, is referred to as water retention function and plays a key role in the mechanical and hydraulic behaviour of unsaturated soils. The relationships s-Sr or s-w are hysteretic as the response of a soil dried from saturated state (main drying) differs from that of the same soil wetted from dry state (main wetting). The main drying and main wetting curves mark out the domain of possible attainable states (hysteresis domain). If a saturated soil is dried to an intermediate degree of saturation and is then wetted, the corresponding s-Sr curve scans the hysteresis domain, from the main drying towards the main wetting curve, and is therefore referred to as scanning curve. The behaviour in the scanning region is often assumed to be reversible, whereas water retention behaviour is irreversible along the main drying and main wetting curve. Earlier work by Ridely (1993), Huang et al. (1998), Vanapalli et al. (1999), Karube & Kawai (2001) have shown that the main drying and the main wetting curves for both reconstituted and compacted soils are void ratio dependent and that main drying and main wetting in deformable soils are then characterised by two surfaces in the space (s, e, Sr) or (s, e, ew), e being the void ratio and ew the water ratio. These findings have been later confirmed by Gallipoli et al. (2003b), Tarantino and Tombolato (2005), Tarantino (2009), and Salager et al. (2010). Romero et al. (1999) were perhaps the first to recognize the interplay between microstructure and water retention behaviour in compacted soils. Figure 35 shows the main drying and main wetting water retention curves at constant void ratio of Boom clay for two dry densities, 13.7 and 16.7 kN/m3 respectively. For water contents greater than 15%, the main drying and main wetting curves depend on void ratio, with the air-entry suction increasing as the dry density increases. In the range of water
129
(a)
Degree of saturation, Sr 0.8
0.6
e=0.940.03 e=1.110.01 e=1.310.05 Reconstituted Function Plot 7
0.4 1
(b)
Water ratio, ew
0.8
0.6
e=0.940.03 e=1.110.01 e=1.310.05 Reconstituted Function Plot 6
Figure 35. Main wetting and main drying retention curves at constant porosity. Intra- and inter-aggregate governing suctionzones (Romero et al. 1999).
0.4 100
Matric suction, sm: kPa
1 000
contents between 5% and 15%, the main drying and main wetting cease to depend on dry density although hysteretic effects have not vanished. Below the water content of 5%, the relationship between suction and water content becomes independent of the direction of the hydraulic path. It is interesting to observe that the separation between the region where water retention is controlled by void ratio and the region where water retention does no longer depend on void ratio is controlled by water content and not suction (Romero & Vaunat 2000). In particular, the water content of wm 15% separating these two regions has a clear physical meaning as it represents the limit between inter-aggregate and intra-aggregate porosity. This water content was referred to as microstructural water content by Romero & Vaunat (2000). As the water content decreases to values lower than wm, the contours of equal suction in the compaction plane become vertical (Figure 11), indicating that pore-water is no longer present in the inter-aggregate pores and has withdrawn into the aggregates. As a change in void ratio occurs mainly at the expenses of the inter-aggregate pores, suction is significantly affected only when water is present in the inter-aggregate pore-space, i.e. when the water content w wm. The fact that water only occupies the intra-aggregate pore-space at w 15% is confirmed by the PSD shown in Figure 10. Results on compacted kaolin presented by Tarantino & Tombolato (2005) essentially confirmed the findings by Romero et al. (1999). Figure 36 reports main wetting data for compacted
Figure 36. Selected main wetting data of compacted and reconstituted Speswhite Kaolin (solid symbols denote compacted samples wetted in a hydraulic fashion) (after Tarantino 2009).
Speswhite Kaolin. The degree of saturation was increased by adding water (hydraulic wetting) and by reducing void ratio at constant water content (mechanical wetting). Only compression data associated with vertical stress equal or greater than compaction vertical stress were assumed to be main wetting data. Selected data presented by Tarantino & Tombolato (2005) are grouped by constant void ratio and it is apparent that void ratio affects water retention behaviour in terms of degree of saturation (Figure 36a). The lower the void ratio, the higher the air-occlusion suction. The effect of void ratio on water retention behaviour of compacted and reconstituted soils has been already observed (Vanapalli et al. 1996; Huang et al. 1998; Romero & Vaunat, 2000; Karube & Kawai, 2001; Gallipoli et al. 2003a). However, Figure 36 calls attention to another important point. The main wetting curves at constant void ratio appear to be independent of the mechanism by which the degree of saturation is increased. In fact, data associated with either mechanical wetting (open symbols) or hydraulic wetting (solid symbols) appear to lie on the same wetting curves at constant void ratio. Similar results were obtained for reconstituted Speswhite Kaolin and reconstituted Barcelona red clayey silt (Tarantino 2009). This implies that the term wetting should be generalised and associated
130
to an increase in degree of saturation and not only water content. The effect of void ratio is much less pronounced if main wetting data are plotted in terms of water ratio (Figure 36b). Similarly to Romero et al. (1999), water retention curves in terms of water ratio ew tend to converge at high suctions. However, it is very interesting to note that the convergence occurs when the water retention curves for the compacted specimens approach the main wetting water retention curve for the same soil in reconstituted state (data from Tarantino 2009). This is clearly illustrated in Figure 37 where all data for compacted specimens are plotted together with the main wetting water retention curve of the reconstituted kaolin. This seems to confirm once again that the compacted soil is made of reconstituted aggregates. It is also worth noticing that the main wetting curves starts to deviate from the curve for the reconstituted soil at around 900 kPa (Figure 37). However, main wetting curves significantly depart from each other at around 400 kPa, the suction at which the degree of saturation of the reconstituted soil cease to increase upon a wetting path (Figure 36a). In other words, pore-water starts invading the inter-aggregate pore space as the reconstituted aggregates become quasi-saturated. Data on reconstituted Barcelona red clayey silt were obtained by Boso (2005). Samples were initially normally consolidated to 100, 300, 500 kPa vertical stress. A first series was air-dried from saturated condition (main drying) and a second series was first air-dried to hygroscopic condition and then wetted (main wetting). Figure 38 shows selected data grouped by constant void ratio. Again, the lower the void ratio, the higher is the air-entry suction (Figure 38a) both along the drying and the wetting path.
(a)
Degree of saturation, Sr 0.9 0.8 0.7
Drying (e=0.400.01)
0.6 0.5
Drying (e=0.460.01) Wetting (e=0.41) Wetting (e=0.38)
(b)
Inter-domain?
Water ratio, ew 0.4
ewm =0.32
0.3
Drying (e=0.400.01) Drying (e=0.460.01) Wetting (e=0.41) Wetting (e=0.38)
0.2 10 100 Matric suction, sm: kPa
Intra-domain?
1000
Figure 38. Selected water retention data of reconstituted red clayey silt normally consolidated to different vertical stresses (vc 100, 300, and 500 kPa) and subsequently dried from saturated conditions or wetted from hygroscopic conditions (after Boso 2005).
Compacted - Hydraulic wetting Compacted - Mechanical wetting Reconstituted
10
100 Matric suction, sm: kPa
1000
Figure 37. Main wetting data of compacted and reconstituted Speswhite Kaolin in terms of water ratio (after Tarantino 2009).
The effect of void ratio on the main drying and main wetting water retention curves is less pronounced when data are plotted in terms of water ratio (Figure 38b). In particular, the convergence of drying and wettin curves seems to ve controlled by water content (ew0.32) and not suction. The similarity between Figure 38b and Figure 35 is noteworthy. For the compacted Boom clay, the convergence of the water retention curves occurred when pore-water withdrew into the aggregates. For the case of the reconstituted Barcelona red clayey silt, it may be tentatively assumed that convergence occurred when pore-water withdrew into microaggregates, i.e. elementary particle assemblages. As discussed in Section 3, aggregation is present also in reconstituted soils similarly to compacted soils, the difference only lying on the size of the aggregates. It is then interesting to compare the water retention curves of reconstituted and compacted Barcelona red clayey silt. Barrera (2002) compacted a sample on the dry side of optimum exhibiting an initial total suction of around 2 MPa (EDO-2). This sample was subsequently placed in an axistranslation oedometer and equalised to a matric suction of 0.8 MPa. Afterwards, the sample was progressively wetted in the axis-translation oedometer. As suction was constantly decreased, the hydraulic path was therefore a main-wetting path.
Water ratio, ew
131
Buenfil (2007) compacted two samples on the dry side of optimum (with an initial suction of 270 kPa), equalised in the axis-translation oedometer at 100 kPa and then wetted until saturation and subsequently dried. Hydraulic paths investigated were therefore main wetting and main drying paths. The comparison between compacted and reconstituted states upon wetting is shown in Figure 39a. At water ratios lower than 0.32, main wetting data from reconstituted and compacted samples superpose. The main wetting curves of the two compacted samples depart from the reconstituted curve at water ratios greater than 0.32. Similarly, the main drying curve on compacted sample (e 0.540.57) tend to converge to the reconstituted curve at a slightly lower water ratio (ew0.28) as shown in Figure 39b. It is remarkable that convergence of compacted and reconstituted curves occurs at approximately the same water ratio where convergence of reconstituted curves at different void ratios is observed in Figure 38. On the other hand, the effect of void ratio on the main drying curves is more significant in the compacted state (compare Figure 38b and Figure 39b). It may be tentatively concluded that both reconstituted and compacted samples are aggregated soils, the main difference lying on the aggregate size. In the compacted samples, large aggregates produce large inter-aggregate pores, much larger
0.8
(a)
0.6
Compacted (e=0.75-0.82, Buenfil 2007) Compacted (EDO-2, Barrera 2002) Reconstituted (Boso 2005)
WETTING
0.4
ewm =0.32
0.2
Compacted (e=0.54-0.57, Buenfil 2007)
(b)
0.6
Compacted (e=0.75-0.82, Buenfil 2007) Reconstituted (Boso 2005)
DRYING
0.4
ewm =0.28
0.2
Compacted (e=0.54-0.57, Buenfil 2007)
0.1
10 100 Matric suction, sm: kPa
1000
10000
Figure 39. Main wetting and main drying data of compacted (solid symbols) and reconstituted (open symbols) red clayey silt samples (after Barrera 2002, Boso 2005, and Buenfil 2007).
than intra-aggregate pores (as evidence by the bi-modal PSD). As pore-water invades the interaggregate space, water retention curves significantly deviate from the curves of the reconstituted aggregates (Figure 39b). On the other hand, aggregates in reconstituted soils are much smaller in size and inter-aggregate and intra-aggregate pores are of the same order of magnitude (as evidenced by the mono-modal PSD). As a result, differences in the inter-aggregate portion of the water retention curves are less significant (Figure 39b). A comparison between water retention behaviour of soils in reconstituted and compacted states can also be made by analysing data obtained on London clay. Marinho (1994) compacted samples at different water contents and dry densities (Figure 40), which were saturated and subsequently dried. He also tested a sample reconstituted from slurry. Suction was measured using the filter paper in all tests. Main drying curves of compacted samples in terms of water ratio ew tend to converge to the same curve as suction is increased, which also corresponds to the main drying curve of the reconstituted specimen. It is very interesting to notice that the water ratio at which the drying curves converge to the reconstituted curve depends on the compaction water content. The specimen LCTS7 and LCTS8, compacted at the higher water contents, essentially superpose to the reconstituted main drying curve in terms of water ratio over the entire range of suction. The sample LCTS6 compacted at w 0.36 converges at suction of around 700 kPa, the samples LCTS2 and LCTS3, were both compacted at w 0.27 and both converged at a suction of around 4000 kPa. It may be inferred that, at high water content, the compacted soil is made of very large aggregates and reduced inter-aggregate pore space (see PSD in Figure 7 and Figure 21). In such a condition, the compacted soil made of large reconstituted aggregates would not significantly differ from the soil reconstituted from slurry. There is thefore no difference between the water retention curves of the reconstituted soil and the soil compacted at high water content. As compaction water content decrease, the size of the aggregates decreases and inter-aggregate pore space increases (see PSD in Figure 7 and Figure 8). As a result, pore-water remains into the inter-aggregate voids over a wide range of suction and differences are observed between compacted and reconstituted soil until water eventually withdraws into the aggregates. For the same reason, the sample compacted by Monroy (2005) at the relatively low water content of w 0.23 (see Figure 40) and subjected to a main wetting path in the osmotic oedometer, showed
Water ratio, ew
Water ratio, ew
132
1.6
(a)
Degree of saturation, Sr 0.9
0.95
1.02
0.765 0.68
0.849 0.929 0.8 1.09 0.952

Oedometer testing As-compacted
0.855
Dry density, d: Mg/m3
1.4
Monroy 2005 LCTS2 LCTS8
0.952 0.95
0.864
0.7
1.2
S
r
=1
.0
LCTS3
=0 Sr .4
LCTS6
S
r
LCTS7
1
S
=0
r
1 0.2 0.24
.7
.8
0.28 0.32 0.36 0.4 Compaction water content, w
0.44
0.48
Figure 40. Samples of London Clay subjected to main drying after saturation (LCTS series) and to main wetting after compaction (after Marinho 1994 and Monroy 2005).
100 Degree of saturation, Sr(%)
Water ratio ew
=0
=0 Sr .5
=0 Sr .6
(b)
1 10 Suction (kPa) 100 1000
Figure 42. Main wetting water retention data of compacted London Clay (after Monroy 2005).
80
60
LCTS7 LCTS8 LCTS2 LCTS3 LCTS6 Reconstituted
40
20
1.2 Water ratio ew
0.8
LCTS7 LCTS8 LCTS2 LCTS3 LCTS6 Reconstituted
0.4
100
1000 Suction (kPa)
10000
Figure 41. Water retention main drying data of reconstituted and compacted London Clay (after Marinho 1994).
a water retention curve that was affected by void ratio only at very low suctions. Figure 42 shows main wetting data obtained from oedometer tests where samples where subjected to compression at constant suction or wetting at constant vertical stress. Data shown in the figure include both mechanical wetting data (degree of saturation
increased mainly due to a reduction in void ratio) and hydraulic wetting (degree of saturation increased mainly due to an increase in water content. In the same figure post-compaction data are also reported as a reference. It can be seen that at suction higher than 40 kPa the water retention curve in terms of water ratio is not affected by void ratio (void ratios are given in Figure 40a). At suctions lower than around 40 kPa water ratio appear to depend on void ratio similarly to the Barcelona red clayey silt (compare with Figure 39a). Finally, it is interesting to show the results from Huang et al. (1998) who investigated the effect of void ratio on water retention behaviour of a reconstituted silty sand including a clay fraction of only 10%. Specimens were one-dimensionally preconsolidated to different normal stresses and completely unloaded prior to being moved to the pressure-plate cells. Samples were therefore dried in pressure-plate cell. Figure 43 shows that main drying water retention curves in terms of water ratio are affected by void ratio over a significant range of suction and that these curves tend to become independent of void ratio when the degree of saturation is close to the residual value. It may be inferred that the low clay content generates little aggregation and that the water retention behaviour is dominated by capillary water in the inter-grain pore space. As a result, water retention behaviour changes significantly as the inter-grain pore space is reduced by preliminary one-dimensional compression.
133
100 Degree of saturation, Sr (%)
80
60
40
PPCT26, e0=0.426 PPCT24, e0=0.474 PPCT22, e0=0.513
20 0.5 Water ratio, ew 0.4 0.3 0.2 0.1 100 Matric suction, s (kPa)
Figure 43. Main dying curves of reconstitued silty sand consolidated to different initial void ratios (after Huang et al. 1998).
CONCLUSIONS
aggregates in compacted soils and can be classified as domains or elementary particle arrangements as a first approximation. As intra- and inter-domain pores in reconstituted soils are of the same order of magnitude, it is difficult to differentiate them on the basis of the mercury intrusion porosimetry data. When comparing the PSD of the same soil in compacted and reconstituted state, it appears that the single modal size in reconstituted soils corresponds to the intra-aggregate modal size in compacted double-porosity soil. In other words, aggregates appear to be made of reconstituted soil. As in reconstituted soils, aggregates are therefore in turn formed by micro-aggregates. Water retention behaviour appears to be directly controlled by the microstructure. In particular, water retention curves appear to be controlled by void ratio when water is present in the inter-aggregate pore space. This applies to compacted soils but it would also seem to apply to reconstituted soils. Aggregation would be present in both compacted and reconstituted soils, the difference lying on the aggregate size, macro-aggregate in compacted soils and micro-aggregate in reconstituted soils. The comparison of the water retention curves of soils in compacted and reconstituted states then appears to confirm the reconstituted nature of the aggregates. ACKNOWLEDGMENTS The author wishes to thank Prof. Pierre Delage, Prof. Cristina Jommi, Dr. Azad Koliji, Dr. Enrique Romero, Dr. Andrew Ridley, and Prof. Paul Simms for the helpful discussions. He also wishes to thank Grainne McCloskey for kindly reviewing the manuscript and the helpful suggestions. REFERENCES
Ahmed, S., Lovell, C.W. Jr. & Diamond, S. 1974. Pore size and strength of compacted clay. Journal of the Geotechnical Engineering Division, ASCE, 100 (4): 407425. Alonso, E.E., Gens, A. & Hight, D.W. 1987. Special problems soils. General report. Proc. 9th Europ. Conf. Soil Mech. Found. Eng., Dublin, 3: 10871146. Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partly saturated soils. Gotechnique 40(3): 405430. Banin, A. & Amiel, A. 1970. A correlative study of the chemical and physical properties of a group of natural soils in Israel. Geoderma, 3: 185198. Barrera, M. 2002. Estudio experimental del comportamiento hidro-mecnico de suelos colapsables. PhD Thesis, Universidad Politecnica de Catalunya, Barcelona, Spain.
The paper has presented a comparison between compacted and reconstituted states. The microstructure was first examined by analysing MIP data. It has been shown that the compaction water content plays a key role in the microfabric of the compacted soil as it controls the size of the aggregates. A simple microstructure model has been presented to explain the relationship between compaction water content and aggregation. After compaction, soils compacted on the dry side of optimum exhibit a bi-modal pore size distribution whereas soils compacted on the wet side of optimum often exhibit (but not always) a mono-modal pore size distribution. However, mechanical and hydraulic effects can significantly modify the pore size distribution, which can evolve from mono-modal to bi-modal or vice versa. The classification of compacted soils in terms of pore size distribution (single- or double-porosity) may therefore be not always appropriate. In addition, pore size distribution may evolve significantly even when no macroscopic effects are observed as occurs when drying the compacted soils from saturated conditions. Reconstituted soils very often exhibit a monomodal pore size distribution even if aggregates are observed in the scanning electron microscope. Aggregates in reconstituted soils are smaller than
134
Boso, M. 2005. Shear strength behaviour of a reconstituted partially saturated clayey silt. PhD dissertation, Universit degli Studi di Trento, Italy. Buenfil, C.M. 2007. Caracterizacin experimental del comportamiento hidromecnico de una arcilla compactada. PhD Thesis, Universidad Politecnica de Catalunya, Barcelona, Spain. Carter, M.R. 2004. Researching structural complexity in agricultural soils. Soil & Tillage Research 79: 16. Christenson, H.K. 1994. Capillary condensation due to van der Waals attraction in wet slits. Phys. Rev. Lett., 73(13): 18211824. Collins, K. & McGown, A. 1974. The form and function of microfabric features in a variety of natural soils. Gotechnique, 24: 233254. Cui, Y.J. 1993. Etude du comportement dun limon compact non satur et sa modlisation dans un cadre lasto-plastique. PhD Thesis, CERMES-ENPC, Paris, France. Cui, Y.J. & Delage, P. 1996. Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique 46(2): 291311. Cuisinier, O. & Laloui, L. 2004. Fabric evolution during hydromecanical loading of a compacted silt. Int. J. Numer. Anal. Meth. Geomech. 28(6): 483499. Delage, P. 2009. Discussion: Compaction behaviour of clay. Gotechnique 59(1): 7577. Delage, P. 2010. A microstructure approach to the sensitivity and compressibility of some Eastern Canada sensitive clays. Gotechnique 60(5): 353368. Delage, P., Audiguier, M., Cui, Y.J. & Howat, M.D. 1996. Microstructure of a compacted silt. Canadian Geotechnical Journal, 33: 150158. Delage, P. & Graham, J. 1995. Mechanical behaviour of unsaturated soils: Understanding the behaviour of unsaturated soils requires reliable conceptual models. Unsaturated Soils. Proc. 1st Int. Conf. on Unsaturated Soils (UNSAT 95), Paris, France (ed. Alonso, E.E. and Delage, P.), Rotterdam: Balkema, 3: 12231256. Delage, P. & Lefebvre, G. 1984. Study of the structure of a sensitive Champlain clay and of its evolution during consolidation. Can. Geotech. J. 21(1): 2135. Delage, P., Marcial, D., Cui, Y.J. & Ruiz, X. 2006. Ageing effects in a compacted bentonite: a microstructure approach. Gotechnique 56(5): 291304. Delage, P. & Pellerin, F.M. 1984. Influence de la lyophilisation sur la structure dune argile sensible di Qubec. Clay Miner. 19(2): 151160. Delage, P., Tessier, D. & Marcel-Audiguier, M. 1982. Use of the Cryoscan apparatus for observation of freezefractured planes of a sensitive Quebec clay in scanning electron microscopy. Can. Geotech. J. 19(1): 111114. Diamond, S. 1970. Pore size distributions in clay. Clays Clay Miner. 18: 723. Diamond, S. 1971. Microstructure and pore structure of impact-compacted clays. Clays Clay Miner, 19: 239249. Dirksen, C. & Dasberg, S. 1993. Improved calibration of time domain reflectometry soil water content measurements. Soil Sci. Soc. Am. J., 57: 660667. Gallipoli, D., Gens, A., Sharma, R. & Vaunat, J. 2003a. An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Gotechnique 53, No. 1, 123136.
Gallipoli, D., Wheeler, S.J. & Karstunen, M. 2003b. Modelling the variation of degree of saturation in a deformable unsaturated soil. Gotechnique 53(1): 105112. Galvani, A. 2003. Resistenza a taglio di un argilla non satura ricostituita in laboratorio. Laurea Thesis, Universita` degli Studi di Trento, Italy. Garcia-Bengochea, I., Lovell, C.W., Altschaeffl, A.G. 1979. Pore distribution and permeability of silty clays Journal of the Geotechnical Engineering Division, ASCE 105(7): 839856. Gens, A., Alonso, E.E., Suriol, J. & Lloret, A. 1995. Effect of structure on the volumetric behaviour of a compacted soil. Proc. 1st Int. Conf. on Unsaturated Soils, Paris 1, 8388. Gomez R., Romero E., Lloret A. & Suriol, J. 2009. Characterising the collapsible response of an in-situ compacted silt. In Proc. of 4th Asia Pacific Conference on Unsaturated Soils (O. Buzzi, S. Fityus & D. Sheng, eds.), CRC Press: 371376. Griffiths, F.J. & Joshi, R.C. 1989. Change in pore size distribution due to consolidation of clays. Gotechnique, 39(1): 159167. Hattab, M., Bouziri-Adrouche, S. and Fleureau, J.-M. 2010. volution de la microtexture dune matrice kaolinitique sur chemin triaxial axisymtrique. Can. Geotech. J. 47(1): 3448. Huang, S., Barbour, S.L. & Fredlund, D.G. 1998. Development and verification of a coefficient of permeability function for a deformable unsaturated soil. Can. Geotech. J. 35(4): 411425. Iwamatsu, M. & Horii, K. 1996. Capillary condensation and adhesion of two wetter surfaces. J. Colloid Interface Sci., 182: 400406. Jommi, C. & Sciotti, A. 2003. A study of the microstructure to assess the relaiability of laboratory compacted soils as reference material for earth constructions. System-based Vision for Strategic and Creative Design, Bontempi, F. (ed.), 3: 24092415. Rotterdam: A.A. Balkema. Jotisankasa, A. 2005. Collapse behaviour of a compacted silty clay. PhD thesis, Imperial College London. Jotisankasa, A., Coop, M. & Ridley, A. 2009. The mechanical behaviour of an unsaturated compacted silty clay. Gotechnique 59(5): 415428. Karube, D. & Kawai, K. 2001. The role of pore water in the mechanical behaviour of unsaturated soils. Geotechnical and Geological Engineering 19: 211241. Koliji, A. 2008. Mechanical behaviour of unsaturated aggregated soils. Ph.D. Thesis, No. 4011, Ecole Polytechnique Fdrale de Lausanne, Switzerland. Koliji, A., Vulliet, L. & Laloui, L. 2010. Structural characterisation of unsaturated aggregated soils. Can. Geotech. J. 47: 297311. Li, Z.M. 1995 Compressibility and Collapsibility of Compacted Unsaturated Loessial Soils, Unsaturated Soils. Proc. 1st Int. Conf. on Unsaturated Soils (UNSAT 95), Paris, France (ed. Alonso, E.E. and Delage, P.), Rotterdam: Balkema, 1: 139144. Lloret, A., Villar, M.V., Sanchez, M., Gens, A., Pintado, X. & Alonso, E.E. 2003. Mechanical behaviour of heavily compacted bentonite under high suction changes. Gotechnique 53(1): 2740. Marinho, F.A.M., 1994. Shrinkage behaviour of some plastic soils. PhD thesis, University of London (Imperial College), London, UK.
135
Monroy, R. 2006. The influence of load and suction changes on the volumetric behaviour of compacted London Clay. PhD thesis, Imperial College, London. Monroy, R., Zdravkovic, L. & Ridley, A. 2010. Evolution of microstructure in compacted London Clay during wetting and loading. Gotechnique 60(2): 105119. Ninjgarav, E., Chung, S.-G., Jang, W.-Y., Ryu, C-K. 2007. Pore Size Distribution of Pusan Clay Measured by Mercury Intrusion Porosimetry. KSCE Journal of Civil Engineering 11(3): 133139. Prapaharan, S., White, D.M., Altschaeffl, A.G. 1991. Fabric of field- and laboratory-compacted clay. J Geotech Eng ASCE 117(12): 19341940. Prapaharan, S., White, D.M. & Altschaeffl, A.G. 1991. Fabric of field- and laboratory-compacted clay. Journal of Geotechnical Engineering, ASCE, 117(12): 19341940. Ridley, A.M. 1993. The measurement of soil moisture suction. Ph.D. Thesis, University of London. Romero, E. & Vaunat, J. 2000. Retention curves in deformable clays. In Experimental evidence and theoretical approaches in unsaturated soils (eds A. Tarantino and C. Mancuso), pp. 91106. Rotterdam: A.A. Balkema. Romero, E. and Simms, P. 2008. Microstructure investigation in unsaturated soils: A review with special attention to contribution a mercury intrusion porosimetry and environmental scanning electron microscopy. Geotech. Geol. Eng., 26: 705727. Romero, E., Gens, A. & Lloret, A., 1999. Water permeability, water retention and microstructure of unsaturated compacted Boom clay. Engineering Geology, 54: 117127. Salager S., El Youssoufi, M.S., and Saix C. 2010. Definition and experimental determination of a soil-water retention surface. Canadian Geotechnical Journal, 47(6): 609622. Simms P.H. & Yanful, E.K. 2001. Measurement and estimation of pore shrinkage and pore distribution in a clayey till during soil-water characteristic curve tests. Can Geotech J. 38:741754. Sridharan, A., Altschaeffl, A.G. & Diamond, S. 1971. Pore size distribution studies. J. Soil Mech. Found. Div., ASCE. 97: 771787. Tamagnini, R. 2004. An extended Cam-clay model for unsaturated soils with hydraulic hysteresis, Gotechnique 54(3): 223228. Tanaka, H., Shiwakoti, D.R., Omukai N., Rito F., Locat J.; Tanaka M. 2003. Pore size distribution of clayey soils measured by mercury intrusion porosimetry and its relation to hydraulic conductivity. Soils and Foundations, 43(6): 6367. Tang, A.-M. & Cui, Y.-J. 2005. Controlling suction by the vapour equilibrium technique at different temperatures and its application in determining the water retention properties of MX80 clay. Can. Geotech. J. 42: 287296.
Tarantino, A. & De Col E. 2009. Compaction behaviour of clay: Authors reply to discussion by P. Delage. Gotechnique, 59(1): 7576. Tarantino, A. (in prep.,a). Effects of aggregation on water retention behaviour of clayey soils. In prep. Tarantino, A. (in prep.,b). A microstructural interpretation of Bigots drying curve. In prep. Tarantino, A. 2009. A water retention model for deformable soils. Gotechnique 59(9): 751762. Tarantino, A. 2010. Basic concepts in the mechanics and hydraulics of unsaturated geomaterials. New Trends in the Mechanics of Unsaturated Geomaterials Lyesse Laloui (ed.). ISTEJohn Wiley & Sons. Tarantino, A. & De Col, E. 2008. Compaction behaviour of clay. Gotechnique 58(3): 199213. Tarantino, A. & Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique, 55(4): 307317. Thom, R., Sivakumar, R., Sivakumar, V., Murray, E.J. & Mackinnon, P. 2007. Pore size distribution of unsaturated compacted kaolin: the initial states and final states following saturation. Gotechnique 57(5): 469474. Tuller, M. & Or, D. 2005. Water films and scaling of soil characteristic curves at low water contents.Water Resour. Res., 41, W09403, doi:10.1029/2005 WR004142. Tuller, M., Or, D. & Dudley, L.M. 1999. Adsorption and capillary condensation in porous media: Liquid retention and interfacial configurations in angular pores. Water Resour. Res., 35(7): 19491964. Vanapalli, S.K., Fredlund, D.G. & Pufhal, D.E. (999. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Gotechnique 49(2): 143159. Vaunat, J., Romero, E. & Jommi, C. 2000. An elastoplastic hydromechanical model for unsaturated soils. In Experimental evidence and theoretical approaches in unsaturated soils (eds A. Tarantino and C. Mancuso), pp. 121138. Rotterdam: A.A. Balkema. Wheeler, S.J. & Sivakumar, V. 1995. An elasto-plastic critical state framework for unsaturated soil. Gotechnique 45(1): 3553. Wheeler, S.J. & Sivakumar, V. 2000. Influence of compaction procedure on the mechanical behaviour of an unsaturated compacted clay. Part 2: Shearing and constitutive modelling. Gotechnique 50(4): 369376. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. 2003. Coupling of hydraulic hysteresis and stressstrain behaviour in unsaturated soils. Gotechnique 53(1): 4154.
136
Unsaturated soil behaviour

Microstructure
Development of microstructure in compacted earthen building materials

C.T.S. Beckett & C.E. Augarde
School of Engineering and Computing Sciences, Durham University, Durham, UK
ABSTRACT: The earthen construction material rammed earth (RE) is effectively a manufactured unsaturated soil. The sources of strength in RE are presumed to be suction, interlock and cementation. Previous work (Jaquin et al., 2009) has proved the link between the first of these features and strength, and cementation is also reasonably well understood. What remains unclear is the nature of the particle interlock, which is controlled by the material microstructure (i.e. the particle and void size distributions and the arrangement of water held at menisci or in bulk) and so the compaction process by which the material is made is of major significance. In this paper we present initial results from laboratory testing on artificial RE materials, to investigate microstructure formation during compaction. Use was made of so-called transparent clay materials, and a specially constructed compaction chamber to enable particle movements to be viewed more clearly. Novel use of a document scanner allowed images of the compacted material to be acquired. Particle Image Velocimetry (PIV) software was used to determine particle displacements from the scanned images. Various difficulties with the procedure are outlined and some results are presented showing the movement of particles and the consequential development of microstructure. 1 INTRODUCTION particle interlock and improves clay bonding by increasing inter-particle connectivity through clay bridges. Therefore, an understanding of the effect of compaction on RE is essential to discover the effects of particle interlock and clay cementation on material strength. Particle Image Velocimetry (PIV) is a technique which allows the movement of particles to be traced from one image to another, by comparing digital images of a particle flow taken at different times during a test. It was originally developed for use in fluid mechanics. In PIV, a test patch of pixels in one image is compared with a corresponding (larger) search patch in a subsequent image, and the best match for the test patch within the search patch is found (Adrian, 1991). In fluid mechanics, double-flash photography of a fluid flow containing seed particles is used to create photographs containing image pairs of each seed particle: the images are then divided into a grid of test patches, and the displacement of each test patch during the interval between the flashes is found (White et al., 2003). This technique can be applied to geotechnical materials as their texture allows for the correlation of test and search patches between images without the need for seed particles, although a soil of sufficient coarseness is required. The movement of the soil particles is then taken to be the centroidal displacement of the test patch A grid of control markers, placed at known intervals, allows
Rammed earth (RE) is a form of unbaked earthen construction. It is formed by placing layers (normally 100 to 150 mm thick) of moist sandy-loam sub-soil inside temporary formwork, which are then compacted. The formwork is then removed to allow the structure to dry, a process through which RE gains much of its strength (Betts and Miller, 1937). Until recently, the design of RE structures was based on past experience. However, the need to consider the use of RE in unusual situations, or subject to unusual loadings has prompted scientific examination of the material (Jaquin et al., 2009). Current research is investigating how RE can be treated as a highly unsaturated soil, and how properties of its microstructure affect the macro-structural properties (for example strength and permeability) (Hall and Djerbib, 2004a; Hall and Djerbib, 2004b; Jaquin et al., 2009). The sources of strength in RE are presumed to be suction, interlock and cementation. Previous research (Jaquin et al., 2009) has proved the link between the first of these features and strength, but particle interlock and cementation remain only loosely understood. Although compaction does not affect the suctions present in a soil dry of the optimum water content (Gens et al., 1995; Dineen et al., 1999; Romero, 1999; Tarantino and De Col, 2008), it is clear that it does increase the
139
the pixel displacements (image-space displacements) to be converted into mm displacements (object-space displacements). The accuracy of PIV depends on the quality of the image and the number of pixels used per test patch: the greater the number of pixels, the lower the error in finding the best match in the subsequent image but the longer the analysis time (White, 2002). The aim of this paper is to combine the use of geotechnical PIV with an analysis of the compaction process in order to determine the role played by compaction in the formation of REs microstructure which could contribute to our understanding of this materials behaviour. 2 2.1 MATERIALS AND METHODS Manufacture of artificial clay
CRATerre institute (Houben and Guillaud, 1996), allowing for the greatest transparency due to the high TCD material content. The labelling of an RE according to its sand:gravel:silty-clay content was first proposed by Hall and Djerbib (2004), as all RE mixes fall within the silty loam soil type category, and so cannot be distinguished by traditional classifications. Adopting this labelling method therefore allows results of testing on different RE mixes to be compared more accurately to previous results. 2.2 Design of the compaction chamber
Although RE is predominantly formed from sand and gravel (Houben and Guillaud, 1996), it has a fine texture due to its clay content. As a result, it is not suitable to be used with PIV analysis, due to an inability to track test patches from one image to another. It was therefore necessary to manufacture an artificial clay to be used as a substitute for the clay in RE which would increase the coarseness and enable PIV tracking. For this investigation, a transparent clay (Trinity College Dublin or TCD material, McKelvey et al., 2004) was selected so that, when mixed with sand and gravel, the resulting material would display the coarseness of a sand but behave like a clayey soil. Tests conducted in McKelvey et al. (2004) have shown that the TCD material alone closely resembles a normally consolidated kaolin clay, making it an acceptable substitute for typical clays used in RE, as kaolins are preferred due to their low swelling capacity (Houben and Guillaud, 1996). The TCD material was manufactured according to the materials and methods specified in McKelvey et al. (2004). A slurry comprising 7% (by mass) of fumed silica in a mixture of liquid paraffin and white spirit (mixed in a 7:3 ratio by mass) was placed inside a Rowe cell and consolidated under a constant vertical pressure of 70 kPa using a rubber bellofram. The fumed silica used in the preparation was Aerosil 300 SP fumed silica, provided by Evonik Industries.Top, bottom and side drainage were used during the consolidation process. The TCD material was added to sand and gravel in the proportions 5:1:4 sand:gravel:TCD (by mass) to create an artificial RE (ARE) material. These proportions were selected as a 5:1:4 mix of sand:gravel:silty-clay represents the highest claycontent for a RE material recommended by the
A sketch of the compaction chamber is shown in Figure 1. The main features of the chamber are two 10 mm thick Perspex panels, front and rear, to allow the compacted material to be viewed, and two 15 mm thick steel sides to prevent bulging. A photograph of the compaction chamber as used in the experiment is shown in Figure 2. PIV analysis uses close-range photogrammetry (CRP) to convert between pixel-space and objectspace coordinates when using a digital camera (White et al., 2003). However, determining the required CRP parameters can be time-consuming and is prone to inaccuracies. In order to remove the need to find these CRP parameters, a document scanner, mounted side-on to the compaction chamber, was used to capture images of the compacted material, rather than a digital camera. As well as having the advantage of not requiring
Adjustable rammer
Rammer carriage Guide rails
Perspex front and rear panels
Steel sides
Steel base
30 0
Figure 1. Sketch of the compaction chamber, showing principal dimensions (in mm).
140
250
Rammer
2.3 Preliminary testing and experimental procedure A series of preliminary tests was carried out to ensure that the scanned images would be adequate for PIV analysis. A mixture of pure sands was used for these tests to avoid wasting the TCD material. The PIV software geoPIV, supplied by Prof. Dave White (UWA), was used to identify the optimum scanning quality for the analysis. It was decided that a setting of 600 dpi be used, as the resulting images were then sufficiently detailed to allow for accurate PIV analysis but not so large as to require too great a computational effort. Further tests were carried out to determine the optimum size (in pixels) for the test patches. Results showed that a size of 128 128 pixels, at a central spacing of 128 pixels (i.e. the patches are in contact without overlapping) produced the clearest data in an acceptable analysis time. As a comparison, analyses were also conducted using sizes of 64 64 pixels at 64 pixel centres, and 256 256 pixels at 256 pixel centres, to create more- and less-dense meshes of test patches respectively. It was found that the more-dense mesh produced too many vectors for the displacement field to be legible, whilst the less-dense mesh did not produce a sufficiently detailed displacement field. As patches should be of size 2n pixels (where n is an integer) (White, 2002), no further patch sizes were investigated. The search boundary is a border of pixels placed around the test patch to determine the size of the search patch (and so the maximum displacement that can be recorded). Search boundary limits of 10, 30 and 50 pixels were tested with combinations of the above patch sizes: a 50 pixel boundary produced too many wild vectors, whilst a 10 pixel boundary could not adequately track particle displacements. 30 pixels was therefore used as a compromise, creating a corresponding maximum displacement for tracking of roughly 1.3 mm. Compaction testing was conducted by placing a level, 110 mm deep layer of ARE material into the compaction chamber. Control markers were placed at known intervals on the front of the Perspex panel to allow displacements to be converted from image to object space. A scan of the uncompacted material was taken to act as a starting point for the PIV analysis before the rammer was set to a drop height of 300 mm above the ARE surface and allowed to drop freely, at the right hand edge of the chamber. The rammers resting height was recorded before it was retracted to allow a new scan of the material to be taken. The rammer was then moved 50 mm to the left and the process repeated, setting the drop height to 300 mm above the previous resting height each time. The rammer was traversed from one side to the other at 50 mm
Scanner mounted sideon to chamber
Clamps
Figure 2. Photograph of the compaction chamber, showing scanner, clamps and raised rammer.
the CRP parameters, the equivalent image size is greatly increased: for example, a scanning density of 600 dpi results in an image of 35 megapixels over an A4 page, as opposed to a modest 4 megapixel image offered by current mid-range cameras. The greater number of pixels allows larger test patches to be used, which increases the tracking accuracy. The size of the compaction chamber was, however, therefore limited by the window of the document scanner, in this case to roughly 330 250 mm. The rim of the scanner window is trimmed so that the scanning glass sits flush against the Perspex panel, and the scanner is held in place by two clamps, anchored to the base of the compaction chamber, ensuring that its position does not change during compaction. Although external RE walls are of the order of 600 mm thick (Walker et al., 2005), internal walls can be roughly 300 mm thick and so their cross sections can be modelled by this apparatus. As RE walls are formed by layering earth in 150 mm deep layers and then compacting (to roughly 100 mm deep) (Walker et al., 2005), a maximum of two RE layers can therefore be accommodated by the compaction chamber. A 2.5 kg flat-head rammer is mounted above the chamber: the drop height is adjustable to ensure that a constant drop of 300 mm is achieved, as is required by the light Proctor test (LPT). The LPT was selected as it more closely approximates construction conditions. The rammer has a 100 100 mm face, and can be moved horizontally across the chamber by means of the rammer carriage (see Figure 1).
141
intervals until the material was fully compacted. A further 110 mm deep layer of ARE material was then added on top of the first and the process of scanning and compacting repeated. 3 3.1 RESULTS AND DISCUSSION Material behaviour
When poured, a mixture of sands will either stratify or segregate, depending on the friction angles of the component sand particles (Makse et al., 1997). This phenomenon is not present, however, if the sands are wet or otherwise cohesive (Samadani and Kudrolli, 2001). Previous work by the author has shown this to be the case. As the ARE material did not show any signs of stratification or segregation, it can be said that it is not acting as a mixture of sands but instead as a cohesive soil. The final depths of the first and second layers were 67.1 mm and 73.7 mm respectively, corresponding to a 39.0% and 33.0% compaction (by volume): this is a favourable comparison to the expected compaction of an actual RE layer, which is roughly 33.3% (Walker et al., 2005), further indicating that the ARE material is a valid substitute for an actual RE mix for compaction testing. The TCD material manufactured for this experiment did not share the transparent qualities of that used in McKelvey et al. (2004). Although this was not a concern in terms of this investigation (as the TCD material was mixed with sand and gravel to form the ARE), it is assumed that the lack of transparency was due to the use of rapid draining during consolidation: less drainage might produce a more transparent material.
Results for the PIV analysis of the scanned images are shown in Figures 35: the displacement field shows the movements of the centroids of the test patches, such that a dot represents no movement. The image behind the displacement field shows the deformed material after that compaction cycle. The axes show the size of the scanning window, in mm (note that the units on the ordinate axis are inverted, as pixels are numbered from the top left corner of the image: displacements have been corrected for this). Figures 3 and 5 show the displacements after two strikes of the rammer, whilst Figure 4 shows displacements half way through the compaction process (after roughly 20 strikes). Figures 3 and 5 show that compaction results in a significant disturbance of the underlying loose material, seen through the large displacement vectors. Recorded displacements in this region are not representative, however, as displacements are greater than 1.3 mm: instead, the highly-disturbed regions in Figures 3 and 5 show where compaction is at its greatest. As expected, these regions are immediately under the impact of the rammer. However, Figures 3 and 5 also show that material is pushed out from the impact region and into the uncompacted material due to the rigid base of the compaction chamber, such that the material is kneaded as well as compacted. Kneading is identified as essential to achieve a well-compacted material (Hillel, 1998) and a strong RE structure (Betts and Miller, 1937). As RE walls are built on top of concrete stem walls to avoid water damage (Walker et al., 2005; Easton, 2007), it is likely that a similar effect occurs in the construction of actual RE structures.
Figure 3. (in mm).
Initial particle displacements (scaled by a factor of 5) during compaction of the first ARE material layer
142
Figure 4. Particle displacements (scaled by a factor of 5) half way through the compaction of the first ARE material layer (in mm).
Figure 5. (in mm).
Initial particle displacements (scaled by a factor of 5) during compaction of the second ARE material layer
The darker material visible in Figures 35 shows where some of the mixture of liquid paraffin and white spirit has been forced out of the TCD material due to compaction. As such, it indicates regions that have been compacted more than others (i.e. regions of higher density). Despite the effect of kneading, it can be seen from Figure 3 that the material towards the top of a rammed layer is denser than that towards the bottom. It is possible, therefore, that the use of a differently-shaped rammer head (for example a heart-shaped rammer (Betts and Miller, 1937)) would produce a more uniform material density, due to a greater penetration of the material and so a more prolonged kneading effect.
Figure 4 shows that a layer underneath one that is being compacted does not deform further (the boundary between layers is clearly visible as a staggered dark band of material running across the figure at roughly 175 mm). This is contrary to existing predictions, which assume that the underlying layers become slightly denser as subsequent layers are compacted above them and as the building settles after construction. This is important in terms of being able to model the material properties of RE, as the compacted density is seen to be independent of the size of the structure. As the scanner is shielded by the Perspex panel on the front of the compaction chamber, the resulting images are quite blurry when compared to images
143
of objects resting directly on the scanning glass. This is a concern as the result of a blurry image is to reduce the apparent texture of the material by making the colours of neighbouring pixels closer to each other. However, as the Perspex thickness is constant the blurring effect will be uniform and so will not lead to an exaggeration of some displacements over others. Although the results show that the scanned images can be used for PIV analysis, it is likely that more accurate analyses can be performed if the Perspex is made thinner (say to 5 mm), although fully-focused images will not be possible due to the need for a front panel. 4 CONCLUSIONS
The tests outlined in this paper have shown that the use of an artificial rammed earth, formed from the TCD material mixed with sand and gravel, is a suitable substitute for RE for compaction testing. It has also been shown that the use of a document scanner is a valid method for obtaining large images for use in PIV analysis without the need for determining close-range photogrammetry parameters or the use of expensive capture devices. The results obtained in this preliminary study are encouraging and suggest that the approach described here may provide insights into the formation of microstructure in compacted soils. ACKNOWLEDGEMENTS The authors thank Prof. Dave White (UWA) for the PIV program geoPIV and Mr. Rob Stow (Durham Tees Valley airport) for supplying the raw materials needed for TCD material manufacture. The author is supported by a studentship awarded by the School of Engineering and Computing Sciences, Durham University. REFERENCES
Adrian, R.J. 1991. Particle Imaging Techniques for Experimental Fluid Mechanics. Annual Review of Fluid Mechanics 23: 261304. Betts, M.C. and Miller, T.A.H. 1937. Rammed Earth Walls for Buildings. Farmers Bulletin 1500.
Dineen, K., Colmanares, J.E., Ridley, A.M. and Burland, J.B. 1999. Suction and Volume Changes Of a Bentonite-Enriched Sand. Geotechnical Engineering 137: 197201. Easton, D. (Completely Revised Edition) 2007. The Rammed Earth House. Vermont (USA): Chelsea Green Publication Company. Gens, A., Alonso, E.E., Suriol, J. and Lloret, A. Effect of Structure on the Volumetric Behaviour of a Compacted Soil. First International Conference on Unsaturated Soils, Paris, France, 1995, UNSAT 95: Hall, M. and Djerbib, Y. 2004a. Moisture Ingress in Rammed Earth: Part 1the Effect of Soil ParticleSize Distribution on the Rate of Capillary Suction. Construction and Building Materials 18 (4): 269280. Hall, M. and Djerbib, Y. 2004b. Rammed Earth Sample Production: Context, Recommendations and Consistency. Construction and Building Materials 18 (4): 281286. Hillel, D. 1998. Environmental Soil Physics. California (USA): Academic Press. Houben, H. and Guillaud, H. (Second ed.) 1996. Earth Constructiona Comprehensive Guide. London (UK): Intermediate Technology Publications. Jaquin, P.A., Augarde, C.E., Gallipoli, D. and Toll, D.G. 2009. The Strength of Unstabilised Rammed Earth Materials. Gotechnique 59 (5): 487490. Makse, H.A., Havlin, S., King, P.R. and Stanley, H.E. 1997. Spontaneous Stratification in Granular Mixtures. Nature 386 (March): 379381. McKelvey, D., Sivakumar, V., Bell, A. and Graham, J. 2004. Modelling Vibrated Stone Columns in Soft Clay. Proceedings of the Institution of Civil EngineersGeotechnical Engineering 157 (3): 137149. Romero, E. 1999. Characterisation and Thermo-Hydro Mechanical Behaviour of Unsaturated Boom Clay: An Experimental Study, PhD Thesis, Universsitat Politecnica de Catalonia. Samadani, A. and Kudrolli, A. 2001. Angle of Repose and Segregation in Cohesive Granular Matter. Physical Review E 6405 (5): 9. Tarantino, A. and De Col, E. 2008. Compaction Behaviour of Clay. Gotechnique 58 (3): 199213. Walker, P., Keable, R., Martin, J. and Maniatidis, V. 2005. Rammed Earth: Design and Construction Guidelines. Watford (UK): BRE Bookshop. White, D.J. 2002. An Investigation into the Behaviour of Pressed-in Piles, Thesis, University of Cambridge. White, D.J., Take, W.A. and Bolton, M.D. 2003. Soil Deformation Measurement Using Particle Image Velocimetry (Piv) and Photogrammetry. Gotechnique 53 (7): 619631.
144
Evolution of meso-structure of intact loess during triaxial shear test

Xiang-Wei Fang Chun-Ni Shen
Department of Architectural and Civil Engineering, Logistic Engineering University, Chongqing, China School of Civil Engineering and Architecture, Chongqing University of Science and Technology, Chongqing, China
Zheng-Han Chen Zheng-Bin Zhou
Department of Architectural and Civil Engineering, Logistic Engineering University, Chongqing, China Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong
ABSTRACT: A series of triaxial shear tests on intact unsaturated loess were conducted using CT-multifunction triaxial apparatus developed in Logistical Engineering University. The distinct CT images and detailed CT data during the evolution of meso-structure were obtained nondestructively. A new method was proposed to identify the yield stress of intact loess using curves of CT-number versus deviatoric stress. Data analyzed from captured images were used to define a structure parameter and an evolution variable which characterized the evolution of meso-structure. An equation describing the evolution of structure was developed. The equation reflected the influences of net cell pressure and suction on the evolution of meso-structure. Keywords: unsaturated soils; intact loess; computed tomography (CT); meso-structure; equation of evolution CT-multi-function triaxial apparatus developed in Logistical Engineering University. Data analyzed from captured images were used to define a structure parameter and an evolution variable which characterize the evolution of meso-structure. An equation on the evolution of structure was developed. 2 2.1 APPARATUS AND TEST PROGRAMME Laboratory equipment
INTRODUCTION
The CT technique provided an excellent experimental method to study the evolution of structure and failure mechanism of soils. Desrues, et al. (1996) observed shear bands in tranial sand specimens by the CT technique. Finno, et al. (1997) studied shear bands in plane strain compression of loose sand under undrainedand drained conditions using stereophotogrammetry. Otani, et al. (2000) utilized CT method to study the failure behaviors of soils. However, the specimens were scanned by CT after unloading in the above mentioned literature. Chen, et al. (2001) developed a modified traxial apparatus utilizing CT to real time observe internal structure of specimens during testing and studied the evolution of meso-structure of expansive soils during loading using the developed apparatus. In this paper, a series of triaxial shear tests on unsaturated intact loess were conducted using
Chen, et al. (2001) developed a modified traxial apparatus utilizing CT to real time observe internal structure of specimens. Chen, et al. (2007) and Fang (2008) further improved the apparatus named as CT-multi-function triaxial apparatus for soils. The CT machine is ProspeedAI horizontal spiral scanning machine produced by GE corporation. The resolutions of space and density of
145
the CT machine are 0.38 mm and 0.3% (3Hu), respectively. 2.2 Test procedure
3.1 The characters of loess initially From Fig. 1, the exist damages of loess were obvious including lots of cavities, cracks, ginger nut, etc. The differences of ME (between 1116.23 1228.43) and SD (94.19 248.03) were obvious. The two specimens which ME exceeded 1200 had ginger nut of high density. SD of the two specimens were also high. ME and SD of the two sections of the same specimen were quite different. High density zone, low density zone, irregular shape cavities, cracks and ginger nut, etc appeared randomly. Therefore, the initial structures showed anisotropy of intact loess and the influence of ginger nut on ME and SD was obvious. 3.2 The consolidation stage Changes of CT images and CT data (ME and SD) showed variations due to the both effect of suction and net cell pressure. If only considering the effects of net cell pressure, the specimens would be contractive and homogeneous, corresponding to the increasing ME and the decreasing SD. If only considering the effects of suction, the ME would decrease and SD would increase because a portion of water (ME 0) draining out of specimens were filled up by air (ME 1000). Therefore, the changes of ME and SD reflected the influences of both suction and net cell pressure. For the three specimens (1#, 2# and 3#) with low suction (equal to 100 kPa), ME increased and SD decreased due to dominant effect of net cell pressure. The changes of the ME and SD, especially the latter one were greater with the increase in net cell pressure. The strength and deformation moduls of soils increased with the increase in consolidation net cell pressure. This was explained from the perspective of evolution of meso-structure. At constant suction of 300 kPa, ME decreased when net cell pressure was 50 kPa, while increased when net cell pressure was 100 kPa and 200 kPa. At constant suction of 300 kPa, SD decreased significantly with the increase in net cell pressure. When suction was equal to 450 kPa, ME decreased but SD was almost unchanged because of low net cell pressure
The intact loess used in this study collected from the depth of 16 to 16.2 m at Pucheng electric power plant, Shanxi Province, China. It is calssfied as Q2 loess in China. A series of triaxial shear tests on intact unsaturated loess were conducted using the improved CT-multi-function triaxial apparatus. The evolutions of internal structure during triaxial shear tests were studied by CT scanning. The intact loess specimens were firstly consolidated at constant suction (100 kPa to 450 kPa) and constant net cell pressure (50 kPa to 200 kPa), then were sheared under drained conditions. Two sections (Section 1 and section 2 which located onethird and two-thirds of height for each specimen) of specimens were scanned before consolidation and during shearing. 3 CT IMAGES AND EVOLUTION OF MESO-STRUCTURE
The grayscales of CT images are closely related to corresponding density of soils according to the laws of CT images. The black zone of scanned CT images represents low density part of the soils (i.e. cavities, cracks). The white zone of scanned CT images represents high density part of the soils (i.e. ginger nut). Six to seven scannings were conducted during tests for each specimen. Each scanning, for one section, CT-number (denoted by ME) and standard deviation of CT-number (denoted by SD) were obtained. ME reflects mean density of all matter of scanning section and SD reflects the variation of density. The bigger ME represents higher density part of specimens and bigger SD represents greater variations of density of specimens. The averages of ME and SD for two sections of each scanning are shown in Table 1. Fig. 1 show scanned CT images during testing of specimen 6#.
Table 1.
CT scanned ME and SD of intact loess during testing. 1# 2# SD 248.0 242.6 187.3 193.2 174.5 184.1 ME 1161.7 1178.1 1229.3 1284.5 1327.5 1369.4 1406.5 SD 147.1 135.7 125.6 116.8 112.2 100.4 93.3 3# ME 1127.2 1304.8 1399.5 1466.0 1526.6 1579.4 SD 157.5 141.8 129.2 118.7 111.6 101.9 4# ME 1218.8 1159.6 1187.3 1217.2 1234.6 1239.2 SD 197.0 193.2 195.3 179.0 161.1 158.9 5# ME 1116.2 1203.3 1243.7 1296.6 1335.5 1370.3 SD 94.2 74.1 76.5 74.6 59.3 53.3 6# ME 1180.2 1187.4 1265.3 1307.5 1359.2 1406.0 SD 143.4 103.9 109.6 97.6 86.8 82.5 7# ME 1182.2 1142.3 1163.4 1187.1 1201.1 1215.6 SD 131.8 131.6 126.2 115.9 108.8 97.7
Scanning ME 1st 2nd 3rd 4th 5th 6th 7th 1228.4 1256.2 1291.0 1324.5 1352.2 1383.8
146
Figure 1.
Scanned CT images of specimens 6# during testing.

1600 1500 1400 ME 1300 1200 1100 1# 2# 3# 4# 5# 6# 7#
of 50 kPa. From consolidation stage, one could see that ME was properly used to describe the evolution of structure during drained triaxial shear tests at constant suction and constant net cell pressure. 3.3 Yield stage
The 2nd to 3rd scanning was taken as the yield stage which the deviatoric stress increased rapidly and the deviatoric strain increased slowly. In this stage, the changes of CT images werent obvious. The density of soils increased and the difference decreased. Fig. 2 shows the relationship between CT data (ME and SD) and deviatoric strain. Fig. 3 shows the relationship between CT data (ME and SD) and deviatoric stress. Based on Fig. 2 and Fig. 3, that the stress increased rapidly and the strain increased slowly before yield were explained from greater slope of ME versus deviatoric strain and small slope of ME versus deviatoric stress. SD decreased generally with the increase in deviatoric strain and stress. Therefore, ME was also properly used to reflect the changes of structure in the yield stage. 3.4 Harding stage
10 Deviatoric strain, S/%
15
20
250
200
1# 2# 3# 4# 5# 6# 7#
150
SD
100
50
10 Deviatoric strain, S/%
15
20
The 3rd to last scanning was taken as harding stage in which the deviatoric stress increased slowly while the deviatoric strain increased rapidly. In this stage, the density of soils increased and the variations of the density decreased corresponding to the increasing ME and the decreasing SD. From CT images, the flaws and cracks were compressed. Most pores were destroyed and compressed, even disappeared; but small portion of pores which distributed isolated area (such as specimens 1# and 6#) or the around of high density zone such as ginger nut (such as specimen 7#) almost unchanged. The sectional area increased gradually during test. From Fig. 2 and Fig. 3, the changes of ME and SD versus deviatoric strain and deviatoric stress in the
Figure 2. ME and SD versus deviatoric strain curves.
harding stage were different from those in the yield stage. Yield stress could be determined according to the difference of ME versus deviatoric stress or deviatoric strain, especially the former one in the yield and harding stages. ME versus deviatoric stress curves can be appoximated by two straight lines and the intersection of the two lines can be taken as the yield stress. 3.5 The characters of meso-structure deformation
The flaws, cracks and pores were compressed gradually in the consolidation and shear stages during
147
1600 1500 1400 ME 1300 1200 1100 0
1# 2# 3# 4# 5# 6# 7#
100
200 300 Deviatoric stress, (13) / kPa
400
500
250
200
SD
150
1# 2# 3# 4# 5# 6# 7#
100
50 0 100 200 300 Deviatoric stress, (13) / kPa 400 500
The structure of loess weaken gradually with the increase in content of water. The process of deformation caused by load indicated that initial structure was weakened and a new structure was produced. If the strength of the new structure were greater than the strength of initial structure, the soils showed harding behaviour, otherwise softing. The initial structure of loess would be changed under loading and wetting. The evolution of structure included not only the damage of initial structure, but also the formation of the new structure. Therefore, an accurate equation on the evolution of structure was difficult to obtained theoretically or experimentally because of complexity of structure evolution. However, CT data could reasonably reflect structure of soils because CT data may reflect the change of density and water within soils. The intact loess were taken as the relatively intact state (RI) with CT-number MEi and standard deviation of CT-number SDi. The loess after shear were taken as the fully adjusted state (FA) with CT-number MEf and standard deviation of CT-number SDf. A structure parameter based on CT data including CT-number and standard deviation of CT-number can be defined as m and m SD f SD SD f SDi (2) ME f ME ME f MEi (1)
Figure 3.
ME and SD versus deviatoric stress curves.
loading for specimens. In the consolidation stage and shear initial stage, the small pores and cracks were compressed initially, then the bigger pores are destroyed and compressed, even disappeared; but small portion of pores changed little. ME and SD for high density and low density zone of loess tended to same value during shear test. This reflected uniformalization of soil structure during evolution. 4 QUANTITIVE STUDY OF EVOLUTION OF MESO-STRUCTURE
CT data including ME and SD could quantitively determine meso-structure characteristic of the specimens. The soils were gradually compressed with increasing the ME and the decreasing SD. From Table 1, the different initial ME and SD of specimens indicated the different interior structure of intact loess. The meso-structure of soils are gradually compressed with the increasing ME and the decreasing SD during shearing. However, CT images could not reflect the changes of ME and SD with strain for specimens at different net cell pressures and suctions well. 4.1 A structure parameter based on CT data
where ME and SD are CT-number and standard deviation of CT-number during shearing, respectively. The structure parameter m based on CT data is a relative value whose magnitude depends on the selected standard of RI and FA. Precision of the parameter m depends on quantaties of specimens. If the differences of structure of RI and FA are small, the greater structure parameter is obtained. For the RI soils, ME MEi, SD SDi, m 1; but for the FA soils, ME MEf, SD SDf, m 0. CT-number was properly used to describe the evolution of structure during drained triaxial shear tests for intact loess. Therefore, equation (1) was used to analyse the evolution of structue of loess. The different selected RI and FA had influence on the value of the structure parameter m, but no influence on the evolution of the structure. 4.2 An evolution variable of structure
Liu (1997) indicated the structure of loess were affected by two main factors: loading and water.
The initial structure parameter m was determined by ME of specimens. Table 1 showed that ME of
148
Fitted D
specimens were affected significantly by the content of water. To describe the evolution of structure during shearing well, an evolution variable of structure is defined as m m D m0 (1 m0 ) 0 m0 (3)
1.00
0.75
0.50
0.25
where m0 and m are the structure parameters at the initial state and at some portion of during shearing, respectively. The evolution variable of structure based on equation (3) corresponded to the variable ranging from m0 to 1. For simplicity, equation (3) also can be written as D m0 m m0 (4)
0.00 0.00 0.25 0.50 Calculated D 0.75 1.00
Figure 5. Fitted D versus calculated D curves.
Equation (4) corresponds to the evolution variable of structure ranging from 0 (RI) to 1 (FA). 4.3 An equation on the evolution of structure
tion. The deviatoric strain and volumetric strain are used to define the equation on the evolution of structure during shearing because the deviatoric strain and volumetric strain are main factors influencing of structure evolution. An exponential function is used to fitting the evolution equation of structure from Fig. 4 D 1 exp[ ( A1 s A2v )] (5)
It was assumed that the specimen after consolidation was taken as initial state corresponding to the evolution variable of structure equaling to 0. Fig. 4 shows the relationships between the evolution variable of structure D and strain. The evolution varable of structure D increased with the increase in net cell pressure and the decrease in suc0.75 1# 2# 3# 4# 5# 6# 7#
where A1 and A2 are both parameters which are function of net cell pressure and suction. A1 and A2 can be denoted according to the realtionship between D and strain a1 A1 e ( s pat ) pat A2 a2 e ( s pat ) pat
3 ua pat
(6)
3 ua pat
(7)
0.50
D
0.25
where a1 and a2 are both constants which are fitted as 1.68 and 0.12, respectively. Substituting equations (6) and (7) into equation (5) yields
0 5 10 Deviatoric strain, S/% 15 20
0.00
0.75
0.50 D
1# 2# 3# 4# 5# 6# 7#
3 ua 1 D 1 exp e pat (a1 s a2v ) (8) ( s pat ) pat
0.25
Equation (8) represents the relationship among the evolution variable of structure D, deviatoric strain and volumetric strain. The equation reflects the influences of net cell pressure and suction on evolution of meso-structure. The fitted D versus the calculated D curve is shown in Fig. 5. From Fig. 5, D was fitted well.
4 6 8 10
Volumetric strain, V/%
0.00
CONCLUSIONS
Figure 4. The evolution variable of structure D versus strain curves.
(1) Every section of the unsaturated intact loess tended to be homogeneous during shearing because
149
the cracks and cavities were reduced gradually and finally disappeared. During loading, ME increased gradually, while SD decreased gradually. A new method was proposed to identify the yield stress of intact loess using curves of CT-number versus deviatoric stress. (2) CT images can only qualitatively describe evolution behavior of meso-structure, but the CT data can quantitively determine meso-structure characteristic of the specimens. Data analyzed from captured images were used to define a structure parameter and an evolution variable which characterized the evolution of meso-structure. An equation on the evolution of structure was developed. In the equation, deviatoric strain and volumetric strain were included. The equation reflected the influences of net cell pressure and suction on evolution of meso-structure. The CT technique makes the quantitative study of meso-structure of soils possible. The technique also provides an experimental basis to establish the equation on the evolution of meso-structure and structure constitutive model of soils. ACKNOWLEDGEMENTS The financial support received from the Natural Science Foundation Project of ChongQing through grant CSTC, 2009BB6174 is gratefully acknowledged.
REFERENCES
Chen Zheng-han, Lu Zai-hua, Pu Yi-bin. 2001. The matching of computerized tomography with triaxial test apparatus for unsaturated soil. Chinese Journal of Geotechnical Engineering 23(4): 387392. Chen Zheng-han, Sun Shu-guo, Fang Xiang-wei, et al. 2007. Development and application of multi-function triaxial apparatus for soil. Journal of Logistical Engineering University 23(4): 15. Desrues J, Chambon R, Mokni M, et al. 1996. Void ratio evolution inside shear bands in traxial sand specimens studied by computed tomgraphy. Geotechnique 46(3): 529546. Fang Xiang-wei. 2008. Study on microstructure, mesostructure and mechanics properties of Q2 loess. Doctors Thesis of Logistical Engineering University. Finno RJ, Harris WW, Mooney MA, et al. 1997. Shear bands in plane strain compression of loose sand. Geotechnique 47(1): 149165. Liu Zu-dian. 1997. Mechanics and engineering of loess. Xian: Shangxi Science and Technology Press. Otani J, Mukunoki T, Obara Y. 2000. Application of X-ray CT method for characterization of failure in soils. Soils and Foundations 40(2): 111118.
150
A model for capillary pressure in unsaturated granular soils within the pendular saturation regime
J.L. Justo & M. Vzquez
Department of Continuum Mechanics, University of Seville, Spain
ABSTRACT: The presence of water in the intergranular voids of partially saturated soils generates a capillary pressure between the grains. This pressure becomes zero for completely dry or saturated conditions. A model for capillary pressure in unsaturated granular soils within the pendular saturation regime has been presented. The model assumes spherical granular surfaces, zero contact angle, grain size 2.7 mm and dominant capillarity (Bond number, Bo 1). The model has been developed for packing of spheres with e 0.35 and e 0.91. The model and experimental results found in the literature have been compared. 1 INTRODUCTION 2 2.1 MODEL Macroscopic point of view
Partly saturated soils are three-phase systems, with solid, liquid an gas phases. The knowledge of the interaction between the three phases and the developed soil packing are basic points to understand its behaviour. This packing has an influence in the stress states associated to the contact between particles, air and water, and controls suction capillary component. When saturation degree is low, the voids occupied by air are communicated between themselves and with the outside. This is the situation in soils near the surface and in granular compacted fills, with grain sizes that range from fine sand up to rockfill. The state corresponding to a saturation degree that generates discontinuous water and continuous air phases correspond to a pendular saturation regime. Matrix suction depends upon the water surface tension and the meniscus curvature radio. When saturation degree decreases, meniscus shrinks in small voids where the curvature radius is reduced. This way, suction increases. Owing to the pore size, matrix suction is higher in clayey tan in granular soils. Within the pendular saturation regime, matrix suction generates contact capillary pressures between particles that provide the soil with an apparent cohesion. This apparent cohesion is the cause of the vertical slopes that appear in some granular soils as loess. Experimental results suggest that the shear strength and volume change of partially saturated soils depend upon matrix suction rather than upon total suction.
Some modifications of Terzaghis law have been proposed to find an expression for effective stress in partly saturated soils. The effect of suction in the increase of effective stress appears in the expression proposed by Bishop (1959):
( ua ) (ua uw )
(1)
In this equation, parameter would be an indicator of the increase of effective stress under the effect of matrix suction. This parameter has been the objective of important investigations, and has been obtained experimentally or by expressions that relate and generalise the experimental results (Bishop et al. 1963, Frendlund et al. 1978, Escario & Sez 1986). In a first approximation, it may be considered that parameter varies in a way similar to the degree of saturation, so that 0 for a completely dry soil and reaches a value 1 for a saturated soil. Many contradictions have been found in the use of equation (1); for example to explain the volume change in collapse, should be negative. Actually, volume change is a complex function of ua and suction and it is not possible to find a single expression for effective stress in partly saturated soils (Jimnez Salas & Justo 1975). 2.2 Microscopic point of view
The relationship between the amount of water in the voids and the increase in effective stress produced by the presence of air-water interphases,
151
Figure 1.
The two packings analysed.
corresponds to an intimate interaction at microscopic level. At this level it is found that the increase in effective stress may be due to several causes, among them the capillary forces generated in menisci, solute suction produced by salts dissolved in the fluid or electrostatic and chemical forces of attraction between particles. In granular soils, the pressure due to capillarity (matrix suction) dominates the other forces indicated above. The model analysed in the present paper corresponds to the contact sphere-sphere in certain packing conditions. It is assumed that the contact angle meniscus-sphere is zero (clean particle surface), that the influence of gravity is negligible (dominant capillarity, Bo 1) and that the saturation regime is pendular. Initially two working configurations have been adopted, one associated with a simple cubic packing (SC; e 0.91) and the other with a cubic packing centred in the body (CCB; e 0.35). The condition of pendular saturation regime implies a range of saturation degree where menisci are independent within every packing configuration, with no contact between adjacent menisci. The continuous phase is gas and the discontinuous one is water. 2.3 Non-dimensional suction
Figure 2.
Model.
Following the notation indicated in the Figure, Laplace-Young equation takes the form: P 1 1 s y (3)
In this expression and y are the curvature radii of the interphase surface in planes xy and yz respectively. The curvature radius of a curve in the xy plane may be expressed as a function of his derivatives: 1
d2y dx 2
1

3 dy 2 2 dx
1
d 2Y dX 2

3 dY 2 2 dX
1 R
(4)
In this expression, X and Y are the non-linear variables, with respect to the radius of the sphere that represents the soil grain. Substituting equation (4) into expression (3) and operating: P R C s
d 2Y dX 2
Laplace & Young, at the beginning of the nineteenth century, formulated the equation regulating the interphase between two non-miscible fluids: 1 1 P s r1 r2 (2)
1

3 dY 2 2 dX
1 Y
(5)
This equation relates the pressure increase with the curvature radii at the interphase (r1, r2) and the surface tension (s). The differential equation that rules the interphase between two non-miscible fluids at the contact between two equal size spheres has been solved. As indicated above, dominant capillarity (Bo 1) and zero contact angles (clean and regular particle surface) have been assumed. Figure 2 represents the described model.
This is a non-dimensional expression of suction at the interphase. Operating in this equation:

2 1 d 2Y dY 2 1 C 2 dX Y dX 3
(6)
This is a non-linear, second order, differential equation. X and Y are the non-linear variables that define the interphase surface. The solution must be found using numerical methods.
152
The following expression is obtained using centred finite differences: 1 Yi 1 2Yi Yi 1 X 2 1 Yi Yi 1 X 1 C (7) Y
i 2 3/ 2
DIMENSIONLESS SUCTION 180 160 140 120 C 100 80 60 40 20 0 0 0,2 0,4 r/R 0,6 0,8
Justo and Vzquez (2010) Cho and Santamarina (2001)
where X is the variable increment in each iterative path and C is the non-dimensional suction defined by equation (5). To solve this equation using an iterative procedure, two initial values are needed. They are obtained knowing that at point 0 (Fig. 2) the meniscus is in contact with the sphere and the contact angle is zero. These values are a function of the meniscus neck radius, r, indicated in Figure 2, and are given by the following expressions:
Figure 3. Dimensionless suction as a function of nondimensional meniscus neck radio.
SATURATION DEGREE THRESHOLD
2( R r ) X0 2 R R r 2( R r )
2 Y0 2X 0 X 0
r2
(8) (9) (10)
1 X0 Y1 X Y0 Y0
Parameter C is a priori unknown. For this reason, it is necessary to solve the iterative procedure (7) for different C values. The solution will correspond to the C value that will give a meniscus curve cutting the Y-axis at right angle. The iterative process has been solved for several values of r/R up to the value corresponding to zero matrix suction. The numerical values have been approximated by an expression of the type (abx)/x2. This has been adjusted by a minimisation algorithm proposed by Nelder-Mead, and formulated using geometrical elements (smplex), implemented in the library of MATLAB. The following approximation is obtained: C (r/R ) 1.96 2.83 (r/R ) (r/R )2 (11)
The model is valid for a pendular saturation regime up to the moment when the menisci start to contact between themselves, giving continuity to the water phase and turning to a funicular saturation regime. Once the basic geometry has been defined and the profile of the interphase curve has been obtained, the water volume enclosed by the menisci for every r/R value and the degree of saturation when the menisci contact may be determined. The pendular saturation threshold has been determined for the two packings included in Figure 1. 3.1 Packing SC (e 0.91) For packing SC (Fig. 1a; e 0.91) the unit cell has a projection area equal to 4R2 and a volume 4R3. Table 1 presents the volume of voids Vh, water Vw and degrees of saturation for a sphere radius R 0.002 mm. From Fig. 4, the boundary value of r/R when the menisci start to contact may be obtained. Table 1 ends with the degree of saturation (18.56%) corresponding to the boundary between pendular and funicular regimes for packing SC. 3.2 Packing CCB (e 0.35)
This approximation gives the non-dimensional suction as a function of the non-dimensional variable r/R (Fig. 3). These non-dimensional conditions make the expression valid for any grain size. Starting from a hypothesis of toroidal interphase surface, Cho & Santamarina (2001) arrive to the expression: C ( ) 2 3 2 (12)
For packing CCB (e 0.35) and the same unit cell considered in the SC case, the values indicated in Table 2 are obtained. From Fig. 5, the limit value of r/R when the menisci start to contact may be obtained. Table 2 ends with the degree of saturation (24.86%) corresponding to the boundary between pendular and funicular regimes for packing CCB.
153
Table 1. r/R 0.10 0.13 0.20 0.27 0.30 0.35 0.40 0.45 0.50 0.55 0.59
Packing SC (e 0.91). Vh (mm3) 108 1.52 1.52 1.52 1.52 1.52 1.52 1.52 1.52 1.52 1.52 1.52 Vw (mm3) 1010 0.02 0.06 0.33 1.14 1.76 3.33 5.78 9.42 14.6 21.7 28.3 Sr (%) 0.01 0.04 0.22 0.75 1.16 2.18 3.79 6.18 9.56 14.2 18.6
Figure 5. Packing CCB: limiting menisci for pendular regime in a section through the centre of the spheres.
Figure 6. Sr(%) vs. .
Sr 1.551 4 Sr 7.96 4
Figure 4. Section through the centre of the spheres, in packing SC, showing the limit of menisci for pendular regime. Table 2. r/R 0.10 0.13 0.20 0.27 0.30 0.35 0.40 0.425 Packing CCB (e 0.35). Vh (mm3) 109 8.32 8.32 8.32 8.32 8.32 8.32 8.32 8.32 Vw (mm3) 1010 0.07 0.19 1.11 3.81 5.87 10.8 17.5 20.7 Sr (%) 0.08 0.23 1.34 4.58 7.06 12.9 21.0 24.9
(13) (14)
Assuming toroidal meniscus surfaces, other authors (Cho & Santamarina 2001) present similar expressions for the same packings. Figure 6 compares both results. Near the threshold of the pendular saturation regime, the toroidal hypothesis starts to split up from the results obtained in this paper. When the degree of saturation is low, the meniscus surface approaches a toroid. 4 CAPILLARY PRESSURES IN THE INTERGRANULAR CONTACT
3.3
Comparison of results
Figure 7 shows the contact between the two spheres that model the soil grains. Two force types appear due to the presence of the air-water interphase in the meniscus. One force is associated to the capillary adhesion and the other to matrix suction. The contact force may be found, establishing the equilibrium with the two cited vectors: Fcont Fts Fsuc (15)
Using a potential approximation to the results obtained, the following expressions relate the degree of saturation with variable r/R, for packings SC (equation 13) and CCB (equation 14).
where Fcont is the force generated at the contact between particles, Fts is the force due to capillary
154
Using the approximated expressions (13) y (14), the capillary pressures for every packing considered are obtained, within the pendular saturation degree, as a function of the degree of saturation, surface tension and grain size. For packing SC:
cont
Figure 7. Contact model.
s 0, 0123 Sr 0,1013 4 Sr R 2

1, 0353 0, 0540 4 Sr (23)
adhesion and Fsuc corresponds to the force generated by matrix suction. Working with the vertical component of the forces: Fcont s 2 sin (ua uw) Ac (16)
For packing CCB:
cont
s 2 0, 0054 Sr 0, 0673 4 Sr R

1, 0353 0, 0813 4 Sr (24)
where Ac is the hatched spherical cap area indicated in the Figure. Developing expression (16): Fcont Y2 s 2 mx (ua uw ) 2 R X mx R (17)
The meniscus geometrical parameters may be substituted by the expressions: Ymx 1,2271 R 0,0179 R (18) (19) X mx 0,5862 3 R 0,5168 2 R 0,0078 R
Equations (21) and (22) are represented in Figures 8 and 9, where capillary pressure tends to zero, as the saturation degree tends to zero, according to what is actually observed. The models developed by Cho & Santamarina (2001) and Tselishchev & Valtsiter (2003) for similar packings, do not predict this fall in capillary pressure, but a linear increase as the water in menisci decreases.
Establishing a unit cell in every packing with an area 4R2 and volume 4R3, the contact capillary pressure will be:
cont
Fcont 4R 2
(20)
Substituting equations (18) and (19) into (17), and using expressions (5) and (12), the following non-dimensional formulations are obtained for every packing considered: [( cont )SC ] ( cont )SC R s
Figure 8. ing SC.
Non-dimensional capillary pressure in pack-
0,0153 2 0,1531 0,3575 1,0353 2 ( cont )CCB R 2 s (21)
[( cont )CCB ]
0,0153 0,1531 2 0,3575 1,0353 (22)
Figure 9. Non-dimensional capillary pressure in packing CCB.
155
CONCLUSIONS
REFERENCES
Baltodano, R. 2006. Tensile strength, shear strength and effective stress for unsaturated sand. PhD Thesis. University of Missouri. Columbia. Bishop, A.W. 1959. The principle of effective stress. Tek. Ukebl. 39: 859863. Bishop, A.W. & Blight, G.E. 1963. Some aspects of effective stress in saturated and partly saturated soils. Gotechnique 13: 177197. Cho, G.C. & Santamarina, J.C. 2001. Unsaturated particulate materials. Particle-level studies. Journal of Geotechnical and Geoenvironmental Engineering. 8496. Escario, V. & Saez, S. 1986. The shear strength of partly saturated soils. Gotechnique 36(3): 453456. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. Shear strength of unsaturated soils. Canadian Geotechnical Journal 15(3): 313321. Jimnez Salas, J.A. & Justo, J.L. 1975. Geotecnia y Cimientos I. Propiedades de los Suelos y de las Rocas. Rueda, Madrid. Matsushi, Y. & Matsukura, Y. 2006. Cohesion of unsaturated residual soils as a function of volumetric water content. Bull. Eng. Geol. Env. 65: 449455. Tselishchev, Y.G. & Valtsifer, V.A. 2003. Influence of the type of contact between particles joined by a liquid bridge on the capillary cohesive forces. Colloid Journal 65(3): 385389.
The curve generated in the air-water meniscus associated to two equal-size spheres in contact has been solved numerically and presented in non-dimensional form. The hypotheses of zero contact angle and dominant capillarity have been assumed. As the saturation degree increases, the meniscus shape splits up more and more from a toroidal shape. Once the meniscus shape is known, the saturation degree at the threshold between pendular and funicular regimes has been obtained for a compact packing (e 0.35) and a loose packing (e 0.91). The resulting degrees of saturation are 24.86% for the compact packing and 18.56% for the loose packing. The model has allowed finding the nondimensional contact capillary pressure, (cont R/ s) for the two indicated packings as a function of saturation degree. The parameters R (sphere radius), s (surface tension) and saturation degree are easily obtainable by laboratory tests. This model reproduces only forces related to capillary effects, and not to chemical bonds or electrostatic forces. The sphere must be smooth, far from the irregular shape associated to some soil grains.
156
An insight into role of clay-fluid molecular interactions on the microstructure and macroscale properties of swelling clays
D.R. Katti, K.S. Katti, P.M. Amarasinghe & S.M. Pradhan
North Dakota State University, Fargo, North Dakota, USA
ABSTRACT: Swelling clays are found in various parts of the world and are responsible for damage to infrastructure and light structures that amount to billions of dollars each year. These clays also have beneficial uses in environmental engineering as clay liners, in petroleum industry as drilling mud, in industry as nano-inclusions in nanocomposites and in pharmaceutical and cosmetics industries. Molecular modeling studies were conducted to calculate interaction energies that describe the mechanism of swelling and their role on maintaining the stability of clay structure. Experimental studies were conducted with clay intercalated with fluids exhibiting a range of polarities from highly polar to non polar. The multiscale approach and results described in this paper open a new methodology to connect molecular scale phenomenon to macroscale properties of clays and would lead to powerful new tools for accurately predicting behavior of clays.
INTRODUCTION
The fundamental study of clay-fluid interactions and their role on clay microstructure and macroscale properties is of considerable importance in the areas of geotechnical engineering, geoenvironmental engineering, materials science, pharmaceutical sciences etc. Development of robust generalized constitutive models to accurately predict the behavior of clays for a wide variety of initial states, fluid types and stress conditions are limited by lack of significant understanding of the nature and role of molecular interactions between clay and fluids that likely influence their properties. This difficulty is further compounded in swelling clays, where highly charged surfaces and interlayer swelling response is significantly influenced by molecular interactions. Advances in computational methods, computational hardware and software and experimental techniques provide researchers with new capabilities to explore the problem in a scientific manner. In this paper, we present results of our modeling and experimental studies at various length scales, undertaken to provide an understanding of how molecular interactions between swelling clay and fluids affect the mechanical and flow characteristics in swelling clays. The first part of this paper deals with results from molecular dynamics simulations and the second part describes results from experimental studies on swelling clay with fluids having a range of polarities from non-polar to highly polar. Our previous experimental work (Katti and Shanmugasundaram 2001) has shown the breakdown of clay particles with increased swelling.
Subsequent discrete element modeling of swelling clay with the incorporation of particle breakdown has shown that increased particle breakdown contributes to increased swelling and swelling pressure (Katti et al. 2009). The origin of particle breakdown likely stems from the swelling of the interlayer that results in exfoliation. The interlayer swelling appears to be closely related to the clayfluid interactions. 2 MOLECULAR MODELING
Molecular dynamics (MD) simulations were conducted to evaluate the interactions between the clay-water-cations in the interlayer. These simulations were conducted using MD software NAMD (Phillips et al. 2005) and visualization and analysis software VMD (Humphrey et al. 1996). Molecular dynamics is a computational method which calculates the time dependent behavior of a molecular system. MD is based on Newtons second law or the equation of motion and provides a trajectory which specifies the variation of position and velocity of individual atoms in a molecular system with time. Individual atoms are characterized as balls with bonds represented by springs. A variety of springs are introduced that capture stretching, angular rotation and torsion. Non-bonded interactions are modeled as van-der-Waals and electrostatic. The potential energy of the system is calculated from a force-field and is used in calculating the trajectory of atoms in the molecular system. In the current work, we have used the
157
CHARMm forcefield (Brooks et al. 1983) which is defined as, E potential

bond
k B (r r0 )2 k A ( 0 )2
ij 4 ij r vanderWaals i j ij qi q j
electrostatic i j dihedral
V D [1 cos( n )]
12 6 ij rij
angle

rij
where kB, kA and VD are the force constants for bond stretching, angle bending and dihedral angle bending. The reference values are r0 and 0 is the phase angle of the dihedral cosine function, and n is the multiplicity. ij and ij are the standard Lennard-Jones potential parameters. The distance between two nonbonded atoms, and the respective charges, is rij, qi and qj. Molecular models of Na-montmorillonite expansive clay mineral (Figure 1) were constructed and CHARMm forcefield parameters for the clay were obtained (Katti et al. 2005; 2007; Schmidt et al. 2005). These parameters were verified by comparing crystal structure of the clay with experimentally obtained structures and other simulations. The accuracy of the clay-water model was verified by comparing the d-spacing for various magnitudes of hydration obtained experimentally. Steered molecular dynamics simulations were conducted to evaluate the mechanical behavior of clay interlayer for various levels of hydration(Katti et al. 2007; Schmidt et al. 2005). All simulations were conducted at atmospheric pressure and room temperature (300 K) conditions. Molecular dynamics simulations were conducted to study the evolution of interaction energies between the clay, water and Na-ions in the interlayer. The molecular model of the clay is
based on our previous work. For the model system, the formula Na0.5Si8[Al3Fe(III)0.5Mg0.5]O20(OH)4 is used. Interaction energies are calculated between the clay sheets, clay sheets and Na ions, Na bound water and clay bound water, clay and Na bound water (Figure 2). Bound water is assumed to be the water molecules within the hydrogen bond distance of 3.5 of the Na atoms or clay molecules. The interaction energies are calculated after introducing layers of water in the interlayer and after the system is minimized and the molecular system is brought to atmospheric pressure and room temperature conditions. The results from simulations showing interaction energies versus the number of water layers and corresponding relationship with interlayer spacing is shown in Figure 3. Each water layer comprises of 32 water molecules in the interlayer. The negative energy values are attractive. The plot in figure 3 shows that with increase in hydration, the interlayer spacing or swelling increases. The interaction energy between clay sheets and the interaction energy between Na and clay are attractive and hold the clay sheets together in the dry state. With increase in water layers, the clay-clay interactions that are primarily van-der-Waals in nature decrease very rapidly and decrease to practically negligible values. The attractive interactions between Na and clay continue to be quite significant even up to 8 layers of water. Beyond about two water
Figure 2. Schematic diagram showing various interaction energies computed.
Figure 1. Molecular model of Na-montmorillonite with three layers of water (initial condition).
Figure 3. Plot of interaction energies versus number of water layers in the interlayer.
158
layers, water also contributes to maintaining the clay structure through attractive interactions between Na bound water and clay bound water. These simulations provide the values of binding energies that keep the clay sheets together. Exfoliation would occur when the binding energy is overcome. 3 EXPERIMENTAL INVESTIGATION OF CLAY-FLUID INTERACTIONS
Experimental investigation of interactions between the Na-montmorillonite clay and fluids with a range of dielectric constants was carried out to investigate the role of these interactions on swelling, swelling pressure, fluid flow and compressibility characteristics of these clays. Na-montmorillonite (swy-2) was obtained from the repository of the clay mineralogy society of America. Seven organic fluids with varying dielectric constants (provided in parenthesis) were used in the study. These are: formamide (110), water (80), methanol (33), acetone (20), chloroform (4.8), TCE (3.4) and toluene (2.4). The high dielectric constant fluids are classified as high polar fluids and the low dielectric constant fluids are termed as low polar or non polar fluids. Fourier transform infrared (FTIR) spectroscopy experiments and x-ray diffraction experiments on clay slurries with various fluids were conducted. FTIR experiments involve impinging infrared radiation on a sample. This energy input results in vibration of the various bonds in the molecular system. The infrared spectrum generated provides a unique molecular signature of the specimen. This spectrum provides information on the nature of bonds and shift in band positions are indicative of influence of new or additional interactions. Detailed information about this study is presented in our previous work (Amarasinghe et al. 2009). This study was also used to experimentally calculate the rate of flow of water molecules in the clay interlayer (Amarasinghe et al. 2008). FTIR study relating silica-water interactions to swelling is reported earlier (Katti and Katti 2006).The results from these studies show that fluids with high dielectric constants have very large interactions with clays and low dielectric constant fluids have low or no molecular interactions with clays. Most swelling and exfoliation occurs with high dielectric constant fluids and little or no swelling occurs with low dielectric constant fluids. Further, swelling pressure under no-volume-change condition, was observed to increase with increasing dielectric constant of the fluid. In order to accurately measure the permeability characteristics of swelling clays, a new permeability device was designed and fabricated. This device which is called as a porous rigid wall flexible wall permeameter allows for conducting flexible wall permeability tests on swelling clays without the
sample or clay microstructure getting distorted by lateral swelling. A photograph of the device is shown in Figure 4. A plot of the coefficient of permeability versus dielectric constant obtained from this device for dry compacted sample at the same density is presented in Figure 5. The results show that dielectric constant of the fluid has a profound effect on permeability characteristics of the clay with low dielectric constant fluids exhibiting almost a million times higher coefficient of permeability values as compared to water. The samples saturated with various fluids were subject to a range of confining stresses and the samples are allowed to consolidate in order to evaluate compressibility and consolidation characteristics. A plot of void ratio versus effective stress for samples saturated with various fluids is presented in Figure 6. For a given effective stress, clay intercalated with a fluid of low dielectric constant, shows higher compression and its behavior is more like that of cohesionless soils. Hence, weak molecular interactions
Figure 4. Photograph of a new device to measure permeability of swelling clays.
Figure 5. Plot of coefficient of permeability of clay sample with dielectric constant of the fluid.
159
macroscale properties of swelling clays is presented and demonstrates the need for incorporating molecular interactions in modeling clays in general and swelling clays in particular. ACKNOWLEDGEMENTS Support from National Science Foundation Grant #0556020, 0114622-Dr. Richard Fragaszy, program director and 0320657(MRI) and 0315513(IMR). REFERENCES
Amarasinghe, P.M, Katti, K.S, and Katti, D.R. (2008). Molecular Hydraulic Property of Montmorillonite: A Polarized FTIR Spectroscopic Study. to appear inApplied Spectroscopy. Amarasinghe, P.M, Katti, K.S, and Katti, D.R. (2009). Nature of organic fluid-montmorillonite interactions: An FTIR spectroscopic study. Journal of Colloid and Interface Science, 337(1), 97105. Brooks, B.R, Bruccoleri, R.E, Olafson, B.D, States, D.J, Swaminathan, S., and Karplus, M. (1983). CHARMMA PROGRAM FOR MACROMOLECULAR ENERGY, MINIMIZATION, AND DYNAMICS CALCULATIONS. Journal of Computational Chemistry, 4(2), 187217. Humphrey, W., Dalke, A., and Schulten, K. (1996). VMD: Visual molecular dynamics. Journal of Molecular Graphics, 14(1), 33-&. Katti, D., and Shanmugasundaram, V. (2001). Influence of swelling on the microstructure of expansive clays. Canadian Geotechnical Journal, 38(1), 175182. Katti, D.R, Matar, M.I, Katti, K.S., and Amarasinghe, P.M. (2009). Multiscale modeling of swelling clays: A computational and experimental approach. Ksce Journal of Civil Engineering, 13(4), 243255. Katti, D.R, Schmidt, S.R, Ghosh, P., and Katti, K.S. (2005). Modeling the response of pyrophyllite interlayer to applied stress using steered molecular dynamics. Clays and Clay Minerals, 53(2), 171178. Katti, D.R, Schmidt, S.R, Ghosh, P., and Katti, K.S. (2007). Molecular modeling of the mechanical behavior and interactions in dry and slightly hydrated sodium montmorillonite interlayer. Canadian Geotechnical Journal, 44(4), 425435. Katti, K.S, and Katti, D.R. (2006). Relationship of swelling and swelling pressure on silica-water interactions in montmorillonite. Langmuir, 22(2), 532537. Phillips, J.C, Braun, R., Wang, W., Gumbart, J., Tajkhorshid, E., Villa, E., Chipot, C., Skeel, R.D, Kale, L., and Schulten, K. (2005). Scalable molecular dynamics with NAMD. Journal of Computational Chemistry, 26(16), 17811802. Schmidt, S.R, Katti, D.R, Ghosh, P. and Katti, K.S. (2005). Evolution of mechanical response of sodium montmorillonite interlayer with increasing hydration by molecular dynamics. Langmuir, 21(17), 80698076.
Figure 6. Plot of void ratio versus effective stress for samples saturated with fluids of high polarity, medium and low polarity.
Figure 7. Scanning electron microscopy images of the clay sample from permeability device after the Namontmorillonite clay is saturated with various fluids.
in the clay-low-dielectric constant fluid system, do not generate swelling forces that would help maintain the soil structure under external stresses. Scanning electron microscopy images were obtained on samples recovered in undisturbed state from the permeameter and are shown in Figure 7. The images reveal that clays with low dielectric constant fluids possess microstructures similar to the dry clay and exhibit no change in particle sizes and show presence of visible voids. However, for a high dielectric constant fluid, the microstructure is remarkably different with significantly smaller particle sizes resulting from particle breakdown and void spaces filled showing the tremendous impact of molecular interactions on microstructure and permeability of the clay. 4 CONCLUSIONS
Evolution of molecular interaction energies and corresponding swelling of Na-montmorillonite interlayer provide a quantitative measure of binding energies in the interlayer. The important role of molecular interactions on the microstructure and
160
Volume changes of swelling clayey soils at microscopic scale level

T. Maison & F. Laouafa J.-M. Fleureau
INERIS, Verneuil-en-Halatte, France Laboratoire MSSMat, UMR CNRS 8579, Ecole Centrale Paris, Chatenay-Malabry, France
ABSTRACT: In order to understand the mechanisms involved in swelling and shrinkage, experimental research was carried out at the microscopic level using Environmental Scanning Electron Microscope (ESEM). This device was used for the observation at a small scale of two natural clays as powder subjected to wetting/drying cycles. The first clay is the Greek montmorillonite, an homogeneous montmorillonite, and the second clay is Romainville clay, an heterogeneous interstratified clay. From image and chemical analysis, their behaviours are analysed and a comparison based on their mineralogical composition is performed. The mineral composition of the clay plays a significant role. We show that the homogeneity of the clay and the presence of carbonates significantly modify the deformation response of the clay to the wetting-drying loading. The different observations also show an evolution which can be decomposed into two distinct and successive steps. The particle size can influence the evolution or the rate of deformation. 1 INTRODUCTION mesoscopic scale of the order of the millimetre unit and the macroscopic scale of the order of the centimetre unit. Some clay features can be obtained at the macroscopic scale. The microscopic experiments usually consist of microscopic observations, often with a Scanning Electron Microscope (SEM). These experiments allow the observation of the microstructure and sometimes of the porosity. However, the samples preparation (freezing, lyophilisation and coating) can disturb in a significant manner the microstructure of the clay. Nowadays, these drawbacks can be overcome with the use of the Environmental Scanning Electron Microscope (ESEM). The control of the hygrometric conditions in the ESEM chamber allows the analysis of the specimen under in situ hygrometric conditions. It does not require any preparation like the SEM observations. Some studies have been carried out on clays with the ESEM (Romero & Simms 2008). The main conclusions are that the macrovoids of the material were filled by swelling bentonite when absorbing water. In order to study the swelling/shrinkage behaviour in clayey soils, ESEM observations were carried out on natural clays (taken in situ). Two clays were studied: an homogeneous natural clay (Greek montmorillonite) (Fleureau et al. 1993, Souli et al. 2008) and an heterogeneous natural clay (Romainville clay). The purpose of this paper is firstly to observe and analyse swelling/shrinkage evolution of the two natural clays with different mineralogical
Global warming, recognized by most of the scientific community, resulted, in the last few years, in volumetric changes of clayey soils in large areas. Such soils have the particularity of dilating (swelling) and contracting (shrinkage) according to the evolution of their water content and mineralogical composition. The consequences are mainly modifications of the physical and chemical properties of the soil which, from a mechanical point of view, can induce non homogeneous deformations and lead to differential displacements causing damage to structures (Mitchell 1976). In this paper, the term clay is only used when speaking specifically of clay minerals or the real clay fraction of the soil. Clays contain various structural scales (aggregates, particles and sheets), which lead to various porosity scales. To understand the relationship or the bridge between the various scales is fundamental to understand the behaviour of the clay (Gens & Alonso 1992). The extrapolation of the behaviour observed at microscopic level to the behaviour at macroscopic level is still an open question. In numerical modelling framework, several constitutive models have been proposed to describe the behaviour of saturated or unsaturated soil (Alonso et al. 1990) and the swelling/shrinkage clay behaviour, but they do not link the microstructure and the macrostructure behaviours together. In this paper, the microscopic scale is considered to be of the order of the micron unit, the
161
composition, and secondly to study the influence of particle size, in the case of the Greek montmorillonite, on the swelling-shrinkage behaviour. The two clays were analysed in powder form (maximum grains size equal to 125 m) of discrete particles. A wetting/drying cycle, with time steps, was applied to the different specimens in the ESEM chamber and their free swelling intensity and hydric-evolution were compared (according to their mineralogical composition). The Greek montmorillonite is mainly composed of smectite, whereas the Romainville clay is composed of non swelling species (kaolinite, illite and carbonates) as well as illite/smectite interlayers (Audiguier et al. 2007). Carbonates are known to prevent swelling with a cementation role. The complex composition of this clay induces a complex swelling behaviour, compared to Greek montmorillonite. A wetting/drying cycle, with time steps, was applied to particles of Greek montmorillonite of different sizes and their free swelling hydric-evolution was compared. 2 MACROSCOPIC CLAYS FEATURES
carbonate, quartz and feldspar. This material was characterized by Audiguier et al. (2007) (Table 2). The swelling intensity for remoulded Romainville clay averages 25%. 3 EXPERIMENTAL PROCEDURE
3.1 Environmental Scanning Electron Microscopy (ESEM): advantages and drawbacks The ESEM is an electronic microscope device which allows observations of samples (soil for instance) under controlled environmental conditions (temperature, pressure, thus hygrometry). Among all the qualities of the ESEM apparatus, a very important one is to maintain the sample unremoulded without any preparation (freezing, lyophilisation and coating as in classical mode). The samples structure and natural water content are then unperturbated. On the other hand, its main drawback is linked to the image quality (contrast and brightness) which strongly depends on the chamber humidity. When dealing with long time wetting-drying cycles, the difficulty is to maintain a constant time step in order to reach a high image definition. The experimental procedure consists in applying regular time steps (varying from 10 to 30 minutes according to the type of clay), in order to let sufficient time for the hydration and dehydration processes to reach stationary state. The rate of wetting is also smoother than the in situ condition and closer to it. The experimental program contains two main phases: during the first phase, the times to reach steady state were chosen as a function of the relative humidity (increments of 10% up to 85% of relative humidity). For the second one, the times to reach steady state were chosen as a function of the pressure (steps of 10 Pa, 5 Pa and at the end 1 Pa) (Fig. 1). With such a process, a good accuracy in hydration of clay particles and in swelling measurements at high relative humidities can be reached. The complete saturation of the observed area is delayed due to the smooth hydration rate. 3.2 Clay powder
The Greek montmorillonite comes from Milos Island, in the Cyclades archipelago, and has a sedimentary origin. As its name indicates, it is composed mainly of smectite, an expansive clay species. The main features of this clay are shown in Table 1. For this clay, about 30% of free swelling were measured on sample with classical geotechnical tests (Fleureau et al. 1993). Romainville clay comes from the East of Paris (France), and has also a sedimentary origin. The clay species present in the specimen are kaolinite, illite and illite/smectite interlayers and it also contains
Table 1. Geotechnical characterisation of Greek montmorillonite (Fleureau et al. 1993, Christidis et al. 1997). Liquid limit, wL,(%) 170 Plastic limit, wP,(%) 60 Plasticity index, Ip 110 Specific surface, (m2/g) 61.76 CEC, meq/100 g 104.4
Table 2. Geotechnical characterisation of Romainville clay (Audiguier et al. 2007). Plastic Liquid limit, wL, limit, wP, Plasticity (%) (%) index, Ip 75 40 35 Specific surface, (m2/g) 195 CEC, meq/100 g 39.5
The Greek montmorillonite and Romainville clay are analysed in powder form and subjected to one wetting/drying cycle. To perform ESEM observations, the powder is put on a double sided carbon tape, for two reasons: (i) to avoid particle displacement during wetting; (ii) to connect the surface to the ground in order to avoid electron overload of the sample, which disrupts the observations. The temperature is kept constant equal to 18C and the pressure is controlled in order to obtain the
162
where Si surface strain at time ti; S0 initial surface of the grain. Under isotropic assumption, the volumetric strain v is equal to: v s where 4 4.1
Figure 1. Relative humidity versus time for the Greek montmorillonite and Romainville clay used in wetting processes.
(2)
EXPERIMENTAL ANALYSIS Results
Figure 2. Deformation definition on the basis of the parameters of time, surface and relative humidity of clay particles.
imposed hygrometry. The initial hygrometry in the ESEM chamber is about 50%. The clay powder is wetted up to 90% with steady steps of different duration according to the type of clay (from 10 to 30 minutes). For greater relative humidities, the water immerses the particles and the observation area. The times to reach steady states are determined for a regular relative humidity rise and fall. At high relative humidities, where the maximum of deformation is reached, the steps are closer. The volume or surface changes are real time followed. The surface deformation induced by swelling and shrinkage is calculated by analyzing ESEM images. The term surface strain is used to define the relative surface change of the particles at time ti (Fig. 2):
s
Si S0 S S0 S0
(1)
The Greek montmorillonite and Romainville clay are compared on the basis of their mineralogical composition. The mean size of particles observed is 50 m. Figures 3 and 4 present the ESEM observations and measurements of surface strain evolution for the wetting/drying experiment performed on the Greek montmorillonite. Between states A and B, the hydration of powder induces swelling until the total saturation of the carbon tape is reached (B). Between states B and C, the dehydration of powder induces shrinkage. At this state, some destructuration can develop, as shown by the particle in the upper part (located by the arrow) of the photograph (Fig. 3). This phenomenon explains the hysteresis between swelling and shrinkage (Fig. 4). The destructuration of the particle involves an increase in the area with respect to the final state of swelling before saturation (state B). The swelling and shrinkage evolution can be decomposed into two sequences (Fig. 4). At each humidity change, the first step is characterised by a discontinuity (jump) in the surface strain (s), followed by constant deformation with time. The swelling intensity measured is 60%. The strain evolves moderately between 50% and 80% of relative humidity, and much more between 80% and 90%: 75% to 85% of the deformation occurs between 80% and 90% of relative humidity. If a zoom is made on very high relative humidities, an average of 40% of strain occurs between 89.4% and 89.9% of relative humidity. This fact seems to confirm the importance of the high relative humidities, and small suctions in the swelling-shrinkage behaviour of clays. The time steps applied for the swelling-shrinkage cycle are of 10 minutes each. Figures 56 present the ESEM observations and measurements of surface strain concerning the wetting/drying experiments performed on the Romainville clay. The states A, B and C are the same as for the Greek montmorillonite. Between states B and C, the dehydration of powder induces shrinkage, but no destructuration (Fig. 5). In this case, it is not this phenomenon which explains the hysteresis between swelling and shrinkage
163
Figure 4. Surface area strain versus relative humidity for Greek montmorillonite. The path AB represents the swelling; the path BC represents the shrinkage.
Figure 3. Greek montmorillonite clay particles evolution versus relative humidity. Three states of the wettingdrying cycle are presented: A) initial state; B) before complete saturation; C) after drying (ESEM images, magnification: 1000).
(Fig. 6). At the beginning of shrinkage and for different steps, the water surrounds the particles and continues to deform. The clay particles boundaries are not visible. This leads to an overestimation of the area, which could explain the gap between the theoretical end of swelling and the theoretical beginning of shrinkage (state B). The swelling and shrinkage evolutions are the same as for the Greek montmorillonite (Fig. 6). The swelling intensity measured is 25%. The evolution and proportions of the strain rate is similar to that of the Greek montmorillonite. The steady steps applied during the swelling-shrinkage cycle are of 30 minutes each one.
Figure 5. Romainville clay particles evolution versus relative humidity. Three states of the wetting-drying cycle are presented: A) initial state; B) before complete saturation; C) after drying (ESEM images, magnification: 1000).
164
Figure 6. Surface area strain versus relative humidity for Romainville clay. The path AB represents the swelling; the path BC represents the shrinkage.
Figure 7. Surface area strain versus relative humidity for Greek montmorillonite small particles. The path AB represents the swelling; the path BC represents the shrinkage.
Particles of 10 m and 50 m size of Greek montmorillonite are compared. Figure 5 present the measurements of the deformation rates of the wetting/drying experiment performed on particles of 10 m. Between states B and C, the dehydration of powder induces the displacement of the particles. This phenomenon explains the hysteresis between swelling and shrinkage (Fig. 7). The displacement of the particle involves a decrease in area in relation to the final state of swelling before saturation (state B). The swelling and shrinkage evolution is the same as for the particles of 50 m. The swelling intensity measured is 15%. The evolution and proportions of deformation rate is similar to that of the particles of 50 m. 4.2 Discussion
The Greek montmorillonite and Romainville clay are compared. For the same size of clay particles, the time needed to reach steady states (10 minutes for the Greek montmorillonite, 30 minutes for the Romainville clay) and the swelling intensity (60% for the Greek montmorillonite, 25% for the Romainville clay) are different. Let us recall that the first one is composed mainly of smectite, and the second one of various components (kaolinite, illite, illite/smectite interlayers, carbonates, quartz, feldspar). Swelling can develop freely in Greek montmorillonite thanks to smectite. Otherwise, swelling can be prevented in Romainville clay due to two parameters: (i) the presence of kaolinite and illite, which are clay species with hard bounds; (ii) carbonates known for constituting cementation. It can be underlined that the swelling intensity measured in ESEM observations is similar to the swelling value measured with classical oedometric free swelling tests (Maison et al. 2009).
The Greek montmorillonite, after a hydration/ dehydration cycle, shows a kind of destructuration of its particle. This destructuration is not observed in the Romainville clay. This phenomenon is perhaps induced by the adsorption of water by the powder. Smectite presents a high absorption rate and can absorb a high quantity of water. Pore water pressure may increase within the set of aggregates until it exerts a stress on the walls of the aggregates and cause their dissociation. The quantitative analyses show that the swelling and shrinkage evolution can be decomposed into two successive steps. The fact that the first step is characterised by a discontinuity (jump) in the deformation may reflect the immediate reactivity of clay to wetting-drying loading. The following strain evolution confirms the importance of time in swelling-shrinkage phenomena. Particles of 10 m and 50 m size of Greek montmorillonite are compared. For the same time to reach steady state, the behaviour is identical. That shows a similar evolution, whatever the size of the observed particle. However, the swelling intensity for particles of 50 m is twice that for particles of 10 m. The ratio for size between normal particles (50 m) and small particles (10 m) is five, whereas the ratio for swelling intensity is two. It would be interesting to lead this test on big particles (100 m for example) in order to confirm or infirm the proportionality characteristic of swelling intensity evolution with particle size. 5 CONCLUSIONS
Swelling-shrinkage cycles were performed in an ESEM in order to quantitatively compare two different swelling clays and the influence of particle
165
size. The observations and measurements show that soil composition influences swelling response to hydration and also swelling evolution. The presence of non expansive components (non expansive clays, carbonates) can be sufficient to reduce the swelling. The evolution can be decomposed into two successive steps as in the case of primary and secondary swelling observed at macroscopic scale. Moreover, the swelling intensity for Romainville clay is similar to the swelling value measured with classical oedometric free swelling tests. These two points show that microscopic measurements, performed on clay powder and during a few days, are quite identical to macroscopic measurements, carried out on macroscopic samples during a few months. The size of the particles plays a role in the swelling-shrinkage rates. Under the same test conditions, particles five bigger show half the swelling intensity. This relation has to be confirmed or infirmed by tests on bigger particles. However, if this fact is confirmed, the swelling of a soil could be the total sum of the swelling of each level of clay structure. ACKNOWLEDGMENT The authors would like to thank Mr Delalain and Ms Moirez (INERIS-DRC-NOVA) for their precious help in laboratory and ESEM experiments. We would like to thank also Ms Audiguier and Mr Cojean from ENSMP, for their help.
REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Audiguier, M., Geremew, Z., Laribi, S. & Cojean, R. 2007. Caractrisation au laboratoire de la sensibilit au retrait-gonflement des sols argileux. Rev. fran. Gotec. 120121: 6782. Christidis, G.E., Scott, P.W. & Dunham, A.C. 1997. Acid activation and bleaching capacity of bentonites from the islands of Milos and Chios, Aegean, Greece. Appl. Clay Sci. 12: 329347. Fleureau, J.M., Kheirbek-Saoud, S., Soemitro, R. & Taibi, S. 1993. Behavior of clayey soils on dryingwetting paths. Can. Geotech. J. 30: 287296. Gens, A. & Alonso, E.E. 1992. A framework for the behaviour of unsaturated expansive clays. Can. Geotech. J. 29: 10131032. Maison, T., Laouafa, F., Fleureau, J.M. & Delalain, P. 2009. Analyse aux chelles micro et macroscopique des mcanismes de dessiccation et de gonflement des sols argileux. In Proceedings of 19e Congrs Franais de Mcanique (CFM09), Marseille. Mitchell, J.K. 1976. Fundamentals of soil behaviour. New York: Wiley. Romero, E. & Simms, P.H. 2008. Microstructure Investigation in Unsaturated Soils: A Review with Special Attention to Contribution of Mercury Intrusion Porosimetry and Environmental Scanning Electron Microscopy. Geotech. Geol. Eng. 26: 705727. Souli, H., Fleureau, J.M., Trabelsi Ayadi, M. & Besnard, M. 2008. Physicochemical analysis of permeability changes in the presence of zinc. Geoderma 145: 17.
166
Influence of the microstructure on the hydro-mechanical behaviour of a natural silty clay

M.M.A. Mascarenha M.P. Cordo-Neto E. Romero
Universidade Federal de Gias, Goinia, Brazil Universidade de Braslia, Braslia, Brazil Universitat Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: The work presented in this paper focuses on the study of the interplay existing between the microstructural, hydraulic and mechanical behaviour of an unsaturated soil. The material studied is a natural silty clay, that presents a high porosity and a collapsible behaviour. The cross coupling between the mechanical and hydraulic responses of the soil have been investigated and completed by the study of the micro-structural changes induced by mechanical and hydraulic actions. This study allowed concluding on the relatively low influence played by the deformations on the hydraulic and microstructural behaviour of this soil. The determination of the hydraulic properties of the soil from its microstructural features has been performed by means of techniques providing the distribution of porosity inside the material. The results obtained encourage strongly the parallel use of testing techniques allowing the determination of the retention curve and the characterisation of the microstructural characteristics of this material.
INTRODUCTION
The increasing interest of geotechnical engineers to incorporate concepts of unsaturated soil mechanics in their current practice and the development of experimental techniques to reproduce in situ conditions in the laboratory have provided the bases for a significant advance in the comprehension of the behaviour of these materials during the last decades. In parallel, several constitutive models have been developed to analyze the behaviour of these soils and the work of Bishop (1959), Matyas & Radhakrishna (1968) and Fredlund et al. (1978) are the first attempts in this direction. At first the unsaturated soils were analyzed by separating the mechanical behaviour, related to the deformability and strength, and hydraulic behaviour, related to the flow and storage. More recently, there has been a need to engage these two aspects for a more complete understanding of this problem (Gallipoli et al. 2003 and Vaunat et al. 2000). With the advent of modern microstructural testing techniques, the studies of hydro-mechanical behaviour of unsaturated soils based on microstructural characterization is gaining prominence, allowing the emergence of constitutive models that engage all three dimensions of soil behaviour.
Therefore, this article focuses on the study of the influence of the mechanical aspects on the soil hydraulic behaviour and the relationship between soil structure and this hydro-mechanical behaviour, verifying the possibility of determination of soil hydraulic properties from techniques to obtain their structural distribution. 2 MATERIAL AND METHODS
2.1 Soil used in the investigation The material studied is a natural silty clay from the basement layer of the Algerri-Balaguer canal (Catalunya, Spain) located in a semi-arid climate zone. The soil presents a high porosity and a marked collapsible behaviour. The results of the geotechnical parameters of this soil are summarized in Table 1, in which liquid limit (wL), plasticity limit (wP), plasticity index (IP), dry unit weight (d), specific weight of the solids (s), void ratio (e), natural moisture of the soil (wnat), natural degree of saturation (Srnat) and classification in accordance with the Unified Soil Classification System (USCS) are indicated.
167
Table 1. Geotechnical parameters of the soil (Mascarenha, 2008). wL (%) 22 wP (%) 15 IP (%) 7 e d kN/m3 s kN/m3 26.5 wnat Srnat (%) (%) 4 15
0.71 15.5
% passing #40 #200 2 m 100 82 16
USCS Classification CL-ML (silty clay of low plasticity)
2.2
Experimental programme and procedures
To determine the water retention curve of this soil, a Dewpoint PotentialMeter (WP4) and an oedometer with controlled suction were used. The oedometer was used to determine the water retention curves for suction values less than 1 MPa, along wetting and drying paths (path ABC under unstressed conditions in Figure 1a). In order to verify the influence of void ratio, another drying path was performed (path GH in Figure 1a) after a loading, wetting and unloading path (CDEFG path in Figure 1a). To obtain intermediate data in the study of the interaction between void ratio and water retention curves, an oedometer test was also performed alternating loading and wetting, as shown in Figure 1b. WP4 was used to obtain the water retention curve of the natural soil for suction values greater than 1 MPa, along wetting and drying paths. To check the influence of void ratio on the water retention curve, one more WP4 test was performed along a drying path for a sample obtained from an oedometer test. The path of the oedometer test is shown in Figure 2. This path was chosen in order to obtain a sample with the same void ratio in which the water retention curve of the oedometer test ECC was determined, along a drying path and after the path CDEFG in Figure 1a. In order to understand the structural behaviour of the soil and to verify the structural changes due to changes in void ratios and/or collapse, mercury intrusion porosimetry tests (MIP) were performed. The tests were performed on samples at their natural state, as well as on samples submitted to wetting-loading-wetting-unloading paths (ABCDE) and loading-wetting-unloading paths (ABCD) as shown in Figures 3a and 3b, respectively. These paths were carried out in an oedometer with controlled suction and a conventional oedometer, respectively. The values of initial suction of the soil samples were obtained by means of psychrometers. In the MIP technique, the relationship between the pore diameter and the pressure applied is obtained by using Equation 1.
Figure 1. Stress paths of the oedometer tests with suction control: a) ECC; b) Ealternated.
Figure 2. Stress paths of the oedometer test to obtain samples for WP4 test.
D
1 4 hg cos hg P
(1)
where D is the pore diameter, P is the absolute pressure applied, hg is the surface tension of mercury and hg is the contact angle between mercury and the pore wall.
168
According to Romero (1999), the injection of non-wetting mercury with contact angle of 140 (hg) and absolute pressure P is equivalent to the injection of water from the pores (desorption curve) by the non-wetting from advance of air with angle of 180 (w) for the same diameter of pores being intruded. Thus, Equation 1 is extrapolated to the matric suction (ua uw), resulting in Equations 3, 4 and 5. ua uw 0.196P w (wmax wr)(1 Snw) wr Sr (1 Snw ) wr Snw wmax (3) (4) (5)
where wmax is the saturation water content of the sample, wr is the water content correction due to voids not penetrated by mercury and Snw is obtained by Equation 6. Snw enw e (6)
where enw is the volume of voids penetrated by mercury (referred to volume of solids) and e is the void ratio of the sample. 3
Figure 3. Stress paths of the oedometer tests to obtain samples for MIP tests: a) with controlled suction b) conventional.
ANALYSIS OF THE RESULTS
An important information obtained through this test is the PSD (pore size density function), which is derived from the distribution of pore sizes. The PSD is obtained from Equation 2, which is useful to discriminate the dominant pore diameters in the sample. PSD de d (log D ) (2)
The mercury intrusion porosimetry (MIP) is also used to study the hydraulic behaviour of soils. According to Prapaharan et al. (1985), the MIP can be used to determine the water retention curve of a soil. The mercury intrusion procedure can become assimilated to the desorption path of the soil water retention curve by applying to an initially saturated sample an increasing external air pressure (nonwetting fluid) to gradually dry the soil.
As for the microstructural behaviour of the soil, Figure 4 shows the intruded void ratios in relation to the pore size of the soil, both for the sample in its natural state (e 0.71) as for the samples collapsed due to the load and wetting (e 0.53 for the stress path in Figure 3a and e 0.45 for Figure 3b). Moreover, the relationship between the PSD and the pore diameter is illustrated in Figure 5. Changes were detected in the macroporosity (macrostructure) with a reduction in pore volume and some shifting towards lower pore sizes when the samples were loaded and soaked. On the contrary, there were no important changes in the distribution of pores of the microstructure of the soil after the loading and wetting, regardless of the path taken, of the level of collapse and of the final amount of voids. It is worth noting that, according to Romero and Vaunat (2000), the microstructure is the part of the soil not affected by the load paths, at least for these low activity soils. For the mathematical representation of the experimental data of the water retention curve obtained by means of WP4 and oedometer tests, the equation proposed by Romero (1999) was used. This author has modified the expression for the
169
Figure 4. Intrusion and extrusion curves of the soil for natural and collapsed samples.
Figure 5. Intrusion and extrusion curves of the soil for natural and collapsed samples.
water retention curve of van Genuchten (1980), adding a correction factor dependent on the suction, which makes the curve tend to a linear relationship at high values of suction. The proposed new equation is presented below. ln 1 w wsat 1 ln 2
ua uw a
Figure 6. Water retention curves of the soil (fitted equation and MIP): a) in relation to water content, b) in relation to the degree of saturation.
w
1 n 1 [ ua u w ]
(7)
where wsat is the soil water content when saturated, a is the suction value to hygroscopic water content and n, m, and are the same parameters of van Genuchten (1980), the latter being related to the entrance air value of the curve. Because of the relationship between the hydraulic and structural behaviour of the soil, it becomes possible to obtain the water retention curve by means of the pore size distribution, as proposed by Prapaharan et al. (1985) and as earlier discussed.
Thus, the curves obtained by MIP (Equations 3, 4, 5 and 6) and the curves determined by Romeros equation (1999), or Equation 7, in drying and wetting paths are shown in Figure 6. The curves show a reasonable agreement on the form, as verified by Prapaharan et al. (1985), Romero (1999) and Buenfil (2007), and which points out the relationship between the microstructure and hydraulic behaviour of the soil. It is observed that the curve obtained by MIP has a better agreement with the wetting path, for high values of suction, and to the drying path, for low values of suction. Romero (1999) obtained similar results. It is emphasized that in obtaining these curves, different techniques and non-wetting fluids were used (air and mercury), which may explain these differences.
170
Considering the fast procedure of obtaining the water retention curve, the interval of suction measurements allowed and the non-occurrence of significant changes in the void ratios, the MIP method presents itself as a good practical alternative for obtaining water retention curves of the soil. It is important to remark that this method does not allow visualization of the hysteresis of the curves, and this is the great disadvantage compared to traditional methods. In order to verify the influence of the deformability on the soil hydraulic behaviour, water retention curves of samples with different void ratios were performed. The stress paths of the oedometer tests performed for this purpose are shown in Figure 1,
which allow obtaining the curves for void ratios of 0.71 and 0.61 (Figure 1a) and for alternated values of void ratios (Figure 1b), respectively. The samples in which the MIP tests were performed were extracted from oedometer test, whose paths are shown in Figure 3, providing void ratios of 0.53 and 0.45. In the WP4 test the sample was extracted from conventional oedometer, specially carried out for this purpose (Figure 2), obtaining a sample with void ratio of 0.61. Figure 7 presents the results of the water retention curves of the soil for different void ratios, using MIP test results. Figure 8 presents the results of the water retention curves of the soil for different void ratios, using the other techniques mentioned above.
a)
b)
Figure 7. Water retention curves of the soil for initial diferents void ratios (MIP tests): a) (ua-uw, w); b) (ua-uw, Sr). Figure 8. Water retention curves of the soil for initial diferents void ratios (other techniques): a) (ua-uw, w); b) (ua-uw, Sr).
171
The curves related to the degree of saturation maintain the shape and the variation of saturation are not greatly affected by void ratio changes. When linking ua-uw, and w, the curves maintain the shape and the water content values are altered only for values close to their saturation water content. In this case, the deformations have practically no influence on the hydraulic behaviour of the soil for matric suctions higher than 30 kPa. As shown in Figure 5, the reduction in the void ratios from 0.71 to 0.45 did not cause a significant variation in the distribution of pores for the microstructure. This behaviour may explain the little influence that the deformations have on the soil water retention curve at matric suction higher than 30 kPa. Thus, for this low activity and collapsible soil, it is not necessary for the simulation of its behaviour the use of complex elastoplastic hydromechanical constitutive models, such as those proposed by Wheeler et al. (2003), Gallipoli et al. (2003), Vaunat et al. (2000), among others. The traditional elastoplastic constitutive models, such as the BBM (Alonso et al., 2009), are sufficient for an adequate representation. In this sense, MIP tests are quite useful, since they ascertain in the use of more advanced models before carrying out more complex laboratory tests. 4 CONCLUSIONS
REFERENCES
Alonso, E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40 (3): 405430. Bishop, A.W. 1959. The principle of effective stress. Lecture delivered in Oslo, Norway in 1995, printed in Teknisk Ukeblad, 106 (39): 859863. Buenfil, C.M.B. 2007. Caracterizacin Experimental del Comportamiento Hidromecnico de uma Arcilla Compactada. Tese de doutorado, Departamento de Engenharia de Terreno e Cartografia, Universidade Politcnica da Catalunia, Barcelona, 466 p. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsaturated soils. Canadian Geotechnical Journal, 15(3): 313321. Gallipoli, D., Wheeler, S.J. & Karstunen, M. 2003. Modelling the variation of degree of saturation in a deformable unsaturated soil. Gotechique, 53: 105112. Mascarenha, M.M.A. 2008. Influncia da Microestrutura no Comportamento Hidro-mecnico de uma Argila Siltosa no Saturada Incluindo Pequenas Deformaes. Tese de Doutorado, Publicao G.TD056/08, Departamento de Engenharia Civil, Universidade de Braslia, Braslia, DF, 158 p. Matyas, E.L. & Radhakrishna, H.S. 1968. Volume change characterists of partially satured soils. Geothechnique, 18 (4): 432448. Prapaharan, S., Altschaeffl, A.G. & Dempsey, B.J. 1985. Moisture curve of a compacted clay: mercury intrusion method. Journal of Geotechnical Engineering, 111(9): 11391143. Romero, E. 1999. Characterisation and Thermo Hydromechanical Behaviour of Unsaturated Boom Clay: An Experimental Study. Tese de Doutorado, Departamento de Engenharia de Terreno e Cartografia, Universidade Politcnica da Catalunia, Barcelona. 405 p. Romero, E. & Vaunat, J. 2000. Retention curves of deformable clays. Proc. Int. Workshop on Unsaturated Soils: Experimental Evidence and Theoretical Approaches in Unsaturated Soils. Trento, Italy, April 1012, 2000. A. Tarantino and C. Mancuso (eds.). A.A. Balkema, Rotterdam: 91106. Van Genuchten, M.T. 1980. A closed-form equation of predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of American Journal, 44: 892898. Vaunat, J. Romero, E. & Jommi, C. 2000. An elastoplastic hydromechanical model for unsaturated soils. Experimental Evidence and Theoretical Approaches in Unsaturated Soils, Tarantino & Mancuso (eds.). Balkema, Rotterdam: 121138. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Gotechnique, 53: 4154.
The curves obtained using mercury intrusion porosimetry show a reasonable agreement in the form to those obtained by using conventional methods, suggesting the coupling between the soil microstructural and hydraulic behaviour of the soil. This technique appears to be a good alternative to other techniques, though it does not allow viewing of the hysteresis of the curves, and this is its greatest disadvantage compared to other techniques used to determine the values of suction. The characteristic curves of the soil used in this study, obtained for samples with different void ratios did not show marked changes, in other words, the deformations have practically a limited influence on the hydraulic behaviour of this low activity soil in relation to the water retention curve. The situation described above may be related to the low micropore size distribution changes with decreasing void ratios from 0.71 to 0.45, as shown by MIP results in the microstructural range. This way, the hydraulic behavior is not only related to the deformation undergone, but also as this strain is reflected in the distribution of soil pores.
172
Stress-strain and strength characteristics
Hydro-mechanical response of an expansive silt-bentonite mixture

M. Ajdari
LAEGO, INPL, Nancy-University, France Department of Civil Eng., School of Engineering, Shiraz University, Iran
H. Nowamooz & F. Masrouri
LAEGO, INPL, Nancy-University, France
G. Habibagahi & A. Ghahramani
Department of Civil Eng., School of Engineering, Shiraz University, Iran
ABSTRACT: To arrive at a better comprehension of the hydro-mechanical behavior of expansive clays, the Soil Water Retention (Characteristic) Curve of a dual porosity silt-bentonite mixture is determined employing vapor equilibrium method and osmotic technique and the validity of these approaches is examined using the results of the filter paper technique. Parameters of the curve are determined employing a modified version of Brooks and Corey (1964) formula for bimodal Soil Water Retention Curves. Moreover, oedometric response of the material is observed for different suction values and volume change of the soil is measured during the free drying and wetting tests under isotropic condition and the results are discussed in details. The role of suction and degree of saturation in the hydro-mechanical behavior of this soil is discussed in the light of experimental results. 1 INTRODUCTION Rowe (2007); Puppala et al. (2006); Villar & Lloret (2004); Likos et al. (2003); Marcial et al. (2002); Ng et al. (2000); Bao & Ng (2000)). In this paper the SWRC of a dual porosity siltbentonite mixture is determined employing vapour equilibrium method and osmotic suction technique and validity of these approaches is examined against the results from the filter paper technique. Next, the parameters of the curve are determined employing a modified version of Brooks & Corey (1964) formula for bimodal SWRCs. Furthermore, wetting induced free swelling in k0 condition is examined and its effects on the behavior of samples are studied. Then, the oedometric response of the material is investigated in a fully saturated condition and also while samples are imposed suctions more than zero. In order to detect the effect of suction on the behavior of expansive soils, the results of SWRC tests are employed to study the volume change of the soil during the free drying and free wetting tests under isotropic condition and the results are discussed in details. 2 MATERIAL AND TEST PROCEDURE
Owing to their low permeability, compacted swelling soils are used for the construction of engineered barriers in waste disposal facilities (Alonso, et al., 2005). Also, structures founded on this type of material suffer damages from minor cracking of pavements or interior finishing of buildings, irreparable displacement of footings and superstructure elements (Chen, 1975). Hence, understanding the behavior of unsaturated expansive clays is essential in the design and construction of light structures, embankments and roads. Most of the experimental works have been carried out either using oedometer apparatus (Alonso, 1995; Tripathy, 2002) or in the triaxial isotropic condition (Blight, 1965; Sharma & Wheeler, 2000). Therefore, additional experimental tests including isotropic free drying-wetting tests will be useful for a thorough understanding of the role of suction and degree of saturation. To study the hydro-mechanical behavior of multiphase media, such as unsaturated soils, one needs to know about the characteristics of the material. Parameters of Soil Water Retention Curve (SWRC) can represent some of these features for unsaturated soils. Some experimental and analytical studies have been carried out to consider the SWRC of expansive soils in recent years (Southen &
In this study, hydro-mechanical response of an artificially prepared highly expansive silt-bentonite mixture was studied. Forty percent of the material
175
consisted of Xeuilley silt from east of France. X-ray diffractometry demonstrated that this silt contains 60% quartz, 20% montmorilonite and 11% feldspar, with the remaining part containing kaolin and mica. The commercially available bentonite contains more than 80% calcium montmorilonite. Table 1 shows the physical characteristics of the material. Samples were statically compacted and then, inundated in an oedometer to observe the free swelling behavior under the oedometric condition. Oedometric loading of the material and also isotropic free wetting and drying cycles were carried out utilizing osmotic technique and vapor equilibrium method. Furthermore, total suction measured by the filter paper technique, ASTM D529894 (ASTM 1995a), was employed to verify the results. In osmotic method, a polyethylene glycol (PEG) solution was used to apply the suction. A semipermeable membrane separates macromolecules of solution and unsaturated soil (Nowamooz & Masrouri (2008); Nowamooz et al., 2009). Osmosis process allows water to exchange across the membrane with the amount of water swap depending on the macromolecule concentration. PEG with a molecular weight of 6000 Da (1 Da 1.6605 1024 gr) was selected for this study. The relationship between PEG concentration and the amount of suction is given by (Delage et al., 1998; Cuisinier & Masrouri, 2005): s 11c2 (1)
Vapor equilibrium method has been founded on the Kelvins equation for perfect gases; that is: s w RT ln(RH ) Mg (2)
In which, s is suction (kPa), R is universal constant for perfect gases (8.31 J 1 mol 1 K 1), is the unit weight of water (9.81 kN/m3), g is the gravitational constant (9.8 m/sec2), M is molecular weight of water (18 10 3 kg mol 1) and RH is relative humidity (%). Thus, a constant value of total suction will be imposed on a soil sample placed nearby a saturated salt solution in an airtight chamber. Different total suctions can be imposed by employing different solutions. It is worth mentioning that the uncertainty in the relative humidity imposed by a salt solution is between 1 and 2%. Therefore, the validity of saturated vapour equilibrium technique is limited to suctions more than 10 MPa (Cuisinier, 2002). A petroleum product named Kerdane was employed to determine the volume of voids, and hence, the variation of volume during free wetting and drying processes. The Kerdane contact angle is more than 90, and therefore, it is a non wetting fluid. Samples were weighed in the wetted situation and while samples had been submerged in Kerdane. Total volume of the samples was then determined using Archimedes rule. 3 SOIL WATER RETENTION CURVE
where, s is the suction in MPa and c is the concentration of PEG in gr of PEG per gr of water. Temperature affects the relationship between suction and PEG concentration and equation (1) is valid when the temperature is maintained at 20 1.5c and for suctions less than 8.5 MPa (Nowamooz & Masrouri, 2009).
Table 1. Properties Liquid Limit (%) Plasticity Index (%) Specific Gravity Initial Dry Density (kN/m3) Initial Water Content (%) Pressure for Static Compaction (kPa) Optimum Water Content in Static Compaction (%) Maximum Dry Density in Static Compaction (kN/m3) Free Swelling (%) Swelling Pressure (kPa) Physical properties of silt-bentonite mixture. Value 87 22 2.67 14.5 15 2000 23 17.8 26 350
For gap graded soils with two or more pore structures, the corresponding SWRC can be bimodal or multimodal. Several studies have been performed on this type of SWRC in recent years (Burger & Shackelford, 2001; Zhang & Chen, 2005; Bagherieh et al., 2009). In the present study, desorption and adsorption branches of the SWRC of silt-bentonite mixture were determined in terms of gravimetric moisture content, w, and also, volumetric water content, using osmotic suction, for suctions less than or equal to 8 MPa, and as well, employing vapor equilibrium method, for suction values more than 8.5 MPa. The results of the drying branch were compared with the results from the filter paper test (Fig. 1). There is a difference between the results of filter paper and osmotic method in the first portion of SWRC as shown in figure 1. This difference indicates the difference between total and matric suctions of the soils measured using filter paper and osmotic methods, respectively. Figures 2 and 3 show the experimental data points describing variation of volumetric water content versus suction during desorption and sorption process, respectively. These bimodal SWRCs
176
imply that the micro pores can be desaturated even in the highly expansive clayey soils. Various empirical equations have been suggested to describe the soil water characteristic curve. Among these equations, the relationships proposed by Brooks & Corey (1964), van Genuchten (1980) and Fredlund & Xing (1994) have been more used in geotechnical engineering. In particular, Brooks &
Corey (1964) simulation is simpler and its parameters are physically meaningful. This equation for unimodal SWRC is written as follows:
(u uw )b r a s r (ua uw ) s
(ua uw ) (ua uw )b (ua uw ) (ua uw )b (3)
Figure 1. suction.
Drying branch of SWRC: water content vs.
In which, ua is pore air pressure, uw is pore water pressure, is a fitting parameter and represents the slope of the curve in a logarithmic space, is saturated volumetric water content, (ua uw)b is bubbling pressure, and is residual water content. For dual porosity soils, with suction surpassing the first air entry value, air starts to fill the macro-pores. At this stage the micro-pores remain saturated until the suction reaches the second bubbling pressure. Hence, to determine the volumetric water content in bimodal SWRCs one can employ superposition principle (Bergur & Shakelford, 2001). In other words, residual water content of macro pores can be used as the saturated water content of micro pores. Thus, to modify the original formula of Brooks and Corey for bimodal SWRCs; the original two-pieces function is replaced by a three-part equation as follows: s (u u ) M rM ( s rM ) a w bM ( ua uw ) ( ua uw )bm m ( ) rm rM rm ( ua uw )
( ua uw ) ( ua uw ) bM ( ua uw ) bM ( ua uw ) ( ua uw ) bm ( ua uw ) bm ( ua uw )
(4)
Figure 2.
0.7
Drying branch of SWRC.
where, rM, rm, (ua uw)bM and (ua uw)bm are residual water content for macro-pores (saturated volumetric water content for micro-pores), residual
Table 2. Parameters of mixture. Parameter (ua uw)bM (MPa) (ua uw)bm (MPa) M m rM rm S sm SWRC of silt-bentonite Drying cycle 2 113 0.38 1.29 0.176 0.01 0.58 0.176
0.6
Volumetric Water Content
0.5
0.4
Wetting cycle 2 83.6 2.52 0.6 0.135 0.00 0.57 0.135
0.3
0.2
0.1
0 10-2
10-1
100
101
102
103
Suction (MPa)
Figure 3.
Wetting branch of SWRC.
177
volumetric water content for micro -pores, bubbling pressure of macro pores and bubbling pressure for micro-pores, respectively. To find a proper curve for bimodal SWRC of the double porosity silt-bentonite mixture employed in this study, (ua uw)bM, (ua uw)bm and S were directly determined from experimental data and rM, rm, m and M were determined by fitting equation 4 to the data points. Parameters of the curves are presented in Table 2 and the fitted curves are shown in Figures 2 and 3. From Table 2 it may be concluded that good agreement obtained between the model made by equation (5) and most of the actual data measured during experiments. 4 OEDOMETRIC RESPONSE
concavity which conforms to the behaviour of dual porosity soils (Nowamooz & Masrouri, (2009); Bagherieh, et al. (2009)). Mechanical loading leads to the destructuring of the hardened samples approaching asymptotically to the saturated normal consolidation line at high applied net stress. 5 ISOTROPIC FREE DRYING RESPONSE
Results from constant suction consolidation tests carried out on the silt bentonite specimens are exhibited in Figure 4. Albeit, all samples had identical initial preconsolidation pressure equal to 2000 kPa (Table 1); yielding of the samples commenced at a net pressure of less than 100 kPa, 300 kPa and 900 kPa for suction values equal to 0, 2 and 8 MPa, respectively (Figure 4). Consequently, the plastic initial swelling causes the soils to soften significantly. Furthermore, this softening occurs sequentially with suction decrease and samples with lower suctions have lower yield stress and lower compressibility (Table 3). This result confirms the hypothesis that the preconsolidation pressure of aggregated soils depends not only on the stress history but also on the soil structure (Koliji, et al., 2008). Also, the saturated sample exhibited a curved normal consolidation line with a slight
In order to study the role of suction on volume change behaviour of the tested material and to exclude the effects of the applied net stress, SWRC tests are replotted in the space of void ratio-suction and Sr-suction as shown in Figure 5. The first portion of the void ratio-suction curve in the free drying process for the silt-bentonite mixture can be approximated by a bilinear curve (Fig. 5). Surpassing the air entry value corresponding to the micro-pores (113.2 MPa); the soil behaves as an elastic material again and modulus of elastic stiffness should be utilized to simulate the volume change of the soil (Khalili, et al., 2004). This behavior implies that this soil is dense with respect to macro-pores and suction hardening (and distance between different normal consolidation lines at different suction values) is negligible in comparison to the wide suction range (4113 MPa). Surpassing the air entry value of micro-pores; dispersed texture of the clay particles changes to flocculated texture (structure formation). Similar to the behavior of other structured soils (Nova, et al., 2003; Callisto & Rampello, 2004), slope of the normal consolidation line of samples that are hardened by suction is always steeper than the slope of intrinsic normal consolidation line. Obviously, wetting causes the initial dispersed texture of the samples to recover again (Basma, et al., 1996), structure degradation, and normal consolidation line comes back to the primary position. It means, not only micro pores can be desaturated; but also, unsaturated behavior for this kind
Figure 4. Oedometric response of silt-bentonite Mixture. Table 3. Suction 0 MPa 2 MPa 8 MPa Results of the oedometer tests. Cr 0.1 0.03 0.02 Cc 0.37 0.37 0.47 v,c(kPa) 100 300 900
Figure 5.
Results of isotropic free drying.
178
Figure 6.
Results of free wetting test.
of soil commences from the bubbling pressure of micro pores. Thus, need for a suitable elastic-plastic model for unsaturated silt-bentonite mixture starts from the desaturation of micro-pores. 6 ISOTROPIC FREE WETTING RESPONSE
The results of the isotropic free wetting test on the silt bentonite specimen are exhibited in Figure 6. Because of the entrapped air, wetting cycle of SWRC does not come to an end at degree of saturation equal to 1. Obviously, the trend of the variation of void ratio with respect to suction follows the shape of the wetting branch of SWRC. Liangtong (2003) and Likos (2004) reported similar results for a slightly expansive soil and bentonite, respectively. As the rate of the water adsorption increases; volume of the specimen grows up more rapidly followed by decrease in the slope of SWRC, rate of volume change diminishes. This can confirm the importance of soil water characteristic curve in applied unsaturated soil mechanics. If wetting process is limited to the suction ranges less than the least air expulsion value (in extremely humid areas) or more than suction coincident with residual water content (in most arid lands), plastic swelling will be negligible (Fig. 6). Otherwise, if suction variation lies between these two extremes, the amount of plastic swelling will be considerable. 7 CONCLUSION
softening was significant. The mild curvature of the normal consolidation lines validated the behavior of materials with double porosity structures. Suction hardening phenomenon was not significant for low suction values. But, surpassing the air entry value of the micro-pores, the soil behaved as an elastic material again and the modulus of elastic stiffness should be utilized to simulate the volume change of the soil. Trend of the variation of the void ratio with respect to suction during the free wetting test follows the shape of the wetting branch of SWRC. Therefore, the degree of saturation plays a major role in the mechanical behavior of unsaturated expansive soils. As the gradient of the water absorption rises; volume of the sample grows up more quickly and with decreasing the slope of SWRC, rate of volume change diminishes. More elaborate and detailed experimental program is required to determine the necessary parameters that are essential for quantitative simulation of the tests results. Further studies are encouraged for evaluating the role of suction and degree of saturation in the hydro-mechanical behavior of expansive clays during drying-wetting cycles. REFERENCES
Alonso, E.E.; Lloret, A.; Gens, A. & Yang, D.Q. 1995. Experimental Behaviour of Highly Expansive DoubleStructure Clay. Proc. 1st Int. Conf. on Unsaturated Soils. Balkema. Paris. Vol. 1, pp. 1116. Alonso, E.E.; Romero, E.; Hoffmann, C. & GarciaEscudero, E. 2005. Expansive Bentonite-Sand Mixtures in Cyclic Controlled-Suction Drying and Wetting. Engineering Geology, Vol. 81, 213226. ASTM D 52980094. 1995. Standard Test Method for Measurement of Soil Potential (Suction) Using Filter Paper. Vol. 4.09, pp. 154159. Bagherieh, A., Khalili, N., Habibagahi, G., & Ghahramani, A. 2009. Drying Response and Effective Stress in a Double porosity Aggregated Soils. J. of Eng. Geology. Accepted. Bao, C.G., & Ng, C.W.W. 2000. Some thoughts and studies on the prediction of slope stability in expansive soils. Proc., 1st Asian Conf. on Unsaturated Soils, Balkema, Rotterdam, The Netherlands, 1532. Basma, A.A.; Al-Homoud, A.S.; Malkawi, A.I.H. & Al-Bashabsheh, M.A. 1996. Swelling-Shrinkage Behaviour of Natural Expansive Clays. Applied Clay Science; Vol. 11; pp. 211227. Blight, G.E. 1965. A Study of Effective Stress for Volume Change. In Symposium on Moisture Equilibria and Moisture Changes in Soils Beneath Covered Area, Butterworths, Sydney, Australia, pp. 259269. Brooks, R.H. & Corey, A.T. 1964. Hydraulic properties of Porous Medium Hydrology paper, No. 3. Civ. Eng. Dep., Colorado State Univ., Fort Colins, Colo. Burger, C.A. & Shakelford, C.D. 2001. Evaluating Dual Porosity of Pelletized Diatomeceous Earth Using
To study the hydro-mechanical behavior of a highly expansive silt-bentonite mixture, the bimodal SWRC of the material was determined using the osmotic method and vapor equilibrium technique. Results were verified by comparing with the results from the filter paper test. A modified version of Brooks and Corey (1964) formula was employed to determine the SWRC parameters. Oedometric response of the soil was studied at different constant suctions. Swelling induced
179
Bimodal Soil-Water Characteristic Curve Functions. Can. Geotech. J., Vol. 38, pp. 5366. Callisto, L., & Rampello, S. 2004. An interpretation of structural degradation for three natural clays. Can Geotech. J., Vol. 41, pp. 392407. Chen, F.H. 1975. Foundations on Expansive Soils. Elsevier Scientific Publication Company. Cuisinier, O. 2002 Comportement hydromcanique des sols gonflants compacts. Thse de I.N.P.L. Nancy, 172 p. Cuisinier, O. & Masrouri, F. 2005. Hydromechanical Behaviour of a Compacted Swelling Soil over a Wide Suction Range. Engineering Geology, Vol. 81, pp. 204212. Delage, P., Howat, M.D. & Cui, Y.J. 1998. The Relationship between Suction and the Swelling Properties in a Heavily Compacted Swelling Clay. Engineering Geology, Vol. 50, pp. 3148. Fleureau, J.M., Kheirbek-Saoud, S., Soemitro, R., & Taibi, S. 1993. Behaviour of Clayey Soils on DryingWetting Paths. Can. Geotech. J., 30, 287296. Fredlund, D.G. & Xing, A. 1994. Equations for the SoilWater Characteristic Curve. Canadian Geotechnical Journal, Vol. 31, pp. 521532. Khalili, N., Geiser, F. & Blight, G.E. 2004. Effective Stress in Unsaturated Soils: Review with New Evidences. Int. J. of Geomechanics. 4(2), 115126. Koliji, A., Vulliet, L. & Laloui, L. 2008 New Basis for the Constitutive Modelling of Aggregated Soils. Acta Geotech. Vol. 3, pp. 6169. Laloui, L. & Nuth, M. 2009. On the Use of Generalized Effective Stress in the Constitutive Modelling of Unsaturated Soils. Computers and Geotechnics, Vol. 23; pp. 2023. Liangtong, Z. 2003. Field and Laboratory Study of an Unsaturated Expansive Soil Associated with Raininduced Slope Instability. PhD Thesis, Hong Kong University. Likos, W.J. 2004. Measurement of Crystalline Swelling in Expansive Clay. Geotechnical Testing Journal; Vol. 27(6); pp. 17. Likos, W.J., Olsen, H.W., Krosley, L. & Lu, N. 2003. Measured and Estimated Suction Indices for Swelling Potential Classification. J. Geotech. Geoenviron. Eng., 129(7), pp. 665668. Loret, B. & Khalili, N. 2002. An Effective Stress ElastoPlastic Model for Unsaturated Soils. Mech. Mater., Vol. 44; pp. 97116. Marcial, D., Delage, P. & Cui, Y.J. 2002. On the High Stress Compression of Bentonite. Canadian Geotechnical Journal, Vol. 39; pp. 812820.
Ng, C.W.W., Wang, B., Gong, B.W. & Bao, C.G. 2000. Preliminary study on Soil-Water Characteristics of Two Expansive Soils. In H. Rahardjo, D.G. Toll, and E.C. Leong, eds. Proc., Asian Conf. on Unsaturated Soils: From Theory to Practice,Singapore,, Balkema,Rotterdam, The Netherlands, 347353. Nova R., Castellanza R. & Tamagnini R. 2003 A Constitutive Model for Bonded Geomaterials Subject to Mechanical and/or Chemical Degradation. Int J Numer Anal Methods Geomech Vol. 27(9), pp. 705732. Nowamooz H. & Masrouri, F. 2008. Hydromechanical Behaviour of an Expansive Bentonite/Silt Mixture in Cyclic Suction-Controled Drying and Wetting Tests. Engineering Geology, Vol. 101, pp. 154164. Nowamooz, H. & Masrouri, F. 2009. Density-Dependent Hydromechanical Behaviour of a Compacted Expansive Soil: Experimental and Analytical Aspects. Engineering Geology, In Press. Nowamooz, H., Mrad, M., Abdallah, A. & Masrouri, F. 2009. Experimental and Numerical Studies of the Hydromechanical Behaviour of a Natural Unsaturated Swelling Soil. Can. Geotech. J., Vol. 46, pp. 118. Puppala, A.J., Punthutaecha, K. & Vanapalli, S.K. 2006 Soil-Water Characteristic Curve of Stabilized Expansive Soils. ASCE, J. of Geotech. and Geoenv. Eng., Vol. 132(6), pp. 736751. Southen J.M. & Rowe, R.K. 2007. Evaluation of Water Retention Curve for Geosynthetic Clay Liners Geotextiles and Geomembranes, Vol. 25, pp. 29. Sridharan, A. & Gurtug, Y. 2004. Swelling behaviour of compacted fine-grained soils. Engineering Geology, Vol. 72, pp. 9 18. Tripathy, S.; Rao, K.S.S. & Fredlund, D.G. 2002. Water Content-Void Ratio Swell-Shrinkage Paths of Compacted Expansive Soils. Can. Geotech. J., Vol. 39; pp. 938959. van Genuchten, M.T. 1980 A Closed Form Equation for Predicting Hydraulic Conductivity of Unsaturated Soils. Soil Science Society of America Journal, Vol. 44(5), pp. 892898. Villar, M.V. & Lloret 2004. Influence of Temperature on the Hydro-Mechanical Behaviour of a Compacted Bentonite. Applied Clay Science, Vol. 26; pp. 337350. Zhang, L. & Chen, Q. 2005. Predicting Bimodal Soil-Water Characteristic Curves. ASCE, J. of Geotech. and Geoenv. Eng., Vol. 131(5), pp. 666670.
180
Deformation behavior in unsaturated soils

R. Baltodano-Goulding
Civil Engineering Department, University of Costa Rica, Costa Rica Construction Engineering Department, Costa Rica Institute of Technology, Costa Rica
ABSTRACT: Vertical deformations were recorded during a series of direct shear tests performed on Ottawa Sand. Since the initial void ratio is known, it is possible to determine the final void ratio, and thereby determine if the specimen experiences a net decrease or an increase in volume. A dense sand specimen tested in a direct shear test will typically compress at relatively small horizontal displacements and then expand at higher displacements. In the other hand, a loose sand specimen is expected to compress during shearing. It was observed that some specimens that were prepared to a void ratio of 0.60 (dense condition) behaved as loose specimens (contraction). This can be explained if it is considered that the critical void ratio is not a constant but a function of the effective normal stress. Since the effective stress varies with suction stress, and this varies with degree of saturation, it can be expected that as the effective normal stress is increased it might be enough to relocate the normal stress onto the loose (contraction) side of the critical state line. This can lead to the thinking that for unsaturated soils not only the normal stress but the mechanisms involved in the unsaturated regimes acting in the specimen affect the deformation behavior of the soil. Because soil strength depends partially on how much deformation has occurred under a particular loading condition, the influence of these results on soil strength should be considered for Geotechnical Engineering Projects. Then a series on shear strength testing on cohesive soils was performed in order to compare to the results observed on granular soils.
INTRODUCTION
The main goal of this work was to investigate the effect suction stress has in the deformation behavior of unsaturated soils. In order to isolate the suction stress contribution to changes in deformation behavior granular soils, where physico-chemical forces are not present, were selected. Effective stress controls soil shear strength and it is influenced by the suction stress, which changes with degree of saturation. Later the stress-strain behavior for cohesive soils was evaluated in order to compare it to what it was observed in granular soils. Shear stress-horizontal and volumetric behavior obtained from direct shear tests, and tensile deformations obtained from tensile strength tests was analyzed for granular soils. Stress-strain behavior obtained from unconfined compression tests in cohesive soils was also analyzed. Finally the behavior changes observed are explained based on critical state theory. 2 MATERIALS AND METHODS
was selected. This material was selected because is inert, so changes in suction stress can be attributed solely to changes in matric suction. Also Ottawa sand grains are rounded making it easier to compare to existing micro-structural theories based on particle geometry, Baltodano-Goulding (2006). Table 1 shows some characteristics for Ottawa sand used for this work. Table 2 shows some characteristics for the cohesive soils used to investigate the soil behavior variation with degree of saturation. Compacted specimens were prepared using Ladds (1978) undercompaction method. This method allows achieving homogeneous specimens and reproducible densities. Individual soil layers (lifts) were compacted to a lower density than the final desired value to normalize the cumulative compaction effect, as each subsequent layer is compacted. The expression used to calculate the percent undercompaction was: U U Nt U N U Ni Ni n 1 nt 1 where UN is the undercompaction percentage, UNi is the percent undercompaction of the first layer, UNt
In order to isolate the suction stress contribution to changes in deformation behavior Ottawa Sand
181
Table 1. Granular soil characteristics BaltodanoGoulding (2006). Soil type F-40 Ottawa sand F-55 Ottawa sand F-75 Ottawa sand d60 (mm) 0.38 0.27 0.26 d10 (mm) 0.25 0.18 0.15 Cu 1.52 1.50 1.73 Cc 0.77 1.09 0.83
tensile deformations were observed to evaluate the change in behavior for specimens at approximately the same density but with changing degree of saturation. Latter, a series of unconfined compression tests were performed on the cohesive soils studied and the stress-strain behavior was analyzed. 3 3.1 RESULTS Shear stresshorizontal displacement behavior
Table 2. Cohesive soil specimen properties BaltodanoGoulding & Korte (2009). Atterberg Standard proctor Limits resulats Classification (USCS) MH CL SC SC SC Dry unit weight (kPa) 15.3 16.8 20.8 16.9 15.5 Opt water content (%) 23.5 19.5 12 16 21
Soil 1 2 3 4 5
LL 55 38 28 40 47
PI 22 15 9 15 18
is the percent undercompaction of the last layer, n is the number of the current layer, nt is the total number of layers. Number of layers and undercompaction percentages were chosen according to the requirements of each particular type of soil and void ratio. The soils were mixed with different amounts of water to produce the target degree of saturation. Since the void ratio and soil type were kept constant the only independent variable is the degree of saturation. There are two dependent variables: suction stress and soil structure, but since the differences in soil structure for the same soil type and void ratio are affected by the interparticle forces captured in the suction stress term, we attribute all changes in soil structure and shear strength observed to the suction stress variation with degree of saturation. Then, vertical deformations were recorded during a series of direct shear tests performed on Ottawa Sand. Since the initial void ratio is known, it is possible to determine the final void ratio, and thereby determine if the specimen experiences a net decrease or an increase in volume. In addition a series of tensile strength tests were performed on the same soil and horizontal displacements were recorded. Using these results different soil behaviors were analyzed. Shear Stress-horizontal displacement, volumetric strain behavior, Critical state, and
Direct shear specimens compacted to a relatively dense state are generally expected to exhibit a peak in shear stress and then experience a residual strength at large deformations. Terzaghi & Peck, (1960). Specimens at approximately the same density behaved different at different degrees of saturation. The relatively dry specimen exhibits peak behavior more characteristic of a densely compacted sand specimen. The relatively wet specimen exhibits non-peak behavior more characteristic of a loosely compacted sand specimen. This behavior may be related to the apparent double-peak behavior mentioned by Baltodano-Goulding (2006). This behavior is more common in granular soils (Baltodano-Goulding 2006) than in fine grain soils where the saturation mechanisms are more complex; however, it was observed as part of this research, in fine soils too. Baltodano-Goulding (2006) mentioned that for specimens with water contents close to the first peak (near the pendular regime), the soil exhibits behavior more characteristic of dense sand behavior. As the water content approaches the second peak (near the capillary regime) the soil starts to exhibit behavior more characteristic of loose sand. The more brittle response of the sand near the pendular regime probably reflects relatively uniform failure of the interparticle water bridges at small levels of strain. The peak in shear stress occurred at a horizontal displacement of about 0.05 in (1.27 mm). The specimen with water content near the pendular regime does not show a peak in shear stress. This behavior is expected for loose specimens but not for dense specimens. It is possible that the deformation behavior of sand might be affected by an increase in the effective normal stress produce by an increase in suction stress that can produce a dense specimen to behave as a loose one. 3.2 Stressstrain behavior
Cohesive soils specimens at approximately the same density tested on unconfined compressions also behaved different at different degrees of
182
saturation. The relatively dry specimen exhibits peak behavior more characteristic of an overconsolidated specimen. The relatively wet specimen exhibits non-peak behavior more characteristic of a normally consolidated specimen. Again specimens with water contents close to the first peak (near the pendular regime), exhibited behavior more characteristic of overconsolidated soils. As the water content approaches the second peak (near the capillary regime) the soil starts to exhibit behavior more characteristic of normally consolidated soils. The more brittle response of the cohesive soils near the pendular regime probably reflects its flocculated structure in contrast with a dispersed structure related to wetter specimens. According to Mitchell (1993), soil structure can be defined as the combined effect of fabric, composition, and interparticle forces. Although, Mitchell (1993) mentioned that the mechanical properties of soils are influenced by the soil type, the structure type, and the stresses, soil structure and interparticle forces are included in suction stress concept. On the other hand, the resulting fabric, after the compaction process of our specimens is done, will be affected by the energy of compaction and the degree of saturation, but since the specimens were prepared at the same void ratio those changes in fabric are due to the change in degree of saturation, so it is also captured in the suction stress concept. The specimen with water content near the pendular regime does not show a peak in shear stress. This behavior is expected for normally consolidated specimens but not for overconsolidated specimens. It is possible that the deformation behavior of cohesive soils might be affected by an increase in the effective normal stress produce by an increase in suction stress that can produce an overconsolidated specimen to change its structure to that of a normally consolidated one. Figure 1 show results obtained for cohesive soil 4. 3.3 Volumetric strain behavior
Figure 1. Stress-strain curves for Soil 4 at different degrees of saturation (void ratio 0.89). Korte-Leiva (2009).
first contract and then dilate as the horizontal displacement increases. However, the specimen with water content near the pendular regime (w 6%) shows initially a decrease in volume and then an increase that exceeds the decrease, thus resulting in a higher void ratio at the end of the test than at the beginning. For higher values of normal stress. The specimen at 6% water content behaves in a similar manner as before, but the initial compression is smaller than at the smaller value of normal stress. The specimen at 18% water content behaves in a manner characteristic of a loose specimen, even though the density and normal stresses have not changed. Both specimens experience compression over the entire measured range of horizontal displacement. The soil deformation behavior observed can be affected by the saturation regime mechanisms involved as well as the increase in effective normal stress due to those mechanisms. These observations are may be interpreted in terms of critical state behavior. 3.4 Critical state line
A dense sand specimen tested in direct shear will typically compress at relatively small horizontal displacements and then expand at higher displacements. The relationship between volumetric strain and horizontal displacement for densely compacted F-75 sand at 6% shows initially a decrease in volume and then an increase that exceeds the decrease, thus resulting in a higher void ratio at the end of the test than at the beginning. For 18% water content and applied normal stress of 5 psi (34.5 kPa) the initial compression of the specimen is both larger in magnitude than that for the dryer specimen and is not fully recovered. The final void ratio is smaller than the initial. Both specimens
Wood (1990) defines the critical state line as the ultimate condition in which shearing could continue indefinitely without changes in volume or effective stress. Atkinson & Bransby (1978) said the critical state line is a single and unique line of failure points for both drained and undrained tests. They added that its crucial property is that failure will occur once the stress states of the specimens reach the line, independently of the path followed on their way to the critical state line. This means that for a given confining stress, the critical void ratio is the void ratio at which the specimen will end up at large strains. This critical void ratio is independent of initial void ratio and confining stress.
183
Wood (1990) mentioned that for dense sand the shear stress reaches a maximum value and if the deformation continues, the shear stress will drop to a smaller value. When this value is reached it will remain constant for all further displacement. During this drop in shear stress the sand continues to expand until the critical void ratio is achieved. At this point continuous deformation is possible at constant shear stress. When loose sand is sheared under constant normal stress the shear stress will increase until it reaches the maximum value. However, if the displacement continues the shear stress will remain the same. Thus, if the same material is tested at loose and dense conditions the curves representing the volume change during shearing meet at the critical state line. Vertical deformations were recorded during the direct shear tests. Since the initial void ratio is known, it is possible to determine the final void ratio, and thereby determine if the specimen experiences a net decrease or an increase in volume. If the initial void ratio is smaller than the critical void ratio the specimen will tend to dilate. If the initial void ratio is greater than the critical void ratio the specimen will tend to contract. It was observed that some specimens that were prepared to a void ratio of 0.60 (dense) behaved as loose specimens (contraction). This can be explained if it is considered that the critical void ratio is not a constant but a function of the effective normal stress. Since the effective stress varies with suction stress, and this varies with degree of saturation, it can be expected that as the effective normal stress is increased it might be enough to relocate the normal stress on the loose (contraction) side of the critical state line. It is more likely for a specimen with a high normal stress to be on the contraction side of the critical stress line, and if an increase due to the suction stress is present, this is even more likely. It is also likely that the specimens prepared to higher water content will show higher effective normal stresses, since the suction stress is greater, thus increasing the likelihood of having a specimen on the contraction side of the critical state line. However, it is important to mention that the effective stress increase due to the suction stress at this range of normal stresses was determined to be minimal. This can lead to the thinking that for unsaturated sands not only the normal stress but the mechanisms involved in the unsaturated regimes acting in the specimen affect the deformation behavior of the soil. 3.5 Tensile deformations
of degree of saturation showed a peak tensile displacement over the entire range of saturation similar for the loose and dense cases and it ranges from about 0.006 inches to 0.014 inches (0.15 mm to 0.36 mm). Peak displacement for the loose specimen appears to decrease at higher saturation. The measured range of displacement at failure is on the same order of magnitude as the mean grain size for the F-75 sand (d50 0.22 mm). For an ideally smooth failure plane in the pendular regime, this suggests that the liquid bridges bonding the particles on either side of the failure plane break when the ratio of particle separation to particle diameter is about 0.68 to 1.63. It is more likely that the liquid bridges fail non-uniformly at much smaller separation distances on a failure plane having roughness on the scale of individual particles. Two observations can be made: (1) the peak tensile displacement generally increases as particle size decreases; and (2) peak tensile displacement tends to be more uniform (constant) with changing saturation for larger particle sizes. The first observation is consistent with Rumpfs (1961) tensile strength theory for the pendular regime, which indicates that for a given degree of saturation, relatively small particles have greater tensile strength at higher separation distances. The second observation was also evident in the tensile strength measurements, which also show a more constant value as a function of saturation for the larger grain sizes. 4 CONCLUSIONS
Results obtained from tensile strength tests for loose and dense F-75 sand in terms of tensile displacement measured at failure as a function
Horizontal displacement (failure plane separation) at failure in tension exhibited double-peak behavior as a function of degree of saturation. For the dense specimens, the two peaks occurred at higher degrees of saturation than the loose ones. The second peak (higher degree of saturation) showed higher horizontal deformation at failure. The twopeak behavior tends to flatten out as the grain size increases. This behavior is consistent with that observed for tensile strength. This behavior is more common in granular soils (Baltodano-Goulding 2006) than in fine grain soils where the saturation mechanisms are more complex; however, it was observed as part of this research, in fine soils too. This observed reduction in shear strength, as the degree of saturation increases in the capillary regime and decreases in the funicular regime, could produce and unsafe condition, if this reduction is not analyzed or considered during the design stage. It is possible the deformation behavior is affected by an increase in normal effective stress due to a suction stress increment. This can produce that a
184
dense or overconsolidated specimen could behave as a loose or normally consolidated one. Wetter specimens behave as expected for loose condition specimens even though the density and normal stress has not changed. The observed deformation behavior can be affected by the saturation mechanisms related to each saturation regime and an effective stress change related to those mechanisms. This difference in behavior can be interpreted in terms of the Critical State Theory. Because effective stress varies with suction stress, which is a function of degree of saturation, it is expected that an increase in effective stress might be enough to relocate the normal stress to the contraction side of the critical state line. It is more likely for a specimen with a high normal stress to be relocated to the contraction side of the critical state line if an increase in suction stress is present. However, it is important to mention that an increase in effective stress due to an increase in suction stress for this range of stresses is small. This can lead to the conclusion that for unsaturated soils not only normal stress but also the structure and mechanisms associated with the saturation regimes have an effect in soil deformation behavior. It was observed that the peak tensile strength generally increases as particle size decreases; and its displacement tends to be more uniform for different degrees of saturation when particle sizes decrease. Suction stress has little effect for the studied granular soils in terms of shear strength, BaltodanoGoulding (2006). However, it was observed that its effect in the deformation behavior it is considerable even when this stresses are not big enough to alter
shear strength behavior. Soil deformation behavior can be affected by the degree of saturation and this can have a significant effect in Geotechnical Engineering projects where deformations play an important role such as embankments, excavations near other structures, tie-back walls or geosynthetics applications. REFERENCES
Atkinson, J.H. and P.L. Bransby. 1978. The Mechanics of Soils: An Introduction to Critical State soil Mechanics. Maidenhead, McGraw-Hill. Baltodano-Goulding, Rafael. 2006. Tensile Strength, Shear Strength, and effective stress for unsaturated sand. Doctotal Dissertation. University of MissouriColumbia. Baltodano-Goulding, R. and Korte-Leiva, D. (2009). Suction Stress Influence on Earth Retaining Structures GeoFlorida 2010 Conference. February 20 to 24, 2010. West Palm Beach, Florida, USA. Korte-Leiva, Diana. (2009). Suction Stress Influence on Earth Retaining Structures Graduation Project. Costa Rica Institute of Technology. Ladd, R.S. (1978). Preparing Test Specimens using Undercompaction. Geotechnical Testing Journal, pp. 1623 GTJDJ. March 1978. Mitchell, J.K. (1993). Fundamentals of Soil Behavior. John Wiley & Sons, Inc. New York. Terzaghi, Karl; Peck, Ralph, B. 1960. Soil Mechanics in Engineering Practice Wiley & Sons Inc. New Jersey. Rumpf, H. 1961. The Strength of Granules and Agglomerates. Agglomeration W.A. Knepper, Interscience: 379418. Wood, D.M. 1990. Soil Behaviour and Critical State Soil Mechanics. Cambridge, England, Cambridge University Press.
185
Collapse potential of a compacted weathered serpentinite from Skiros Island, Greece

M. Bardanis & S. Grifiza
Edafos Consulting Engineers S.A., Athens, Greece
ABSTRACT: The paper presents the findings of an experimental investigation of the collapse potential of a weathered serpentinite from Skiros island, Greece. Index properties, compaction characteristics and collapse properties of the material were obtained and are presented in the paper. The collapse potential was investigated by oedometer tests with wetting under constant vertical stress. As expected, the drier the soil compacted, the larger the collapse it exhibited upon wetting, even at densities higher than 95% of the maximum dry density according to Standard Proctor compaction test. Apart from the test results the paper presents also the experimental method used for quality control of the specimens prepared for collapse testing and the results of matric and total suction measurements on the compacted samples, immediately after compaction. 1 INTRODUCTION of disturbed material was brought to the laboratory of EDAFOS S.A., which was responsible for the experimental programme. The sample was homogenised in the lab and five samples were used to measure index properties. Table 1 summarises the results of these tests along with the average values and standard deviation for each property. According to USCS Samples 1 and 3 are characterised as low-plasticity clays, Sample 2 as low plasticity silt, Sample 4 as silty sand and Sample 5 as clayey sand. On average the five samples have a liquid limit of 27.4%, plasticity index of 8.4, and are characterised as CL according to USCS. Figure 1 shows Casagrandes plasticity chart with experimental points corresponding to the five samples used in the experimental investigation. As far as grain-size distribution is concerned, Samples 1 to 3 contain more silt than sand and a very small percentage of clay while Sample 4 contains more
Table 1. Summary of index properties of the weathered serpeninite from Skiros Island, Greece. Sample wL (%) Ip Gs Sand (%) Silt (%) Clay (%) USCS 1 27.6 10.0 2.58 44.8 53.1 2.0 CL 2 25.9 3.4 2.58 45.5 50.7 3.8 ML 3 28.0 11.9 2.60 44.0 52.8 3.2 CL 4 30.0 5.8 2.62 49.6 46.7 0.4 SM 5 25.3 11.0 2.66 59.3 39.3 1.4 SC Average 27.4 1.9 8.4 3.7 2.61 0.03 48.6 6.3 48.5 5.8 2.1 1.3 CL*
The use of a weathered serpentinite has been planned for the construction of the shoulders of the Ferekambos earth-dam on the island of Skiros, Greece. The Ferekambos dam is a 25 m high earthdam with a sandy clay core 5 m wide at the crest and 35 m wide at its widest point. The core is to be constructed of a sandy clay found as top cover in the location of the dam and the reservoir. The shoulders of the dam are to be constructed from weathered serpentinite found in abundance in the area of the dam immediately below the sandy clay and down to depths exceeding 25 m, as well as in the form of outcropping material on the valley sides. The nature of the weathered serpentinite and the form it is found in the field ready to collapse upon wetting or excavation and transport has called for an extensive experimental investigation of its collapse potential upon wetting after dynamic compaction, the results of which are presented in the paper. Several researchers have already investigated the collapse potential of materials used for the construction of earth-dam shoulders and especially rockfill (for instance Alonso & Oldecop, 2000), but collapse test results from serpentinites are scarce if any to the authors knowledge. 2 SAMPLING, INDEX PROPERTIES AND COMPACTION CHARACTERISTICS
As part of the investigation of the collapse potential of the weathered serpentinite a large quantity
* Theoretical, based on the average index properties.
187
60 50 Plasticity Index Ip 40 30 20 10 CL-ML 0 0 10 CL ML 20 30 OL/ML 40 50 OH/MH 60 70 80 90 100 CL CH A-Line
with standard compaction effort. An automatic compactor was used. Two compaction tests were performed in 4 diameter moulds. Optimum water content was found on average 11.1% and maximum dry unit weight 19.3 kN/m3. The Proctor compaction curves are presented in Figure 3. 3 SAMPLE PREPARATION AND TESTING METHOD
Liquid Limit wL (%)
Figure 1. Casagrandes plasticity chart with experimental points for weathered serpentinite (empty points) and their average (solid point).
100 90 80 70 Passing (%) 60 50 40 30 20 10 0 0.001 0.01 0.1 Sieve opening (mm) 1 10 1 2 3 4 5
Figure 2. Grain size distribution curves of weathered serpentinite.

21 Dry Unit Weight (kN/m3)
20
Sr=100
19
90
18
80% 30% 40% 50% 60% 70%
17 5 10 Moisture Content (%) 15
Figure 3. Standard Proctor compaction curves of weathered serpentinite.
sand than silt and even less clay. In Sample 5 sand is considerably more than silt. The grain size distribution curves of the five samples are presented in Figure 2. Unit weight-moisture relationships were obtained using the Proctor compaction test
The tests performed for the investigation of the collapse potential of the weathered serpentinite were oedometer tests with wetting under constant vertical stress. The samples used were dynamically compacted in the same automatic compactor that the Standard Proctor compaction tests were carried out. Prior to compaction all the material used was oven dried, broken and passed through ASTM sieve No 4. After compaction at each selected water content the compacted samples were extruded from the moulds using a hydraulic extruder and the specimens to be tested were carefully trimmed using the oedometer cutting ring which was slowly and carefully pushed in the compacted sample with the sides cut off by a sharp knife. As part of the quality control of the testing programme, trimmed samples were always visually inspected after the completion of the test for voids on the periphery of the samples as may in fact happen if the cutting ring is pushed too fast in a dynamically compacted sample. If such voids were observed at the end of the test after the specimen was taken out of the ring, the test results on that specimen were discarded and the particular test was repeated until the quality of the specimen was of the required level with no voids observed on the periphery. In Figures 4a and 4b a sample discarded due to the existence of voids and a sample with no voids on the periphery are presented respectively. The procedure followed was considered necessary in order to employ dynamic compaction with the same automatic compactor for the samples to be tested. Using static compaction as an alternative was anticipated to be unrepresentative of the as compacted state obtained from a Standard Proctor compaction test. Dynamic compaction by hand in the ring of each oedometer cell was also considered insufficient as it would not allow homogeneity from sample to sample both in terms of density and water content. Despite the problem of the voids on the periphery of the samples which may require repetition of several tests (in the experimental programme undertaken approximately 30% of the tests were repeated), the adopted procedure was anticipated to be the one ensuring maximum homogeneity between samples and compatibility with the compaction characteristics obtained from
188
Figure 4. a) Sample (one of 63.5 mm diameter in this case) discarded due to large voids on its periphery (indicated by dashed black lines), and b) sample (one of 50 mm diameter in this case) accepted as there are negligible macro-voids on its periphery.
a Proctor test. A small but critical detail is that the sections of the compacted sample used to trim the specimens must not contain the surfaces between different layers of soil compacted in the mould. Even if scarification of the surface of the previous layer has taken place (as suggested in available standards for laboratory compaction), still it will be very difficult to trim a specimen through the part containing the interface between consecutive layers without creating large voids on the periphery of the specimen. It must be noted however that the procedure for checking the quality of the samples on the basis of the existence of voids on the periphery after the completion of each test was only possible because the particular material undergoes relatively small deformation during testing. If it were a more compressible material (or the same material but in a looser state) this method could not have been adopted. Once each specimen was prepared, it was taken with the ring and placed in the oedometer cell. Lower and upper porous stones were dry before placing the rings containing the samples. Using wet saturated porous stones (as when performing a typical oedometer test) was anticipated to induce part of the collapse deformation of the sample due to wetting prior to the addition of water in the cell
at the required stress, especially in the case of the specimens with the lowest water content. On the other hand, possible initial drying of the specimens from coming in contact with the dry porous stones was considered negligible as the porous stones have too large pores to be able to sustain capillary forces large enough to draw water from the compacted weathered serpeninite. Had the material under consideration been coarser, this could have been a problem. Once the samples were placed in the cells and the cells assembled and placed on the oedometers, the upper opening of the cells was wrapped with cling film to minimise drying through the porous stones to the atmosphere in the laboratory environment. Still, some drying may actually have occurred as the cling film cannot seal completely the opening but this effect was the same on all samples with all water contents tested and so it was considered that it affected all the tests in the same way. Also a small increase of the target moisture content of the samples was implemented to counter for that effect. As far as one-dimensional consolidation is concerned, pneumatic oedometers were used with a capacity for maximum applied force of 15 kN and fixed-ring cells with lightly lubricated, smooth and polished inner surface rings, with a 63.5 mm or 50 mm internal diameter and 20 mm height. Maximum applied stress with the 63.5 mm diameter rings can be 4.74 MPa and with the 50 mm diameter rings 7.64 MPa. Compliance tests for loading were performed on all oedometers for each combination of oedometer and cell for the actual loading sequence to be followed, so that corrections could be applied to the deformation measurements. In addition to typical compliance tests for loading, special compliance tests were performed for the effect of wetting on the dry porous stones. As in the case of loading, where a compliance test shows how much of the deformation recorded at each load is actually due to specimen deformation and how much due to apparatus and cell deformation, the compliance tests for wetting were meant to show how much of the deformation recorded upon wetting is due to deformation of the porous stones and how much due to swelling or collapse of the sample itself. So, oedometer cells containing brass calibration disks in place of the soil specimens were assembled and placed on the oedometers with dry porous stones as would be the case during the wetting tests of the actual soil specimens. Vertical stress was raised to the required value and then the cells were filled with water. It was observed from the first three tests performed that even in the case of the highest stresses (7.6 MPa) the deformation recorded due to wetting of the porous stones was less than 0.002 mm (always positive) while at low
189
stresses it was obviously less than 0.0005 mm as the displacement gauges of 0.001 mm accuracy did not indicate any change. Given that this type of compliance test takes much more time than the regular compliance test for loading (as after wetting at each load the oedometer has to be unloaded, the cell disassembled and the porous stones left to dry completely before proceeding to the wetting at the next load) these results were considered enough and it was decided that compliance due to wetting of the porous stones would not be taken into account. Still it is recommended for laboratories engaged in swelling/collapse testing to perform such tests sporadically on the actual equipment used in each specific laboratory in case that wetting collapse of the porous stones may actually prove significant. Three types of specimens were prepared; at 2.5% dry of optimum moisture content, at optimum and at 2.5% wet of optimum. For each case of moisture conditions a deviation of 0.25% water content was considered acceptable. As far as dry unit weight was concerned it had to be higher than 95% of the maximum obtained from the Standard Proctor test. For each of the three moisture conditions one type of test was performed with 24 h time intervals between consecutive loads. That was the one that wetting took place only with the top cap load only. In the case of all other tests, the load was increased as soon as primary consolidation (plus a safe margin of deformation to make sure) of the material was completed. As a rule this lasted between 4 and 8 hrs depending on the load and moisture condition. The use of computer controlled pneumatic oedometers made it possible to programme such a loading sequence that could continue overnight without a need for personnel to be present in order to apply the next load. This decrease of time intervals between consecutive loads was considered necessary to ensure that by the time the specimen reached the vertical stress that wetting would take place, only the minimum amount of water would have escaped through the cling film due to evaporation in the laboratory conditions. The final stress of the loading sequence at which wetting would take place on each specimen was left overnight to ensure that any excess pore water pressures would dissipate for certain prior to wetting. Once consolidation at each stress that wetting would take place had completed, the cling film was removed from the opening of the cell and deaired, deionised water was added typically within 1015 s and the deformation occurring started to be recorded. Water was added until both the specimen and the top cap were fully underwater in the cell. Given that the compliance tests for wetting showed negligible deformation due to wetting of the dry porous stones, the deformation due to wetting was
taken as the difference between the initial reading immediately before wetting and the final reading after 24 hrs. The strain due to wetting was taken as the ratio of the deformation due to wetting over the height of the specimen at the stress prior to wetting, corrected for compliance of the apparatus. Part of the deformation recorded during this amount of time may actually have been creep due to the stress sitting on the specimen. Still for a time of only 24 hrs this was considered negligible. After wetting for 24 hrs the samples were unloaded to zero stress in steps twice as large as the ones during loading and then removed from the cells for calculation of final water content and final dry mass. Post-wetting loading behaviour at this stage of the research was not investigated. The results of the tests are presented and discussed in the next section. 4 RESULTS AND DISCUSSION
The strain induced due to wetting of the specimens is plotted against the vertical stress sitting on the specimen during wetting in Figure 5. Vertical stress is plotted in linear scale in Figure 5a and in logarithmic scale in Figure 5b. As expected, the drier the soil initially the larger the collapse due to strain at high vertical stress. For water content 2.5% dry of optimum, from 250 kPa of vertical stress the collapse strain started to increase rapidly up to 4000 kPa where a tendency for stabilization of the collapse strain with vertical stress is observed. Maximum collapse strain reached 6.4%. Zero deformation upon wetting was observed at 125 kPa. For lower stresses swelling was observed but its maximum value was only 0.19% for the minimum applied stress of 8 kPa. For the optimum water content a similar behaviour was observed although the increase in collapse strain is less rapid and the maximum collapse strain less than in the previous case (it reached a maximum value of 2.7%). The rapid increase started between 500 and 750 kPa and lasted up to 4000 kPa where stabilization of the collapse strain occurred again. Zero deformation upon wetting was observed at 225 kPa. For lower stresses swelling was observed but its maximum value was only 0.34% for the minimum applied stress of 8 kPa. Finally for water content 2.5% wet of optimum, the collapse strain observed even at vertical stress as high as 7600 kPa was practically negligible (the maximum value obtained was 0.07%). In this particular case of so low collapse strains, it may be argued that there was essentially no strain due to wetting as the deformation observed after wetting for 24 hrs was only secondary compression deformation of the sample (especially at these high stresses). Zero deformation upon wetting
190
was observed at 500 kPa. For lower stresses swelling was observed but its maximum value was only 0.09% for the minimum applied stress of 8 kPa. If swelling pressure is defined as the vertical stress sitting on a specimen at which neither swelling nor collapse will occur upon wetting then the values corresponding to each moisture condition are: 125 kPa, 225 kPa and 500 kPa for dry of
(a)
0 -1.0 0.0 Swelling/Collapse (%) 1.0 2.0 3.0 4.0 5.0 6.0 7.0 Optimum-2.5% Optimum Optimum+2.5% Vertical Stress (kPa) 2000 4000 6000 8000
(b)
1 -1.0 0.0 Swelling/Collapse (%) 1.0 2.0 3.0 4.0 5.0 6.0 7.0
Vertical Stress (kPa) 10 100 1000 10000
Optimum-2.5% Optimum Optimum+2.5%
(c)
1 -0.5 0.0 0.5 1.0
Vertical Stress (kPa) 10 100 1000
optimum, optimum and wet of optimum respectively. Given the higher density of the compacted material at the optimum moisture content, the larger swelling pressure of the samples compacted at this moisture content than that of the samples compacted at moisture content dry of optimum is understandable, even though the latter have higher initial suction and lower degree of saturation. Still, the higher value of the swelling pressure obtained for the samples compacted wet of optimum seems to make no sense. If the values of the observed swelling and collapse strains are taken into account it may be attributed to the fact that the material at this moisture condition has saturated due to loading under constant moisture content. This point is further discussed below. Prior to testing, the total and matric suction was measured in samples of each moisture condition immediately after compaction using a chilled mirror hygrometer with temperature correction and a laboratory tensiometer with 100 kPa airentry pressure ceramic tip respectively. The results of these measurements are presented in Table 2. Several measurements of matric and total suction were made on the same sample prepared at each moisture content. Measurements included both different points on the extruded compacted sample and several measurements at each point. The chilled mirror hygrometer was calibrated and also prior to each measurement buffer KCl solutions of 0.2M concentration were used to ensure correct total suction measurements. As expected, initial matric suction immediately after compaction increased with decreasing moisture content with a similar but not quite as accurate trend for total suction. Generally osmotic suction (defined as the difference between total and matric suction) would be expected to be the same, irrespective of compaction moisture. On the other hand the measurement of total suction remains tedious even when using a chilled mirror hygrometer. Also, for compaction
Table 2. Initial suction measurements on compacted samples of weathered serpentinite. Moisture condition Dry of optimum Wet of Optimum optimum
Swelling/Collapse (%)
Optimum-2.5% 1.5 2.0 Optimum Optimum+2.5%
Distance from optimum moisture content 2.5 0.25% 0 0.25% 2.5 0.25% Total suction 384 46 (kPa) Matric suction 69 2 (kPa) Osmotic 314 47 suction (kPa) 234 27 30 5 205 25 239 30 12 2 227 31
Figure 5. Strain due to wetting versus vertical stress prior to wetting at each of the three moisture contents tested for vertical stress in a) linear scale, b) logarithmic scale, and c) logarithmic scale but only in the range 1 to 1000 kPa.
191
moisture at optimum and at wet of optimum conditions the tensiometer responded fairly rapidly and maintained a constant reading for several hours indicating correct and accurate measurement. In the case of dry of optimum compaction moisture, the instrument responded rapidly at first and then quite slowly until it reached a practically constant value after almost 24 h. The long equilibration time may indicate that the true matric suction for this moisture condition is perhaps higher. Both these characteristics (accuracy of the chilled mirror hygrometer and response behaviour of the tensiometer for dry of optimum compaction moisture) may explain the higher osmotic suction obtained for this moisture condition. The osmotic suction for optimum and wet of optimum compaction moisture seems well defined at approximately 200 kPa. Still for the dry of optimum compaction moisture a value of approximately 300 kPa was measured. In any case it may be stated that the osmotic suction of the compacted serpentinite mixed with distilled deionised water is fairly well defined between 200 and 300 kPa. The matric suction is considered accurately measured for the optimum and wet of optimum compaction moisture but may have been underestimated for the dry of optimum compaction moisture. Perhaps a tensiometer with higher air-entry pressure ceramic tip will allow more accurate measurements in this range of moisture in the future. These measurements of initial suction may explain the observed behaviour at each moisture condition regarding the obtained swelling pressure value. A matric suction of 12 kPa can be assumed to decrease practically to zero under constant water content oedometer tests as is the case prior to reaching the stress at which wetting will occur, even for a material of low compressibility as compacted weathered serpentinite. On the other hand, exactly because the material has low compressibility it may be postulated that the suction measured at optimum and dry of optimum moisture conditions have not decreased considerably. As a result, in the case of moisture conditions at optimum and at dry of optimum, collapse and swelling observed reflect the increase in degree of saturation at each density. At wet of optimum moisture conditions however whatever swelling or collapse is observed reflects a combination of the presence of whatever clay minerals in the soil andin the case of collapse at large stressesthe secondary compression of the material during the further 24 h the load spent on the specimen during wetting. This could be an explanation for the high value of swelling pressure obtained for the compacted material in moisture conditions wet of optimum. It is planned that samples of weathered serpentinite will be tested for the clay minerals they contain.
CONCLUSIONS
The weathered serpentinite to be used in the Ferekambos Dam exhibited considerable collapse due to wetting at high stresses both in the case of specimens prepared 2.5% dry of optimum water content and in the case of specimens prepared at optimum water content. Specimens prepared at 2.5% wet of optimum essentially did not exhibit any collapse at high vertical stresses. Rapid increase of collapse strains starts at 250 kPa for moisture conditions dry of optimum and at stresses between 500 and 750 kPa at optimum. The latter stress is marginally higher than the maximum anticipated vertical stress in the shoulders of the Ferekambos Dam where weathered serpentinite is going to be used which is only 550 kPa (prior to impoundment). These results indicate that the material to be used during construction should be strictly compacted at moisture conditions wet of optimum, as even at optimum there is a possibility for large collapse strains to occur at the layers with the higher vertical stress anticipated in the shoulders of the dam. Increased density achieved with field compaction equipment (given the higher compaction energy) and field optimum moisture content decreasing due to higher field compaction energy work in favour of collapse strains njot appearing even at the maximum vertical stress anticipated in the dam. In any case, strict control of compaction moisture should be applied to ensure that compaction moisture will be higher than at least 1% wet of optimum, so that considerable collapse strains do not appear in the layers with the maximum anticipated stress after impoundment of the dam. In all moisture conditions where collapse appeared, collapse strains due to wetting seem to stabilise for stresses larger than 4000 kPa. This is especially the case for moisture conditions at optimum and at dry of optimum where large collapse strains were observed. ACKNOWLEDGEMENTS Owner of the Ferekambos Dam on Skiros island is the Agricultural Development and Foods Ministry of Greece. Prime contractor for the dam project was Elliniki Hydrokataskevi S.A.. The authors wish to acknowledge the help received from the Edafos S.A. Lab technicians, Mr C. Vagenas and Mr M. Tsoukaladakis. REFERENCE
Alonso, E.E., Oldecop, L.A., 2000. Fundamentals of rockfill collapse, Rahardjo et al. (eds), Unsaturated Soils for Asia, A. A. Balkema, Rotterdam, pp. 313.
192
Collapsibility study for tropical soil profile using oedometric tests with controlled suction
J.C.B. Benatti & M.G. Miguel R.A. Rodrigues & O.M. Vilar
The State University of Campinas, Campinas, Brazil University of So Paulo, So Carlos, Brazil
ABSTRACT: The aim of this research is to present and discuss the results of oedometric tests under controlled matric suction, for lateritic soil samples from an unsaturated soil profile, typical of 14% of the region of Campinas, in the State of So Paulo, Brazil. The soil profile consists of colluvial, lateritic and collapsible silty clay, with approximately 6.5 m thickness. The water level is approximately 17 m. The textural, structural and compositional characteristics of the soil profile demonstrate two distinct sources in the upper layer (from 1.5 m to 5.5 m) as well as in the lower layer (at 6.5 m). The former has a sedimentary source and the later has an igneous and/or metamorphic source. However, in the lower layer, the structural characteristics of the rock source prevail over the upper layers, with the evident presence of igneous origin minerals. The hematon/goetan argilon aggregates are the most common, followed by ferruginous concretions, magnetite grains and pyroxene pseudomorphs, suggesting a source near the Diabase. All these components are involved in a matrix with clay minerals, oxides and iron hydroxide and an expressive quantity of voids. Although there is the presence of cimentation, from iron oxides, the particles, surrounded or not by hematite and goethite, and microagregates are linked by capilarity forces. This fact indicates that the soil suffers disaggregation when wet. Disturbed and undisturbed samples were collected from this profile, at a 1.5 m depth. Six specimens were made from the undisturbed samples and submitted to oedometric tests. The specimens were wet and submitted to axis-translation technique, with pressure equipment and a pressure control panel. The imposed matric suction values were: zero, 25 kPa, 50 kPa, 100 kPa, 200 kPa and 400 kPa. The levels of net stress applied were 6 kPa, 12.5 kPa, 25 kPa, 50 kPa, 100 kPa, 200 kPa, 400 kPa, 800 kPa and 1600 kPa. Experimental results, obtained by oedometric tests, indicate collapse index varying between 2% and 18.7%; indicating collapsibility. Moreover, the experimental results were inserted into a theoretical model to estimate the LC curve and evaluate its application on this lateritic soil.
INTRODUCTION
Tropical soils have certain peculiarities that differ from soils originating from temperate climates. The typical climatic conditions of tropical regions lead to the formation of soils such as lateritic soils, which are characterized in their formation by the intense migration of particles under the action of infiltration and evaporation, resulting in a porous superficial horizon, where the most stable minerals such as- quartz, magnetite, ilmelite and kaolinite almost exclusively remain. Clay and silt particle aggregation is common in these soils due to the action of iron and aluminum oxides and hydroxides, which gives lateritic soil the mechanical and hydraulic behavior characteristics which are not appropriate for its texture. (Committee on Tropical Soils of ISSMFE, 1985; Gidigasu, 1976).
Some unsaturated soils are characterized by a further, sudden and large vertical displacement suffered, when subjected to an increasing degree of saturation, under almost constant total stress, caused by the collapse of soil structure. Therefore, such soils are usually called collapsible. Jennings and Knight (1957) provide a hypothesis for the collapse phenomenon: the structure of a collapsible soil to be loaded in its natural moisture content, compresses gently, without substantial change in volume, resisting compression efforts between the grains, without large relative movement between them. But when this loaded soil is wet and reaches a critical moisture content, the resistance bonds weaken and the structure collapses. The grains or micro-aggregates are kept stable by the presence of suction (capillary forces and adsorption) and/or cementing agents (oxides,
193
hydroxides and/or aluminum and carbonate). When wet these soil structure stabilizing forces lose intensity, allowing the relative displacement between particles, so that they start to fill the soil structure voids (Dudley, 1970). Collapse study in tropical soils is usually carried out using single or double conventional oedometer tests, without taking into account the influence of matric suction on soil behavior. The use of oedometer tests with suction control is vitally important for collapse study in unsaturated soils. By using this technique, it is possible to predict soil behavior for known values of matric suction, eliminating the problem of not knowing the initial moisture content of the specimen, which is a value that significantly influences the increase of specimen collapse. Conventional oedometer tests performed on studied soil showed that it is a collapsible soil when tested using field moisture. The study of thin and impregnated soil plates at various depths showed that the link between grains of soil is basically due to the presence of matric or adhesiveness without the presence of cementations that may help stabilize the soil structure. Thus, due to the great importance of matric suction on the mechanical behavior of the studied material, an option was made to use oedometer testing with suction control, making it possible to verify the real contribution of this variable in the resistance of the material. Therefore, this paper analyzes the results of oedometer tests carried out with suction control for the superficial layer of the colluvial and lateritic soil profile at the Experimental Field of Soil Mechanics and Foundations, UNICAMPEFSMF, located in Campinas / SP, Brazil, and aims to assess the influence of matric suction in the hydro-mechanical behavior of this soil and to verify the applicability of the elastoplastic model proposed by Alonso et al (1990) for the samples tested. 2 2.1 MATERIALS AND METHODS
n 68.2%, degree of saturation, Sr 33.7%, liquid limit, LL 48% and plasticity index, PI 10.7%. Figure 1 shows the granulometric curves obtained from tests with and without the use of deflocculants during the sedimentation phase of the joint granulometric test, and Figure 2 shows the retention curves of water in the soil for soil profiles studied, obtained by the axes translation technique and presented by Miguel & Vilar (2009). As can be seen from the results obtained, the soil has high porosity, typical of laterized tropical soils. The granulometric curves of Figure 1, analyzed together, indicate the presence of a structure with micro-fine particle aggregates, forming clumps (or flakes). Fine particles aggregate due to physicochemical attractions and cementations present in the soil profile, due to the leaching processes that occurred. The retention curves show a soil with very low air intake values, around 1 kPa. Figure 3 shows a photograph obtained from thin impregnated plates carried out for the studied soil in an undisturbed condition. There is, as shown previously, a large amount of pores from these soils which are in general effective, or in other
100% 90% 80%
ND - z=1.5m
Percent by mass
70% 60% 50% 40% 30% 20% 10% 0% 0.001 0.010 0.100
WD - z=1.5m
1.000
10.000
Grain-size (mm)
Figure 1. Granulometric curves obtained with the use of deflocculants (WD) and with no use of deflocculants (ND).
Gravimetric Water Content (%)
Soil studied
50
Wetting z = 1.5m
The studied soil was removed from the Experimental Field of Soil Mechanics and Foundations of UNICAMPEFSMF. The soil profile of EFSMF, up to 6.5 m deep, has a layer of silty clay-sandy, porous, of colluvial origin, composed predominantly of clay minerals like kaolinite, quartz and iron oxides (hematite) probably acting as cementations between the grains of clay and sand (Miguel & Vilar, 2009). The main features of the material studied are: specific gravity natural nat 11.8 kN/m3; specific weight of solids, s 29.89 kN/m3; field moisture content, w 24.2%, voids, e 2.14, porosity,
45 40 35 30 25 20 0.01 0.1 1 10 100 Matric Suction (kPa)
Wetting z = 2.5m Wetting z = 3.5m Wetting z = 4.5m Drying Drying Drying Drying z = 1.5m z = 2.5m z = 3.5m z = 4.5m
1000
10000
Figure 2. Curves of water retention for the soil profile studied, according to Miguel & Vilar (2009).
194
Figure 3. Impregnated thin plates obtained for the studied sample. (Miguel & Vilar, 2009).
the implementation of suction control. So as to carry out the test, the saturated specimen is brought into contact at the base, with the membrane or the high air entry value porous plate, and a common porous plate is placed on top. Air pressure is applied in the chamber and is expected to stabilize specimen water output. Loading is carried out conventionally, and suction is kept constant, using this technique for the translation of Hilf axes (1956). The technique proposed by Hilf (1956) uses a strategy to apply air pressure higher than atmospheric pressure. After adjusting the pressure, the difference between the applied air pressure and water pressure (ua-uw) remains the same as it had before the application of pressure in the chamber, with the advantage of not having more negative water pressure. 2.3 Model used
words communicating. The presence of quartz grains surrounded by hematite material, forming lumps was also noted. Cementations due to the presence of iron oxides, appears forming clumps, but it does not keep them connected, showing that the soil after undergoing the process of hemispheric dominance, suffered transportation, which broke the connection between the lumps, but maintained them in loco. The transportation hypothesis of the material is also maintained due to the hexagonal shape of the grains. The soil structure is stabilized by the presence of a suction force, also called adhesiveness, which gives the sample a meta-stable equilibrium condition. The presence of a structure under meta-stable equilibrium, with high porosity and in a non-saturated condition makes it obvious that this is a potentially collapsible soil. Simple oedometer tests, performed without suction control, showed that it is a highly collapsible soil when wet with field moisture, and loaded. To quantify the collapse of the soil, the formulation proposed by Jennings & Knight (1975) was used, which defines the potential collapse (PC) as the ratio between the variation of voids before and after the wetting, and initial void index of the specimen, in percentage. For simple oedometer tests, performed without suction control, PC values in the order of 20.0% were obtained. 2.2 Oedometer tests
Alonso et al (1990) proposed a model called the Barcelona Basic Model (BBM). The model was developed from a study in which different characteristics of unsaturated soils are analyzed in a joint way (Alonso et al, 1987). According to Alonso et al (1987), collapsible soils display behavior if their humidity is low. When samples of these soils are subjected to different suctions and each one to compression, several curves of stress-strain will emerge with their respective over-consolidation stresses The values of these over consolidation stresses, once united on the plane (p, s), resulted in points which, when interpolated, will generate a curve as shown in Figure 4. The points (p0*)1 (p0*)2 and (p0)1 and (p0)2, represent the over-consolidation stresses 1 and 2, the saturated and unsaturated soils, respectively. The A1B1 and A2B2 curves are called plastification or LC yield curves (loading-collapse), because they represent irreversible volumetric paths for L and C, i.e. for loading and collapse.
Oedometer tests with controlled suction were carried out using an oedometer chamber similar to the one developed by Escario & Saez (1973). This chamber is made up of a high air intake, porous stone adjusted to the base of the equipment for
Figure 4. Loading and wetting stress path in the plane (p, s) (Alonso et al, 1987).
195
The two paths (loading or collapse) shifted the plastification curve A1B1 to A2B2. The translation of the curve implies in an increase of the elastic region, because the soil accumulates plastic deformation along L or C paths. In the structural arrangement of Figure 4, both L and C will lead the soil to the same compaction state i.e., shift the A1B1 plastification curve to A2B2, by changes in load or suction in the soil. Equations 1, 2 and 3, presented by Alonso et al (1990) define the plastification curve proposed by Alonso et al (1987).
* ( s ) p0 p0 s c pc p s ( s ) (0) r (1 r ) e
1 0.9 0.8 0.7
e/e0
0.6 0.5 0.4 0.3
( 0 )
(1)
1
0 kPa 25 kPa
10
50 kPa
100 -ua(kPa)
100 kPa 200 kPa
1000
400 kPa
(2) (3)
Figure 5. Compressibility curves obtained from oedometer tests with controlled suction.
12 10 8
PC (%)
r
(s ) (0)
where s matric suction, the recompression straight slope of the saturated soil; s the recompression straight slope of unsaturated soil, (0) the virgin straight slope of saturated soil, (s) the virgin straight slope of the unsaturated soil, the parameter that controls soil stiffness, p0 the over-consolidation stress of unsaturated * soil; p0 over-consolidation stress of saturated soil and pc reference stress. 3 PRESENTATION AND RESULTS ANALYSIS
6 4 2 0 0 200 400 600 800

-ua(kPa)
25 kPa 50 kPa 100 kPa 200 kPa 400 kPa
1000
1200
1400
1600
Figure 6. suction.
Values of potential collapse depending on
The results obtained for the oedometric tests carried out with suction control are presented in figure 5. The void index values were normalized (e/ e0, where e is the void index value at the end of the stage and e0 is the initial void index value of the sample), so as to make the samples uniform. The collapse potential for suction control tests is defined based on the methodology developed for double oedometric tests, in other words, as the difference between the compressibility curve obtained for known suction values and the compressibility curve obtained for the wet soil. It can be seen, in Figure 5, that the compressibility of the soil depends on the matric suction value imposed, and therefore the initial moisture content of the sample. It can also be seen that there is little difference between the curves, for suctions over 25 kPa. Observing the retention curve of this soil (Figure 2), for suction values between 40 kPa and 1000 kPa, there is little variation in moisture
content for major variations in matric suction. Therefore, a small change in water content leads to a considerable increase of matric suction in the soil, within this range. On the other hand, for suction values ranging between 0 kPa and 25 kPa, a small increase in suction causes a sharp decrease in the moisture content of soil. This may explain the difference between the compressibility curves for the 0 kPa and 25 kPa suction values. Figure 6 shows PC values as a function of matric suction for the soil studied. It indicates that the PC values are greater, the higher the matric suction values applied, and therefore, the smaller the moisture content values of the samples. The maximum PC values occur for vertical tension around 200 kPa. When vertical stress increases a lot, the PC seems to tend towards a value close to 3.0%, regardless of the values of the matric suction applied. The soil is considered problematic in relation to collapse, according to Jennings & Knight (1975),
196
for practically all vertical stress and suction values. However, it can be seen that there is a big difference in the PC values found for the testing carried out with 25 kPa and 400 kPa. This shows the importance of oedometer testing with suction control. It is possible to classify a soil as non-collapsible in oedometer testing, which does not take this variable into account, even if it is for certain soil conditions. It can be stated that the collapsibility is not a soil characteristic, but a soil condition, since its potential is directly linked to the moisture content of the soil, which can be modified. Virtual pre-densification stress values can be obtained for the compressibility curves presented in Figure 5. The pre-densification stress in tropical soils is called virtual, as it does not reflect the maximum stress that the soil has already been submitted to, such as in temperate climate soils. Vargas (1973) states that the pre-densification stress value in tropical soils indicates a rupture of cementations, and therefore should not be mistaken for the pre-densification stress of temperate climate soils. Figure 7 shows the variation of virtual pre-densification stress with suction. Increase of virtual over-consolidation stress of the soil due to the increase of matric suction applied can be verified. This behavior shows that, as previously presented, the structure of these soils is supported mainly by adhesiveness (suction), and not so much by the presence of cementations bonds between the grains. Figure 8 shows the values of the stiffness parameters () for each value of applied matric suction. This parameter is obtained from the approximation by a straight of virgin stretch, in oedometric tests. Parameter is an indication of the soil stiffness, since it is defined from the virgin slope. A decrease in the stiffness value with the increase of applied matric suction can be seen. This behavior was also
0.21
0.19 (s)
0.17
0.15
0.13 0 100 200 300 400 500 Matric Suction (kPa)
Figure 8. Variation of parameter (s) with suction.
observed in other tropical soils (Futai & Almeida, 2002, Machado, 1998; Vilar & Davies, 2000). The model proposed by Alonso et al (1990) foresees the reduction of parameter (s) with increasing suction and requires r 1.0. The value of r calculated from the experimental results of (s) for this soil is r 1.49, so the model should not show a good fit to experimental data. Thus, it can be said that the Barcelona Basic model does not foresee the behavior of this soil. Other authors (Futai & Almeida, 2002, Machado, 1998) also noted the inapplicability of the BBM model to the tropical soils that they studied due to the stiffness parameter, and proposed changes to foresee the decrease in stiffness with the increase of matric suction. Thus, the models proposed by Futai & Almeida (2002) and Machado (1998), modified by Alonso et al (1990) can foresee both increased and decreased stiffness with increasing suction and should present a more coherent setting for the soil studied. 4 CONCLUSIONS
400
300
200
100
0 0 50 100 -ua(kPa) 150 200
Figure 7. Virtual pre-densification stress variation with matric suction.
The following conclusions can be reached based on this work: It was found that the compressibility of the soil is proportional to the imposed matric suction, and consequently the moisture content of the sample. The higher the values of potential collapse PC, the higher the values of matric suction applied, reaching values of about 11.5%. There was a tendency to have maximum PC values near vertical stress of about 200 kPa, apart from obtaining a constant PC value for high values of vertical stress. There was an increase in virtual over consolidation stress with increasing matric suction, indicating that the structural stability of the soil is due to that variable. Soil stiffness tends to decrease with the increasing matric suction imposed, contrary to the theory of the model proposed by Alonso et al (1990).
197
ACKNOWLEDGMENTS The authors would like to express their sincere thanks to FAPESP (Fundao de Amparo Pesquisa do Estado de Sao Paulo) for their support towards research 2005/00164-7 and 2008/00016-6 and the master scholarship 2008/02162-0. REFERENCES
Alonso, E.E.; Gens, A.; Hight, D.W. 1987. Special Problem soils. Proc. 9th E.C.S.M.F.E. Dublin General Report, Section 5, 5.15.60. Alonso, E.E.; Gens, A.; Josa, A. 1990. A constitutive model for partially saturated soil. Gotechnique 40, n 3, p. 405430. Committee on Tropical Soils of ISSMFE. 1985. Peculiarities of Geotechnical Behavior of Tropical Lateritic and Saprolitic Soils Progress. Edile: So Paulo/SP. Dudley, J.H. 1970. Review of collapsing soils. Journal of the Soil Mechanics and Foundation Division. ASCE, Vol. 96, n. SM3, p. 925947, may, 1970. Escrio, V. & Sez, J. 1973. Measurement of the Properties of Swelling and Collapsing Soils under Controlled Suction. 3th I.C.E.S., Haifa, 195200. Futai, M.M. & Almeida, M.S.S. 2002. Collapsible Soil: a Theoretical and Experimental Studies. The Electronic Journal of Geotechnical Engineering. EUA.V.7. 2002. Gidigasu, M.D. 1976. Laterite Soil Engineering: Pedogenesis and Engineering Principles. Developments in Geotechnical Engineering, 9. Elsevier Scientific Publishing Company.
Hilf, J.W. 1956. An investigation of Pore-Water Pressure in Compacted Cohesive Soils. Technical Memorandum N. 654, Bureau of Reclamation U.S. Department of Interior, Denver, Colorado. Jennings, J.E. & Knight, K. 1957. The additional settlement of foundations due to a collapse of structure of sandy sub soils on wetting. In: IV International Conference On Soil Mechanics And Foundation Engineering. London. Proceedings. v.1, p. 316319, 1957. Jennings, J.E. & Knight, K. 1975. A Guide to Construction on or with Materials Exhibiting Additional Settlement due to a Collapse of Grain Structure. Proc. 4th Regional Conference for African on Soil Mech. Found. Eng., Durban, 99105. Machado, S.L. 1998. Apply of Elastoplasticity concept to Unsaturated Soils. Thesis (in Portuguese). EESC/ USP. 1998. Miguel, M.G. & Vilar, O.M. 2009. Study of Water Retention Properties of a Tropical Soil. Can. Geotech. J., 46, 2009. Vargas, M. 1973. Structural Instable Soils in Southern Brazil. Proc. 8th I.C.S.M.F.E, Moscow, vol. 2, 239246. Vilar, O.M. & Davies, G.I. 2002. Collapse Behavior Analysis of a Clayey Sand Using Different Testing Procedures. In: 3a Intern. Conference on Unsaturated Soils, 2002, Recife. Proc. 3a Intern. Conference on Unsaturated Soils. Amsterdam: Balkema, 2002. v. 2, p. 571576.
198
Stress path effects on shear stiffness, damping ratio and volumetric behaviour of normally consolidated unsaturated compacted clay
M. Biglari
International Institute of Earthquake Engineering and Seismology (IIEES), Tehran, Iran
A. dOnofrio & C. Mancuso M.K. Jafari & A. Shafiee
Dipartimento di Ingegneria Idraulica, Geotecnica ed Ambientale, University of Naples Federico II, Italy International Institute of Earthquake Engineering and Seismology (IIEES), Tehran, Iran
ABSTRACT: An experimental study has been carried out to investigate the effects of stress path on initial shear stiffness, G0, initial damping ratio, D0, and volumetric behaviour of Zenoz kaolin, an unsaturated silty clay, tested both in normally consolidated and over-consolidated conditions. To this purpose, a fixed-free Resonant Column-Torsional Shear device (RCTS-ns) working in suction controlled conditions has been used. Specimens have been compacted at the same void index by wet tamping, using the undercompaction technique. G0 and D0 have been measured almost continuously along complex stress paths. Similar values of initial shear stiffness, initial damping ratio, specific volume (v), and specific water volume (vw), are obtained in normally consolidated state when a given value of mean net stress and suction are reached after different stress paths. 1 INTRODUCTION initial soil stiffness, G0, and damping ratio, D0, on different low plasticity soils, along various isotropic stress paths in the mean net stress,( ua), suction,(ua uw), plane. The project focused on this aspect because a strong link was expected among stress, volumetric state, volumetric history and shear stiffness, as for saturated materials. Vassallo et al. (2007b) focused on modelling the G0 variation with ( ua), (ua uw) and their histories. Pineda et al. (2008) presented results of tests performed by two small-strain dynamic equipments, a resonant column device, and bender elements system. The initial shear modulus was continuously monitored along drying and wetting paths using bender elements installed inside a desiccator. These results were then compared with that measured in resonant column: G0 measured by RC tests result higher than that measured by bender elements, probably because affected by total suction effects. Ng & Yung (2008) carried out isotropic compression tests and measured the multidirectional shear wave velocities using modified unsaturated triaxial apparatus equipped with 3 pairs of bender elements. They found that shear wave velocities increases less than linearly with mean net stress, shear wave velocities also increases non-linearly with matric suction. Xu et al. (2008) extended the work of Ng & Yung
Shear modulus and damping ratio are key parameters for predicting the dynamic response of soils, for the design and the analysis of earthquakeresistant foundations, vibrating machine foundations, and for the analysis of static and dynamic soil-structure interactions. Although measurements of shear modulus and damping ratio by resonant column tests have been performed since several decades, little attention has been devoted to the small strain behaviour of unsaturated soils and literature works based on suction controlled tests are relatively recent. Cabarkapa et al. (1999) measured the initial shear modulus by bender element installed in a triaxial cell working in suction-controlled conditions. The data pertaining to normally consolidated samples showed a G0 increase as suction increases. To analyze the effects of suction and compaction variables on the mechanical response of soils a broader project was established at Universit di Napoli Federico II, where a resonant columntorsional shear device (RCTS-ns) working under controlled suction conditions has been developed. Mancuso et al. (2002); Vassallo (2003), Vassallo et al. (2007a), & dOnza (2007), measured
199
(2008) and studied the effects of drying and wetting paths at constant mean net stress using the same device. In their study, drying and wetting effects on multidirectional shear wave velocities and degree of stiffness anisotropy were investigated. Similarly to what happen for the Soil Water Retention Curve, they observed a hysteretic behaviour of the shear wave velocities measured at different matric suction along drying and wetting paths. According to previous literature studies (dOnza et al. 2008), the effect of suction on the initial shear stiffness of soils seems to be significantly affected by the soil plasticity. Since there were no accurate shear stiffness data on medium to high plasticity soils, Biglari et al. (2009) carried out some resonant column tests in suction controlled condition on medium plastic mixed materials in order to quantify the effect of soil plasticity on G0. This paper show some of the experimental results collected during this wide experimental programme, carried out on compacted Zenoz kaolin using a resonant column device working in suction controlled condition. The experimental measurements have been performed at the soil dynamic laboratory of University of Naples Federico II, Italy, as a part of a project under way at International Institute of Earthquake Engineering and Seismology (IIEES), Iran. The experimental results showed in the paper are aimed to illustrate the influence of stress path on the initial shear stiffness and damping ratio taking into account for the volumetric behaviour.
it to a virgin state at a relatively low stress levels. Quite homogeneous initial testing conditions have been achieved (see Table 1). 2.2 Experimental device
Tests were performed using a suction controlled RCTS-ns developed at the University of Naples Federico II. This apparatus independently controls the cell, pore-water and pore-air pressures, by using three electro-pneumatic regulators. The suction control is based on the axis translation technique, with a maximum applicable suction of 500 kPa. Water and air pore pressures are both applied at the base pedestal in order to not modify the fixedfree torsional restraint condition typical of RC tests. The total volume changes are measured by monitoring the axial and average radial strain, respectively by an LVDT coaxial to the specimen and by a differential pressure transducer (DPT) gauging the changes in the level of a water bath surrounding the specimen. A system composed of two doublewalled burettes connected to another DPT monitors the water content changes in the soil. Details on the device are reported in Vinale et al. (1999), Mancuso et al. (2002), and Vassallo (2003). A high capacity tensiometer (HCT) designed in University of Napoli Federico II (Rojas et al., 2008) was also used in order to measure the after compaction suction on a specimen compacted in a way identical to that used with the RCTS-ns. 2.3 Experimental program
TESTED MATERIAL, EXPERIMENTAL DEVICE AND EXPERIMENTAL PROGRAM
2.1 Tested material A commercial Iranian kaolin, Zenoz kaolin (ZK) was adopted as testing soil; it is a silty clay extracted from a mine in North West of Iran. The soil has a clay content of about 18% and a silt fraction of about 60%. It is an inorganic silty clay with a plasticity index of 12%. The soil powder was initially dried in a desiccator with silica gel. Then cylindrical samples, 36 mm in diameter and 72 mm in height, were prepared in a mould by compacting soil at a water content of 11.9% (i.e., 3.5% less than the optimum water content of the standard Proctor compaction test) and dry density of 12 kN/m3. All samples were compacted in seven layers by moist tamping, using the under-compaction technique (Ladd 1978). This procedure resulted in a specific volume of 2.17 and a degree of saturation of 27%. The intention of the adopted preparation method was to produce samples with a low value of after-compaction yield stress, in order to bring
The tests here discussed are reported in Table 1. The data presented in this paper refers to three different tests: ZK1 at 50 kPa, test ZK2 at 150 kPa, and tests ZK3 at 300 kPa of suction. During these tests, after a preliminary equalization stage, each specimen was isotropically consolidated at constant suction while measuring the initial shear stiffness and initial damping ratio almost continuously by resonant column tests at very small strain. Two tests also included stages of wetting at constant mean net stress (ZK2 and ZK3). The mentioned stress path allowed to eliminate the over consolidation
Table 1. sample
Experimental program and initial properties of d Initial(s) ( ua) (s)history vw (kN/m3) (kPa) (kPa) (kPa) 50 150 300 450 350 350 none 50 150 50
Tests (%) v
ZK1 12.0 2.15 1.3 12.0 ZK2 11.9 2.17 1.3 11.9 ZK3 12.0 2.17 1.3 11.9
200
400 350
ZK1 ZK350-300 B ZK250-150 ZK2350-150 ZK3350-300 ZK2 ZK3
(a)
Change in specific water volume vw
0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0.00 -0.01 -0.02 0 0.01 0.00 -0.01 -0.02 -0.03 -0.04 -0.05 -0.06 -0.07 -0.08 -0.09 0 20 40 60 80 100 120 ZK1 ZK2 ZK3 20 40 60 80 100 120 ZK1 ZK2 ZK3
Suction (kPa)
300 250 200 150 100 50 0 0 A
ZK3350-150
ZK150-50 50 100 150 200 250
ZK2350-50 ZK3350-50 ZK3 450-50 ZK1450-50 ZK1

350-50
300
350
400
450
500
Mean net stress (kPa)
(b)
Change in specific volume v
Time (hr)
Figure 1.
Stress Path.
effect on initial shear stiffness and damping ratio measured. Figure 1 shows the paths followed during the tests in the ( ua): (ua uw) plane. As shown in Figure 1, the stress paths approach to same points in ( ua):(ua uw) plane (e.g. points ZK2350150 & ZK3350150 and ZK135050, ZK235050 & ZK335050, and paths ZK135050 to ZK145050 & ZK335050 to ZK345050) via different history of mean net stress and suction. This testing strategy was chosen in order to investigate the effect of stress path on volumetric and stiffness parameters. Further details about soil properties, specimen preparation, the experimental program, and procedures can be found in Biglari et al. (in press). The initial equalisation have been performed by applying positive values of the pore water and pore air pressures at the specimen base, and waiting for the water pressure stabilisation while measuring the volume and water volume changes with time. Equalisation was assumed at stabilisation of both the above variables. The constant suction isotropic stress paths have been performed in water drained condition applying a mean net stress loading rate of 2 kPa/hr. During consolidation specific volume, (v 1 e), specific water volume, (vw 1 Sre), G0, and D0 changes have been measured almost continuously. The wetting stage have been finally carried out by applying finite variations to the pore water pressure at the specimen boundary and measuring v, vw evolution in both ZK2 and ZK3 and the G0 and D0 value in ZK3 at equalisation point. 3 3.1 EXPERIMENTAL RESULTS AND DISCUSSION Equalization stage
Time (hr)
Figure 2. (a) Variation of specific water volume during equalization stage (b) Variation of specific volume during equalization stage with time in hour.
suction (si) measured by the HCT (si 240 kPa), during the equalization stage at 300 kPa of suction water flowed out of the sample, while in the equalization stages at 150 kPa and 50 kPa of suction water flowed in (Fig. 2a). The changes of specific volume against time plotted in Figure 2b for the various tests seem to exclude the occurrence of severe creep phenomena for the tested soil during the equalisation time. Regardless of the suction value, reduction in specific volume was observed during all equalization stages. This reduction in specific volume is due to collapse behaviour in the ZK1 and ZK2 tests, and to compression by some coupled SI-LC movements in the tests ZK3. A detailed description of the stress paths followed during the equalization stage, the estimate position of the LC yield curve produced by the initial compaction process and its evolution during equalization stage can be found in Biglari et al. (in press). 3.2 Volumetric behavior during isotropic compression and wetting stages The changes in specific volume, v, and specific water volume, vw, versus mean net stress occurred in the three tests during isotropic compression and
The change of specific water volume, vw, with time, occurred during the equalization stages of all the tests discussed in this paper are plotted in Figure 2a. As expected from the after compaction
201
wetting paths are plotted in Figures 3a, and 3b respectively. A detailed inspection of the ramped consolidation plots in Figure 3 indicates that when the yield stress at a particular suction was exceeded, the soil state fell on an isotropic normal compression line. Figure 3 also shows some yielding for ZK1 and ZK2 specimens even though samples were theoretically normally consolidated after the equalization stage, in the authors opinion this behaviour can be due to minimal creep strains accumulated during the long-lasting equalization stage. The experimental data reported in Figure 3 also show that same stress points (in terms of mean net stress and suction) are associated to same specific volume and specific water volume, independently of the stress path followed to reach the final stress state. This is for example the case of points ZK2350150 and ZK3350150 (both at ( ua) 350 and (ua uw) 150) plotted in Figures 3a and 3b. This observation is not fully confirmed by the points ZK135050, ZK235050, and ZK335050, that do not overlap perfectly, even if related to the same stress conditions, but these may be reasonably considered as worthless differences (maximum difference are between ZK135050 and ZK335050 points
which are vw 0.02 and v 0.04) caused by a non perfect homogeneity between different samples and thus a normal experimental dispersion. According to the BBM (Alonso et al, 1990), it is possible to assume independency between volumetric behaviour and stress path in these series of data since the final stress points lie on the same yield locus (Fig. 3). It is worth noting that the two final stress points at s 50 and 150 kPa with the same ( ua) 350 kPa, are all reached by wetting: the same conclusion (i.e. the uniqueness of the hydro-volumetric state at the same stress point and yield locus) is not intended here applicable to wetting paths preceded by drying stage. 3.3 Initial shear stiffness measurements during isotropic compression and wetting stages
(a) 2.25
Specific volume v
2.15 ZK2 2.05 1.95 1.85 1.75 1.65 ZK1 ZK2 ZK3 0 100 200 ZK3350-300 ZK2350-150 ZK3350-150 ZK2350-50 ZK3350-50 ZK1350-50 300 400 B
50-150
ZK350-300
ZK150-50
ZK3450-50 ZK1450-50 500
Figure 4 presents the G0 measurements versus mean net stress collected during the tests ZK1, ZK3 and some parts of ZK2 test. In these tests G0 and D0 were measured at scheduled time by resonant column tests at very small shear-strain level (i.e. multistep sequence). Since drained conditions were constantly achieved, the G0 and D0 measurements were attributed to the current mean net stress and suction. The rate of loading in isotropic loading stage was selected to obtain an acceptable compromise between the dissipation of excess pore-water pressure throughout the sample and the reduction of creep strains. Resonant column tests were carried out along all the paths continuously, except for ZK2 test during which unfortunately because of unidentified reason G0 and D0 measurements failed in isotropic compression stage at constant suction after ( ua) 150 kPa (ZK2150150). The G0 data measured after the equalization at given suction values are presented in Figure 4. During the ZK3 test after the equalization stage at 50 kPa
(b) 1.45
Specific water volume vw
1.43 1.41 1.39 1.37 1.35 1.33 1.31 1.29 1.27 0 100 200 300 400 ZK250-150 B ZK350-300 ZK2350-150 ZK150-50 ZK3350-50 ZK2350-50 ZK1350-50 ZK1450-50 ZK3350-150 ZK3450-50
150 130 110 ZK1 ZK2 ZK3 ZK3450-50 ZK3350-150 ZK1450-50 ZK3350-300 ZK3350-50 ZK1350-50
G0 (Mpa)
90 70 50 ZK350-300 B 30 ZK250-150 ZK150-50 10 0 100 ZK2150-150
ZK3350-300
ZK1 ZK2 ZK3

500
200
300
400
500
Mean Net Stress (kPa)
Figure 3. (a) Variation of specific volume (b) Variation of specific water volume, versus mean net stress, during isotropic compression stage at constant suction and wetting stage at constant mean net stress.
Figure 4. Initial shear stiffness measured during isotropic compression stage at constant suction and wetting stage at constant mean net stress.
202
of suction, the mean net stress was increased from ( ua) 350 to 450 kPa to compare the shear modulus measured along this isotropic compression path with that measured on the same stress path in ZK1 test. In normally consolidated conditions, G0 shows a non linear increase with mean net stress. The values pertaining to tests performed at different constant suction levels, describe different parallel curves showing higher G0 values at higher suction values. In the test ZK3, the wetting path from s 300 kPa up to 50 kPa leaded to collapse (Fig. 3), being the sample in a normal consolidated state (ZK3350300). Hence, the ZK3 test shows an almost constant G0 value on wetting from 300 kPa to 50 kPa notwithstanding the significant reduction of suction. One further interesting observation is the similarity between initial shear stiffness of ZK1 test in the range from the point ZK135050 to ZK145050 and ZK3 test from ZK335050 to ZK345050, suggesting the independency of initial shear stiffness on previous stress path (Fig. 4). A slight difference which is observed between starting points of this path (ZK135050 and ZK335050) can explained by hardening in ZK3 sample due to creep strains accumulated during the long-lasting equalization stages at 50 kPa of suction. According to these data it is possible to assume that in soil samples with similar loading collapse, LC, lines, initial shear stiffness is independent on stress path. 3.4 Damping ratio measurements during isotropic compression and wetting stages As mentioned in previous sections, resonant column tests were continuously performed during isotropic consolidation and wetting stages. Small strain material damping ratio of specimen was measured by half-power bandwidth method from resonant column results in frequency domain. Small strain material damping is the damping inherent in the materials which can be measured in linear shear-strain domain (less than l almost 105) without any slippage of particles or plastic deformation. The magnitude of this damping ratio is small comparing to the damping ratio in higher shear strain ranges. Figure 5 shows the results of initial damping ratio variation versus mean net stress for ZK1, ZK2 and ZK3. Along the loading paths D0 first slightly increases and then start to decrease when the normally consolidated state is achieved. In the normally consolidated range D0 shows a non linear decrease as the mean net stress increases and values corresponding to tests performed at different constant suction levels, describe different parallel curves showing lower D0 values at higher suction values.
3 2.5 2
ZK1 ZK2 ZK3 ZK2150-150
D0 (%)
1.5 ZK250-150 ZK150-50 B 1 ZK350-300 0.5 0
ZK3450-50 ZK3350-50 ZK3350-150 ZK1 350-50 ZK3350-300 ZK1450-50
100
200
300
400
500
Mean Net Stress (kPa)
Figure 5. Initial damping ratio measured during isotropic compression stage at constant suction and wetting stage at constant mean net stress.
One of the possible explanations for the observed slight increase in D0 around the yielding point may be obtained by looking at the evolution in the specific volume and the specific water volume. As a matter of fact, during the isotropic compression two different phenomena take place: the specific volume reduction and the degree of saturation increase. Reduction in specific volume causes a decrease in D0 and an increase in degree of saturation lead to a D0 increase. Hence, interaction of these two contrasting phenomena may have resulted in temporarily D0 increase around yielding stress corresponding to a faster rate of reduction in volume and a non complete constant-suction consolidation caused by a slight non-equalization of pore water pressure throughout the sample. As the compression path proceeds and the net stress increases, global hardening of mean net stress mainly influences D0 trend. Furthermore, the initial damping ratio variation with mean net stress is not significant. For Zenoz kaolin it varies between 1% and 2%. 4 CONCLUSIONS
Data from series of controlled suction resonant column tests on samples of compacted Zenoz kaolin were used to investigate the effects of the generalized stress path (mean net stress and suction) on the volumetric behavior (v and vw), initial shear stiffness (G0) and damping ratio (D0) in unsaturated conditions. To this purpose, series of Resonant Column tests were carried out using a torsional shear device (RCTS_ns) working in suction controlled conditions available at the University of Naples Federico II, Italy. This machine allowed to acquire almost continuous measurements of G0 and D0 along complex stress paths including: (a) initial equalization stages up to
203
suction values of 50, 150 and 300 kPa; (b) a constant suction isotropic compression stage up to a 350 kPa of mean net stress; (c) a wetting path up to the 50 kPa of suction; (d) a constant suction (50 kPa) isotropic loading up to 450 kPa of mean net stress. The mentioned stress path allowed to eliminate the over consolidation effect on initial shear stiffness and damping measured during the stage (b), (c) and (d). As a matter of fact, before wetting, the mean net stress and suction histories applied are expected to have moved the volumetric state of the sample on the LC yield locus of the BBM. Starting from this stress state wetting (i.e., suction reducing) at constant mean net stress is expected to have caused collapse with the current LC continuously moving and following the current generalized stress state. These new tests results suggest that initial shear stiffness, initial damping ratio and volumetric behaviors of normally consolidated samples of the Zenoz kaolin are stress path independent, since G0, D0, v, and vw measured at the same mean net stress and suction value reached by different stress path show the same values. REFERENCES
Alonso, E.E., Gens, A., & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40(3): 405 430. Biglari, M., dOnofrio, A., Mancuso, C., Shafiee, A., & Jafari, M.K., Small strain stiffness of Zenoz Kaolin in unsaturated conditions, Submitted to Canadian Geotechnical Journal, paper No. 09211. Biglari, M., dOnofrio, A., Mancuso, C., Shafiee, A., & Jafari, M.K. 2009. Small strain behavior of unsaturated plastic material in suction control RCTS test. In Proceedings of 4th Asian Pacific Conference on Unsaturated soils, Newcastel, Austuralia, 2325 November 2009. CRC Press/Balkema, In press. Cabarkapa, Z., Cuccovillo, T., & Gunn, M. 1999. Some aspects of the pre-failure behaviour of unsaturated soil. In Proceedings of the Second International Conference on Pre-failure Behaviour of Geomaterials, Torino, Italy, 2729 September 1999. A.A. Balkema, Rotterdam. Vol. 1, pp. 159165. DOnza, F. 2007. Influenza dello stato di parziale saturazione dei terreni sulla risposta sismica locale. Ph.D. thesis, University of Napoli Federico II, Naples, Italy. DOnza, F., dOnofrio, A., & Mancuso, C. 2008. Effects of unsaturated soil state on the local seismic response of soil deposits. First results, Unsaturated Soils: In Proc. 1st European Conf. on Unsaturated Soils, Durham, UK (ed. Toll, D.G., Augarde, C.E., Gallipoli, D. & Wheeler, S.J.), 24 July 2008. Leiden: CRC Press/ Balkema, pp. 531536.
Ladd, R.S. 1978. Preparing testing specimens using undercompaction. Geotechnical Testing Journal, 1: 1623. Mancuso, C., Vassallo, R., & dOnofrio, A. 2002. Small strain behavior of a silty sand in controlled-suction resonant column torsional shear tests. Canadian Geotechnical Journal, 39: 2231. Ng, C.W.W., & Yung, S.Y. 2008. Determination of the anisotropic shear stiffness of an unsaturated decomposed soil. Gotechnique, 58(1): 2335. Pineda, J.A., Lima, A., & Romero, E.E. 2008. Influence of hydraulic paths on the low-strain shear modulus of a stiff clay. First results, Unsaturated Soils: In Proceedings of 1st European Conf. on Unsaturated Soils, Durham, UK (ed. Toll, D.G., Augarde, C.E., Gallipoli, D. & Wheeler, S.J.), 24 July 2008. Leiden: CRC Press/Balkema, pp. 519524. Rojas, J.C., Pagano, L., Zingariello, M.C., Mancuso, C., Giordano, G., & Passeggio, G. 2008. A new high capacity tensiometer: First results, Unsaturated Soils: In Proc. 1st European Conf. on Unsaturated Soils, Durham, UK (ed. Toll, D.G., Augarde, C.E., Gallipoli, D. & Wheeler, S.J.), 24 July 2008. Leiden: CRC Press/Balkema, pp. 205212. Vassallo, R. 2003. Comportamento di terreni costipati non saturi a piccole, medie e grandi deformazioni. Ph.D. thesis, University of Napoli Federico II, Naples, Italy. Vassallo, R., Mancuso, C., & Vinale, F. 2007a. Effects of net stress and suction history on the small strain stiffness of a compacted clayey silt. Canadian Geotechnical Journal, 44: 447462. Vassallo, R., Mancuso, C., & Vinale, F. 2007b. Modelling the influence of stressstrain history on the initial shear stiffness of an unsaturated compacted silt. Canadian Geotechnical Journal, 44: 463472. Vinale, F., dOnofrio, A., Mancuso, C., Santucci De Magistris, F., & Tatsuoka, F. 1999. The prefailure behaviour of soils as construction materials. In Proceedings of the Second International Conference on Pre-failure Behaviour of Geomaterials, Torino, Italy, 2729 September 1999. A.A. Balkema, Rotterdam. Vol. 2, pp. 9551007. Xu, J., Ng, C.W.W., & Yung, S.Y. 2008. Drying and wetting effects on shear wave velocity of an unsaturated soil. First results, Unsaturated Soils: In Proceedings of 1st European Conf. on Unsaturated Soils, Durham, UK (ed. Toll, D.G., Augarde, C.E., Gallipoli, D. & Wheeler, S.J.), 24 July 2008. Leiden: CRC Press/ Balkema, pp. 525530.
204
Shrinkage during wetting of fine-pored materials: Does this accord with the principle of effective stress?
G.E. Blight
University of the Witwatersrand, Johannesburg, South Africa
ABSTRACT: According to the principle of effective stress, unsaturated porous materials will shrink if they become dryer and swell if they become wetter. An interesting situation arises when such a material approaches a completely dry state. If the negative pore water pressure that has caused the material to shrink, disappears because there is insufficient water to sustain it, does the material expand? If the material is then re-wetted sufficiently for the negative pore water pressure to be re-established, is the material then re-compressed? The paper describes a set of experiments designed to establish what happens and reports the result. 1 INTRODUCTION According to the well-known Kelvin equation (Blight, 1966) p can be represented by p 311 log10H [in MPa] (2)
According to the capillary model of an unsaturated porous material, the pore space is partly filled with air and partly with water. Because of the curvature of the air-water interfaces, or menisci, the pressure in the water is less than that in the air, and if the air-filled pore space is connected to the atmosphere, the pressure in the water will be sub-atmospheric and could be an absolute tension. Figure 1 (Marshall, 1959) shows the capillary model in diagrammatic form, with the pressure in the pore water represented by the (negative) height h of a suspended column of water, i.e. the difference between pore air and pore water pressures is p (ua uw) wh pore water suction (1)
where H is the relative humidity of the pore air. Thus p increases as H decreases, as shown in Table 1. In the table, the radius of the air-water menisci has been calculated from p 2T/r (3)
where T is the surface tension of water and r is the meniscus radius. According to the principle of effective stress, in a homogeneous porous material that is subjected to a change of isotropic effective stress I, the corresponding isotropic linear strain will be I/E(1 2) (4)
where E is the elastic modulus and v is Poissons ratio. If the material is mechanically unconfined, i.e. 0, and subjected to a change of suction p, the change in effective stress will be I p where is Bishops effectiveness parameter, and hence p/E(1 2) (4a)
Figure 1. material.
Capillary model of partly saturated porous
As shown by Table 1, p can become very large, reaching 620 MPa when H 1%. Although will decrease as H decreases, I might still be equal to several tens of MPa when the porous material is almost dry. Assuming that the capillary model remains
205
Table 1. Values of H and corresponding values of p calculated from equation (2). H% 100 98 95 85 75 60 45 30 15 5 1 p MPa 0 2.7 6.9 21.9 38.8 68.9 108 163 256 405 622 Meniscus radius r mm 103 (micron) Menisci planar 5.3 2.1 0.66 0.37 0.21 0.13 0.09 0.06 0.04 0.02
valid at low values of H, a stage must be reached at which p disappears, either because there is no longer enough water to form menisci, or because there are no longer voids small enough for the required radius of meniscus to form. At this stage, does the strain also disappear and the material expand? On the other hand, if H is subsequently increased and sufficient water condenses in the fine pores to re-establish menisci, does the strain reappear as p re-establishes itself ? To try and answer these questions, an experiment was undertaken. 2 THE EXPERIMENT
It was decided to start the measurements from a standard state of dryness and to allow the concrete specimens to take up water, and then dry them out in stages and observe their strain behaviour as they approached dryness. Drying at 50C is often used to establish water contents of organic materials, e.g. grain, timber and municipal solid waste, as relatively little of the volatile constituents (except water) is lost at 50C. The concrete slabs were therefore dried out at 50C for 24 hours, and after cooling to 20C and taking zero strain measurements on both sides of each slab, groups of slabs were exposed in vacuum desiccators to relative humidities of 86, 93 and 98%, all controlled by saturated salt solutions in trays at the base of the desiccators. The desiccators were kept in a temperature-controlled laboratory at 20 1C. The strain versus time graphs appear in Figures 2 to 4. All of the specimens initially shrank before starting to expand after about 3 to 5 days. After 60 days none of the specimens had returned to their initial length. At 60 days the specimens stored at a relative humidity of 86% were interchanged with those stored at 93% and those stored at 98% had their saturated salt solution replaced by distilled water. The latter specimens then swelled slightly past their initial length and also sharply increased in water content. Thus all specimens complied with the expectation, expressed in the introduction, that shrinkage would occur as water menisci were reestablished by absorption and condensation of water drawn from the surrounding air. Referring to equation (4a) p E(1 2) The concrete would have behaved exactly like a soil, in that the rebound strain would have been less than the initial compressive strain. Taking the elastic modulus of the rebounding concrete as E 40 000 MPa and 0.1, p 32 000 MPa. The maximum initial compression strain varied from about 350 to 550 and hence the maximum initial compressive stress was from 11 to 17 MPa. It was also to be expected that at 60 days the initial compression strain had not been recovered. It was only after the concrete had been allowed to swell freely in an atmosphere close to RH100% that the initial length was regained. As the menisci reformed by condensation, the radii would have been very small, possibly corresponding to a relative humidity as low as 15%. At this stage, the parameter would also have been as small as 11/256 to 17/256, i.e. 0.04 to 0.07. As the relative humidity within the pores of the concrete increased towards the ambient relative humidity in the desiccator, the suction would have reduced and would have increased progressively.
Small changes of strain are difficult to measure on unconfined soil specimens without losing soil solids which, in turn makes measurements of small changes of water content difficult. For this reason, it was decided to use concrete as a surrogate soil and make the measurements on small concrete slabs on which strains could be measured to an accuracy of 10 microstrain using targets and a Demec gauge. The specimens were sawn from cores taken from bridge structures that were 40 years old when sampled, the object being to use a very old concrete in which no re-activation of hydration of the cement was likely to occur. The test specimens were sawn from the cores and consisted of slabs 10 mm thick, 150 mm long and 50 mm wide. The Demec targets were mounted on both faces of the slab with a gauge length of 100 mm, so that any bending effects caused by uneven shrinkage or expansion could be detected. The small 10 mm thickness was used to reduce the drainage path length to only 5 mm and hence shorten equilibration times. Changes of water content were measured by weighing the specimens.
206
Figure 2. Behaviour of specimens A during exposure to relative humidities of 86% to 98% at 20C after drying at 50C.
Figure 3.
Similar to Figure 2, but for specimens B.
207
Figure 4.
Similar to Figures 2 and 3, but for specimens C.
Figure 5. Behaviour of specimens after re-drying at 50C followed by exposure to 98% relative humidity, at 20C. In each case, zero strain and water content are the same as in Figures 2, 3 and 4. (The numbers (1), (2) and (3) allow water content and linear strain curves to be matched.).
208
To further investigate and confirm the phenomena recorded in Figures 2 to 4, after 83 days the specimens were again oven-dried at 50C and the original measurements were repeated, except that all specimens were now exposed to RH 98%. The results of the first seven days of this stage of testing are shown in Figure 5. In this figure, the zeros of measurement for strain and water content are the same as in Figures 2 to 4. The measurements at 90 days were recorded after removing the specimens from the 50C oven, placing them into a sealed desiccator over silica gel to cool and, when at 20C, measuring initial masses and strains. After this, the silica gel in the desiccators was replaced with a saturated solution of potassium sulphate to provide the RH 98% atmosphere. Comparing Figure 5 with Figures 2 to 4, it will be seen that the residual shrinkage at 90 days was approximately equal to the maximum shrinkage originally recorded at about three days, while the residual water content at 90 days was also approximately the same as the water content at three days. Thereafter, further shrinkage took place up to 92 to 95 days, after which the specimens again began to swell. It therefore seems that if an expansion occurs as the last vestiges of unbound water are removed from the finest pores, it is likely to be much smaller than the compression that occurs on re-wetting.
CONCLUSION
The concept of the capillary model illustrated in Figure 1 leads to the expectation that when a partly saturated porous material is dried, the water menisci in the pores will withdraw into the finer pores and the material will shrink until the last vestiges of water have been withdrawn from the finest pores. When this happens, a slight expansion will occur. When the dry material is incrementally wetted by condensation from humid air drawn into the air filled pore space, water menisci will reform, starting in the finest pores, exerting a compressive stress on the material and causing it to shrink. As the water menisci grow larger and their radii reduce, the material will start to swell. The experiment described in this paper supports this conclusion. REFERENCES
Blight, G.E. 1966. Strength characteristics of desiccated clays. Journal of Soil Mechanics and Foundation Engineering, ASCE, 92(SM6): 1937. Marshall, T.J. 1959. Relations between water and soil. Technical Communication No. 50, Commonwealth Bureau of Soils, U.K.
209
Shear strength of an unsaturated silty sand

F. Casini, P. Minder & S.M. Springman
Institute for Geotechnical Engineering, IGT-ETH, Zrich, Switzerland
ABSTRACT: This paper presents a series of direct shear tests performed on a silty sand at three different gravimetric water contents. The soil was sampled from test pits south of Ruedlingen in North East Switzerland, where a landslide triggering experiment was carried out on a 3740 steep forested slope through infiltration of extreme artificial rainfall. The aim of this work was mainly to establish a correlation between the degree of saturation and slope stability. Direct shear tests were performed on reconstituted samples with a direct shear apparatus under undrained conditions for the water phase. Three of the samples were saturated after the shear phase in order to analyse the effect of wetting on shear strength. The results were interpreted with the assistance of a soil water retention curve (WRC) and an analytical slope stability analysis was performed to apply the laboratory results to the field experiment. 1 INTRODUCTION the slope. Some surface movements were detected during this extreme event, although failure did not occur. Subsequently, a range of measures was implemented, such as relocating the distribution of the sprinklers to provide more rainfall to the upper part of the slope, so that a failure mechanism was triggered in March 2009, incorporating about 150 m3 of debris (Springman et al. 2009; Casini et al. 2010; Askarinejad et al. 2010). It is essential to establish a suitable model for the shear resistance of the soil in order to conduct stability analysis. Shear tests performed by Minder (2008) on reconstituted samples are presented in this paper. They are interpreted based on a WRC by considering the influence of vertical load and water content. A simple slope stability analysis is also performed that takes the hysteresis of the WRC into account in respect to the suction acting on the shear plane. 2 2.1 MATERIAL TESTED Soil characterisation
Landslides are a common natural hazard in steep mountainous terrain, such as that of Switzerland (Springman et al. 2003; Lateltin et al. 2005; Springman et al. 2009). Experience in Hong Kong shows that they occur when short-duration rainfall is of high intensity (Brand et al. 1984; Finlay et al. 1997; Fannin et al. 2005). The location of the failure is strongly influenced by spatial and temporal variations in seepage induced pore water pressure (Fannin et al. 2000). A clear understanding of the likely failure mechanisms is required for a reliable stability analysis, and characterization of the soil strength is essential over an appropriate stress range (Fannin et al. 2005). A slip can be caused by: a transient increase in pore water pressure generated by subsurface seepage (Johnson & Sitar 1990); static liquefaction induced by small perturbation that leads to flow failure (Anderson & Sitar, 1995); reduction in negative pore pressures that requires the suction history of the soil to be taken into account (Springman et al. 2003). A landslide triggering experiment was carried out near Ruedlingen in North East Switzerland along the river Rhine in autumn 2008 and spring 2009 to replicate the effects of a heavy rainfall event from May 2002, in which 100 mm rain fell in 40 minutes (Fischer et al. 2003), causing 42 superficial landslides. The slope was subjected to extreme rainfall by artificial means in October 2008 over a period of 4 days, with a rainfall intensity of 15 mm/h for the first two days and an intensity of 30 mm/h for the last two days. The sprinklers were distributed at constant spacing along the central line of
A series of sieve analyses of the soil from Ruedlingen has been carried out by Casini et al. (2008), paying special attention to its spatial variability (Springman et al. 2009; Casini et al. 2010). The soil can be classified as medium to low plasticity silty sand (ML) according to USCS. The activity, IA, is higher than 1.25 in the upper part of the soil profile and decreases from IA 1.25 to IA 0.75 with increasing depth. Activity of the soil comes from the chloritic-smectitic clay fraction (Colombo 2009). The mean physical properties are summarised in Table 1.
211
Table 1. w (%) 19.6
Mean physical properties of the in-situ soil. IP (%) 9.9 s (kN/m3) 25.8
silt
(kN/m3) 16.7
sand
e () 0.85
gravel
Sr (%) 60.6
100 percentage passing P [%] 80 60 40 20 0
in situ laboratory
Figure 2. Sample preparation: (a) cloddy soil after mixing with water; (b) sieved soil; (c) shear box filled with soil; (d) compaction with hydraulic press.
0.01 0.1 1 10 diameter D [mm]
100
Figure 1. Comparison between in situ and laboratory grain size distribution.
The soil was reconstituted based on a representative grain size distribution (Figure 1), by mixing the fractions of the natural soil and omitting the fractions larger than 2 mm. 2.2 Sample preparation
Figure 3. Sample homogeneity: (a) sample w 20%; (b) sample w 25%.
The hand-mixed soil was passed through a 4 mm sieve before it was stored for 24 hours in a sealed bag in order to avoid large clods for being placed in the direct shear apparatus. After this homogenisation phase, the wet soil was statically compacted in one 2 cm thick layer to a void ratio e 0.9 by means of a hydraulic press. Figure 2 illustrates the steps of sample preparation. The samples were statically compacted at three different target water contents of w 15%, 20% and 25%. The compaction stress required to reach the same void ratio of e 0.9 varied with water content. It ranged from 10 kPa (w 25%) to 100 kPa (w 15%) but was similar for samples with the same water content. Depending on the remoulding water content, different degrees of sample homogeneity were obtained. Figure 3 shows a comparison between a homogeneous sample (w 20%) with a wetter, macroporous sample (w 25%). The shear tests were performed with a standard direct shear apparatus at three vertical stresses v 20 40 80 kPa, respectively. Three of the samples were saturated after 6 mm horizontal
displacement in order to analyse the effect of the wetting on the mobilised shear strength. The samples were sheared with a velocity v 0.033 mm/min. This shear rate is fast enough to represent undrained conditions for the water phase. 3 3.1 EXPERIMENTAL RESULTS Shear tests
The results of the shear phase are shown in Figures 4, 5 and 6 in terms of the relation between normalised shear stress (/v), vertical displacement v and shear displacement h. A negative vertical displacement indicates an increase in specimen height. Highest stress ratios are exhibited by two samples at v 20 kPa and w 15 and 20% (Figure 4a). Both specimens show an increase in height (dilatancy) during shearing (Figure 4b). The soil reduces in volume during the saturation of specimens at w 20 and 25%, indicating wetting collapse (Figure 4b). In the subsequent shearing stage, the stress ratios are in good agreement with the saturated sample.
212
1.5 v= 20 kPa 1
1.5 v= 80 kPa 1
v
v
0.5
(a)
0.5
(a)
0 -0.5 0 0.5 1 1.5 2
2 4 6 8 Shear displacement h (mm)
10
Vertical displacement v (mm)
0 -0.5 0 0.5 1 1.5 2
10
w=15% w=20% w=25% w=w sat saturation saturation
(b)
w=15% w=20% w=25% w=wsat saturation
(b)
10
10
Figure 4. Tests with v 20 kPa: (a) stress ratio and (b) vertical displacement versus shear displacement.
1.5 v= 40 kPa 1
v
vertical stress leads to a lower void ratio resulting in a higher degree of saturation and therefore less suction. Secondly the overconsolidation ratio (OCR) is decreasing as the vertical stress increases. 3.2 Soil water retention curve
0.5
(a)
0 -0.5 0 0.5 1 1.5 2
10
w=15% w=20% w=25% w=wsat
(b)
10
The sample under v 80 kPa also reduces in volume during the saturation phase (Figure 6). The stress ratio for the equal water content decreases as the vertical stress increases. This is due to the combination of two effects. Firstly, increasing
The Water Retention Curve (WRC) in terms of water ratio ew Vw/Vs, where Vw is the volume of water and Vs is the volume of solids, and suction s ua uw; was obtained using a Fredlund Oedometer (Figure 7) under suction controlled conditions for a reconstituted sample with an initial void ratio e0 0.83 and w0 17%. The volumetric state variable ew Sre tends towards the void ratio with increasing the degree of saturation Sr. The water ratio ew is obtained directly in laboratory measuring the inflow or outflow of water into a soil with a known volume of solids. In addition, ew remains constant in water undrained tests in an equivalent concept to isochoric tests, where e remains constant (Romero & Vaunat 2000). The results are reported in Figure 8. A suction s 70 kPa was applied as first step. Water was pressed out of the sample initiallly, which means that the compacted soil had a suction less than 70 kPa. The following three points seem follow a straight line in the plane ews (which could be interpreted as a scanning curve). The other points in wetting and in drying follow the respective main curves.
213
eters of the subcurves (i, ni, mi) are subject to the conditions (i 0, mi 0, ni 1. The parameter mi is assumed to be mi 1 1/ni. A bimodal retention function was chosen, because the soil was compacted on the dry part of the Proctor curve, which usually leads to a bimodal distribution of voids. This assumption is also supported by Mercury Intrusion Porosimetry (MIP) tests performed by Colombo (2009). The comparisons between the model for drying and wetting curves and the laboratory results are reported in Figure 9 and the resulting parameters are given in Table 2. 4.2 Shear strength envelope
Figure 7. Oedometer-type pressure plate apparatus (after Perez Garcia et al. 2008).
80 begin 60 end
main wetting point main drying point scanning curve point
The shear strength of an unsaturated soil can be espressed using many different approaches (e.g. Miao et al. 2001; Vaunat et al. 2002). A prediction based on the WRC was chosen here (berg & Sllfors 1997; Springman et al. 2003): n ua)tan Sr (ua uw)tan
40
20
where n is the total normal stress, ua is the air pressure, uw the pore water pressure, Sr is the degree of saturation and is the friction angle of the material. The quantity Sr(ua uw) was obtained using the WRC. For each test known ew and e, it was obtained the suction and the degree of saturation Sr ew/e. The comparison between the prediction and the laboratory data are shown in Figure 10. The meas100
scanning curve point main wetting main drying
s (kPa)
0 0.2
0.4
0.6 0.8 ew=Sr*e=Vw/Vs
1
80 60
Figure 8.
WRC of the reconstituted Ruedlingen soil.

s (kPa)
4 4.1
DATA INTERPRETATION Water retention curve
40 20 0
The data reported in Figure 8 was modelled with a multimodal Durner model (1994) that describes the retention characteristics of soils with heterogeneous pore systems. The multimodal retention function is constructed by a linear superposition of subcurves of the Van Genuchten type:
k ew ewres 1 wi ewsat ewres i 1 1 ( s )ni mi
0.2
0.4 0.6 ew=Sr*e=Vw/Vs
0.8
Figure 9. curve. Table 2.
Multimodal function for the water retention Parameter durner model. w1 a1 (kPa1) n1 0.27 0.86 3.3 3.3 a2 (kPa1) n2 0.014 0.022 7 2.9 ewsat 0.85 0.85 ewres 0.15 0.15
(1)
where k is the number of subsystems that form the total pore-size distribution, and wi are weighting factors for the subcurves, subject to 0 wi 1 and wi 1. As for the unimodal curve, the param-
drying wetting
0.53 0.65
214
100
' =31
[kPa]
50
50 100 'n =n+Sr (ua-uw) [kPa]
150
Figure 10.
Shear strength envelope using the WRC.
ured points are reasonably well fitted with a linear envelope using Equation 2 with a 31. 4.3 Slope stability analysis
Figure 11. Depth z for FoS 1 for the drying and wetting branch is plottet as a function of suction.
In order to apply the laboratory results to the test side in Ruedlingen a simplified slope stability analysis is carried out with shear strength determined under unsaturated conditions for homogeneous ground. Extending the approach of Springman et al. (2003) to a laterally limited slide, the factor of safety FoS becomes:
c*
FoS
cos
1 2K
z d cos( )
z tan 1 K
z d cos ( )
3

(3)
z tan
where c* Sr(ua uw) tan , is the inclination of the slope, z the depth of the sliding surface, d the width of the slope, K the earth pressure coefficient. The groundwater table is assumed to lie below the sliding surface and the lateral sides can fully mobilise the shear resistance according to Equation 2. In the following parametric evaluation of Equation 3 a slope inclination of 38, a width of d 8 m, a void ratio e 0.9 and an earth pressure of K 1 sin()( 31) were chosen as representative values for the field conditions. The depth, for which the factor of safety reaches unity, is plotted in Figure 11 versus the. All points to the left of the curves represent stress states where the shear stress exceeds the shear resistance. Shallow slides are stable due to the apparent cohesion provided by suction, whereas very deep slides are unlikely to happen due to the friction mobilised on the planes of the failure mechanism. The difference between drying and wetting conditions originate from different degrees of saturation and unit-weight at same suction. When plotting the critical depth z (FoS 1) versus the degree of saturation (Figure 12), the difference
Figure 12. Depth z for FoS 1 for the drying and wetting branch is plotted as a function the degree of saturation.
between the drying and the wetting branch of the WRC becomes more evident. During a rainfall event, the soil is saturated gradually. When a suction s 2.9 kPa and Sr 0.55 are exceeded, a sliding surface is formed at a depth of about 1.12 m. 5 CONCLUSION
A triggering landslide experiment was conducted in north-east Switzerland. In order to characterise the strength of the soil, a series of direct shear tests were conducted at three constant water content and vertical stresses. Three samples were also saturated for a better understanding of the effect of saturation.
215
The results were interpreted with the help of a water retention curve obtain under suction controlled conditions. They are well fitted with a Mohr Coulomb criteria with the use of Bishop stress with parameter Sr and a critical state friction angle of 31. During the saturation phase, all three samples showed a decrease in height indicating a wetting collapse that decreases as the water content increases. The normalised shear stress (/v) decreases, for the same gravimetric water content; as the vertical stress increases due to a lower void ratio and hence a higher degree of saturation and lower suction. A slope stability analysis was performed using the same approach as Springman et al. (2003) but extended to a laterally limited slide. A critical depth z 1.12 m is obtained for a suction of 2.9 kPa. All the other depths investigated needed less suction for a safety of factor FoS 1.The approach used is quite simple with all the limitations of limit equilibrium analyses. But it can be applied to predict the zone of potential failure, which in the field experiment was located at a depth z 1.25 m. The suction mobilised at failure was lower than the critical value determined here. This is due to a more complex response of the soil in the field in comparison to the constraints of the laboratory sample and its reconstitution. ACKNOWLEDGEMENTS This research was funded by the Competence Centre for Environment and Sustainability (CCES) within the framework of the TRAMM - Project. The authors are grateful to Amin Askarinejad, Lorenzo Colombo, Peter Kienzler, Marco Sperl, Ernst Bleiker for their contribution to the field and laboratory studies. We are also most grateful to the Gemeinde Ruedlingen and their President Katy Leutenegger for giving permission to carry out this experiment on their land. REFERENCES
Anderson, S.A., Sitar, N. (1995). Shear strength and slope stability in a shallow clavey soil regolith. In Haneberg W.C., Anderson S.A. (eds): Reviews in Engin. Geology, Vol. X: Clay and Shale Slope Stability. Askarinejad, A., Casini, F., Kienzler, P., Springman, S.M. 2010. Comparison of the in situ and laboratory water retention curves for a silty sand (this issue). Brand, E.W., Premchitt, J., Phillipson, H.B. 1984. Relationship between rainfall and landslides in Hong Kong. Proc. 4th Int. Symp. on Landslides 1: 377384. Casini, F., Jommi, C., Springman, S.M. 2010. A laboratory investigation on an undisturbed silty sand from a slope prone to landsliding. Granular Matter 12(3): 303316.
Casini, F., Kienzler, P., Askarinejad, A., Springman, S.M., 2008. Laboratory results. IGT CCES-TRAM: 113. Colombo L. (2009), Large shear box for analysing strength mobilisation in unsaturated conditions. Master Thesis, Politecnico di Milano. Durner W. (1994), Hydraulic conductivity estimation for soils with heterogeneous structure. Water Resources Research. 30(2): 211223. Fannin, R.J., Eliadorani, A., Wilkinson, J.M.T. 2005. Shear strength of cohensionless soils at low stress. Gotechnique 55(6): 310. Fannin, R.J., Jaakkola, J., Wilkinson, J.M.T., Hetherington E.D. 2000. The hydrological response of soils to rainfall at the Carnation Creek watershed, Brititsh Columbia. Water Resour. Res. 36(6): 14911494. Fischer, C., Lpez, J., Springman, S.M. 2003. Remediation of an eroded steep slope in weathered sandstone after a major rainstorm. Intern. Conf. on Land., Hong Kong:. 878883. Finlay, P.J., Fell, R., Maguire, P.K. 1997. The relationship between the probability of landslide occurrence and rainfall. Can. Geotech. J. 34(6): 811824. Johnson, K.A., Sitar N. 1990. Hydrologic conditions leading to debris flow initiation. Can. Geotech. J. 27(6): 789901. Lateltin O., Haemmig Ch., Raetzo H., Bonnard Ch. 2005 Landslide risk management in Switzerland. Journal of the International Consortium on Landslides, 2(4). Miao, L., Yin, Z., Liu, S., 2001. Empirical function representing the shear strength of unsaturated soils. Geotech. Testing Journal 24: 220223. Minder, P. (2008). Shear resistance of silty sand from the Rdlingen monitoring and triggering test site. Semester thesis, ETH, Zurich. berg, A., Sllfors, G. 1997. Determination of shear strength parameters of unsaturated silts and sands based on the water retention curve. Geotech. Test. J. 20(1): 4148. Perez-Garcia, N., Houston, S.L., Houston, W.N., Padilla, J.M. 2008. An oedometertype Pressure Plate SWCC Apparatus. Geotechnical Testing Journal 31: 19. Romero, E., Vaunat, J. 2000. Retention curves of deformable clays. Intern. Work. On Unsat. Soils: Exper. Ev. & Theoretical Approaches in Unsat. Soils, Trento, Italy: 91106. Tarantino A, Mancuso C (eds), Balkema Rotterdam. Springman, S.M., Jommi, C., Teysseire P. 2003. Instabilities on moraine slopes induced by loss of suction: a case history. Gotechnique 53(1): 310. Springman, S.M., Kienzler, P., Casini, F., Askarinejad, A. 2009. Lanslide triggering experiment in a steep forested slope in Switzerland. In Hamza M., Shaien, M., El-Mossallamy Y. (eds):: Proc. of the 17th Intern. Conf. on Soil Mech. and Geotech. Engin., Egypt, 59 Oct.. IOS Press. Vaunat, J., Romero, E., Marchi, C., Jommi, C. 2002. Modeling the shear strength of unsaturated soils. 3rd Intern. Conf. on Unsaturated Soils, 2: 245251. Balkema.
216
Dilatancy and shear strength of an aged compacted fill

Abraham C.F. Chiu & X. Zhao
Key Laboratory of Ministry of Education for Geomechanics and Embankment Engineering, Hohai University, Nanjing, China
ABSTRACT: The dilatancy and shear strength of an aged compacted fill was studied by conducting a series of suction-controlled triaxial tests on its intact and reconstituted samples. The frameworks of intrinsic properties of reconstituted soil and average skeleton stress are used to interpret the effects of bonding and partial saturation on the behaviour of the fill. Similar to other bonded soils, a non-linear peak envelope is observed for the intact state and the shear strength increases with suction. At low stress level, the pre-yield intact state exhibits brittle and dilative behaviour and the corresponding peak strength envelope lies above the intrinsic critical state line (ICSL) of saturated reconstituted state. This extra shear strength may be contributed by the post-construction bonding and dilatancy of the fill. Pre-yield and post-yield intact states exhibit different stress-dilatancy relationships. It seems that the bond breakage causes the stress-dilatancy relationship of intact state shifting towards that of reconstitute state. The effect of suction on the mechanical behaviour of the fill is similar to that of bonding. The suction provides additional attractive forces at the inter-particle contacts, which results in the augmentation of yield stress and peak shear strength, and reduction of dilatancy at failure.
INTRODUCTION
Compacted fill materials are unsaturated in nature and post-construction bonding and cementation between particles due to the physico-chemical processes can be developed during the lifetime of earth structures. Thus, it is crucial to study the coupled effects of partial saturation and bonding on the mechanical behaviour of aged compacted soils in order to understand properly the long-term performance of earth structures. Past studies have been focused on the mechanical behaviour of either unsaturated compacted soils without bonding (Wheeler & Sivakumar 1995, Cui & Delage 1996, Ng & Chiu 2001 & 2003, Melinda et al. 2004, Jotisankasa et al. 2009) or saturated bonded soils (Burland 1990, Leroueil & Vaughan 1990, Cecconi & Viggiani 2001, Lo & Wardani 2002, Jovicic et al. 2006, Cotecchia et al. 2007). Burland (1990) suggested that the intrinsic properties of reconstituted soil can be used as a reference framework for interpreting the behaviour of corresponding bonded soil. Further, Lo & Wardani (2002) showed that the high peak strength of a saturated cemented silt is related to the negative dilatancy (or dilative response) at failure. Recently, a few laboratory studies have presented the stress-strain behaviour and shear strength at the critical states for unsaturated bonded soils (Rinaldi & Capdevila 2006, Toll et al.
2008). Further experimental evidence is required to examine the interplay between the bonding and suction on the peak strength of unsaturated bonded soils. The framework of two independent stress state variables is commonly used for interpreting the mechanical behaviour of unsaturated soils (Fredlund & Morgenstern 1977). The most common choices of stress state variables are net stress ( ua) and matric suction (ua uw), where is total normal stress, ua is pore-air pressure and uw is pore-water pressure. Recently alternate forms of stress state variables have been proposed (Jommi 2000; Wheeler et al. 2003; Gallipoli et al. 2003; Tarantino 2007), for example ps = ( 1 2 3 ) / 3 ua Sr (ua u w ) s n (ua u w ) (1) (2)
where ps is average skeleton stress, 1, 2 and 3 are principal stresses, s is modified suction, n is porosity and Sr is degree of saturation. The average skeleton stress represents the average stress acting on the solid phase of unsaturated soil. The advantages of using the average skeleton stress are (i) a smooth transition between the saturated state and the unsaturated state, i.e., at zero suction, average skeleton stress becomes effective stress, and (ii) account for the hydraulic hysteresis effect on the mechanical behavior of unsaturated soil.
217
In this study, a series of suction-controlled triaxial tests was conducted on intact and reconstituted samples of an aged compacted fill. The term fill in the remainder of this paper will mean aged compacted fill. The combined effects of partial saturation and bonding on the dilatancy and shear strength of the fill is interpreted based on the frameworks of the intrinsic properties and average skeleton stress. 2 LABORATORY STUDY
Soil block samples were taken from a compacted fill slope at a depth of 1.5 m. The average dry density and in-situ water content of the intact samples are 1560 kg/m3 and 19%, respectively. The fractions of sand, silt and clay are 5%, 70% and 25%, respectively. The liquid and plastic limits are 43% and 25%, respectively. The specific gravity is 2.72. According to the Unified Soil Classification System (USCS), the fill is classified as clay of low plasticity (CL). The maximum dry density and optimum water content determined from the standard Proctor test are 1670 kg/m3 and 19%, respectively. The air-entry value of the samples compacted to the insitu dry density determined by the pressure plate test is below a suction of 50 kPa (Ji 2007). Reconstituted samples were prepared by mixing the intact samples at a water content of about 1.5 times the liquid limit (Burland 1990). The slurry was then placed in a consolidometer and loaded under one-dimensional condition to a state where reconstituted specimens could be cut without disturbance. Intact specimens were trimmed from the block samples by wire saw. Triaxial tests were carried out on saturated specimens of 39 mm in diameter and 80 mm in height and unsaturated specimens of 50 mm in diameter and 100 mm in height. Suction-controlled compression and shear tests were conducted in a twin-cell triaxial stress path apparatus. The testing apparatus is controlled by three water-pressure controllers and one air-pressure controller. Axis translation technique (Hilf 1956) is used to control the suction. Pore-water and pore-air pressures are applied at the base of the specimen through a 5 bar high air-entry ceramic disk and at the top of the specimen through a low air-entry porous filter, respectively. An inner cell made of glass is used inside a conventional Bishop and Wesley triaxial cell. The volume change of soil specimen is measured by monitoring the amount of water flowing into or out of the inner cell with an Imperial College type volume change measuring device. The double cell system was calibrated for the compliance errors including the movement of loading ram, the deformation of inner cell and drainage lines due to the variation of cell pressure and temperature, and the compressibility of water.
The test program and conditions are summarised in Table 1. A total of 14 drained tests were conducted and the first alphabet I and R of specimen no. represent intact and reconstituted specimens, respectively. For saturated tests, the specimens were first saturated with de-aired water inside a vacuum chamber for at least 24 hours. After installing the specimens in the triaxial cell, pore-pressure parameter B was checked. Back pressure saturation would be applied if a minimum B-value of 0.98 was not reached. Thereafter the specimens were isotropically compressed to confining pressures ranging from 25 kPa to 400 kPa. Filter paper side drains were used to accelerate compression. After compression the specimens were sheared under drained condition at a strain rate of 1.2%/hour. The test procedures for unsaturated tests consist of three stages: (i) suction equalisation, (ii) isotropic compression under constant suction, and (iii) drained shear under constant suction. To speed up the suction equalisation, the specimens were first airdried in the laboratory under ambient temperature. Before assembling in the triaxial cell, the dimensions and mass of air-dried specimens were measured by a caliper and balance, respectively. Furthering drying (or suction equalisation) of the specimens was continued under a suction of 100 kPa and a cell pressure of 25 kPa inside the triaxial cell for 2 to 7 days. The equalisation was terminated when the rate of change in volumetric water content of specimens was smaller than 0.005/day. Thereafter the cell pressure was applied at a loading rate of 5 kPa/ hour. After reaching the target cell pressures given in Table 1, the specimens were left for 24 hours to ensure complete dissipation of excess pore-water pressure. Then the specimens were sheared at a strain rate of 0.24%/hour under constant suction.
Table 1. Specimen no. R-0-50 R-0-100 R-0-200 R-0-400 R-100-50 R-100-200 I-0-25 I-0-50 I-0-100 I-0-200 I-0-400 I-100-50 I-100-100 I-100-200 Test program and conditions. Suction (kPa) 0 0 0 0 100 100 0 0 0 0 0 100 100 100 Confining pressure (kPa) 50 100 200 400 50 200 25 50 100 200 400 50 100 200 Void ratio after compression 0.793 0.703 0.647 0.581 0.821 0.742 0.742 0.736 0.698 0.671 0.630 0.675 0.703 0.687
218
YIELDING
Figure 1 shows the isotropic compression test results under constant suction of unsaturated samples as void ratio (e) plotted against mean net stress (p). p is defined as (1 2 3)/3 ua. The isotropic compression curves of saturated samples are also shown in the figure for comparison. The compression curve of saturated reconstituted samples represents the intrinsic normal compression line (INCL) of the fill that is used as a reference frame to interpret the effect of bonding on its mechanical behaviour (Burland 1990). The compression curve of saturated intact samples lies above the INCL. Leroueil & Vaughan (1990) defined the space above the INCL as a structure permitted space that may be used as a measure to reflect the extent to which the soil is bonded. Compared to typical natural soils, the structure permitted space of the fill is relatively small and a relatively weak bonding may be anticipated. Well-defined yield stresses are observed in the compression curves (see arrows in Figure 1). The yield stresses for the unsaturated reconstituted samples (suction 100 kPa) and saturated intact samples (suction 0 kPa) are 65 kPa and 70 kPa, respectively. No distinct yielding is observed in the compression curve of unsaturated intact samples (suction 100 kPa). It is anticipated that the yield stress should be greater than 200 kPa, which is the maximum applied stress used in the compression test. Thus, the yield stress increases with increasing suction for both intact and reconstituted samples. For intact samples, the yielding may be attributed to the beginning of breakdown of inter-particle bonding. Consider the soil microstructure, suction may provide additional stabilising forces at the inter-particle contacts, as a result a higher stress is required to break the bonding (in intact samples) or initiate the slippage at the inter-particle contacts (in reconstituted samples) and a higher yield stress is expected for intact or unsaturated samples.
1 0.9 Voidratio,e 0.8 0.7 0.6 0.5 1
Intact (s = 0 kPa) Intact (s = 100 kPa) Reconstituted (s = 0 kPa) Reconstituted (s = 100 kPa) INCL (s = 0 kPa)
STRESS-STRAIN BEHAVIOUR AND DILATANCY
Figures 2 and 3 show the typical stress-strain curves of intact and reconstituted samples. It should be noted that specimens R-100-50, I-0-50, I-100-50 and I-100-200 are compressed below the yield stress (or pre-yield state) before shearing. As expected the shear strength and initial stiffness of the intact samples are higher than those of reconstituted samples. Further the shear strength and initial stiffness of unsaturated samples are higher than those of saturated samples. Only two specimens (I-0-50 & I-100-50) exhibit brittle behaviour, but the brittleness is mild as the percentage for the drop of strength is below 15% of peak strength. A tendency to dilate is observed for these two pre-yield specimens at large strain. Regarding the axial-volumetric strain relationship, intact and unsaturated samples are less contractive than reconstituted and saturated samples, respectively. Figure 4 shows the effect of bonding on the stress-dilatancy relationship for the saturated samples. Dilatancy is defined as the ratio of plastic strain increments (dvp/dsp), where dvp and dsp are increments in plastic volumetric and deviatoric strains, respectively. The Rowes stress-dilatancy relationship (Rowe 1963) is also shown in the figure for comparison. It should be noted that specimens I-0100 and I-0200 are in the post-yield states before shearing, i.e. certain degrees of bonding were damaged after isotropic compression. The figure
600 Deviatoricstress,q(kPa)
(a)
400
R-0-50 R-0-200 R-100-50 R-100-200 I-0-50 I-0-200
200
0 0 5 10 Axial strain (%) 15 20
Volumetricstrain(%)
R-0-50 R-0-200 R-100-50 R-100-200 I-0-50 I-0-200
(b)
10
100
1000
10000
10 Axial strain (%)
15
20
Mean net stress, p (kPa)
Figure 1. Isotropic compression curves of intact and reconstituted samples.
Figure 2. samples.
Typical stress-strain curves of reconstituted
219
shows that the dilatancy is stress dependent which decreases with increasing effective stress ratio (q/p), i.e. less contractive at higher q/p. For reconstituted samples, the dilatancy approaches zero at failure, i.e. no volumetric change. However the intact samples (except specimen I-0-400) reach a negative dilatancy at failure. Further two distinct stress-dilatancy relationships may be identified for the intact samples. Despite some scatter of the data for specimen R-0-50 at q/p smaller than 1, it seems that the test data of post-yield intact samples (I-0-100 & I-0-200) fall closely within a narrow band of the test data of reconstituted samples. On the other hand, the test data of pre-yield intact samples (I-0-25 & I-0-50) fall within another band which is above that of the reconstituted samples.
1000 Deviatoric stress,q(kPa) 800 600 400 200 0 0
I-0-50 I-0-200 I-100-50 I-100-200 R-0-50 R-0-200
(a)
10 Axial strain (%)
15
20
8 Volumetric strain(%)
I-0-50 I-0-200 I-100-50 I-100-200 R-0-50 R-0-200
(b)
As the bond breakage continues to evolve, the stress-dilatancy relationship of a partly damaged state (or post-yield intact state) may move down from that of the virgin state (or pre-yield intact state) towards the completely damaged state (or reconstituted state). The test results are consistent with those of other naturally bonded soils, e.g. a pyroclastic soil (Cecconi & Viggiani 2001) and a stiff North Sea clay (Jovicic et al. 2006). Figure 5 shows the effect of suction on the stress- dilatancy relationship for the reconstituted samples. The dilatancy is plotted in terms of mean net stress (p) and average skeleton stress (ps) in Figures 5a and 5b, respectively. Both saturated and unsaturated specimens approach a zero dilatancy at failure. In addition, a higher mean net stress ratio (q/p) is required for unsaturated samples to reach zero dilatancy than that for saturated samples. If the stress states of the unsaturated samples are expressed in terms of ps, the test data will fall within a narrow band for both saturated and unsaturated samples despite some scatter of the data for specimen R-0-50 at q/ps smaller than 1. Figure 6 shows the effect of suction on the stressdilatancy relationship for the intact samples. The dilatancy is plotted in terms of ps. Unlike reconstituted samples, the dilatancy at failure decreases (more negative) with increasing suction for intact samples. For q/ps greater than 1.3, the pre-yield intact samples (I-0-25, I-0-50, I-100-50, I-100-100 & I-100-200) and post-yield intact samples (I-0-100 &
2.0 1.5
(a)
2
Dilatancy 1.0 0.5 0.0 -0.5 0.5
R-0-50 R-0-100 R-0-200 R-100-50 R-100-200 Rowe's stress dilatancy
0 0 5 10 Axial strain (%) 15 20
Figure 3. samples.
2.0 1.5 Dilatancy 1.0 0.5 0.0 -0.5
Typical stress-strain relationships of intact
1.0 1.5 Stress ratio (q/p)
2.0
2.0
R-0-50 R-0-100 R-0-200 I-0-25 I-0-50 I-0-100 I-0-200 Rowe's stress dilatancy
(b)
1.5 Dilatancy 1.0 0.5 0.0 -0.5 0.0
R-0-50 R-0-100 R-0-200 R-100-50 R-100-200 Rowe's stress dilatancy
0.5
0.0
0.5
1.0 q/p'
1.5
2.0
1.0 q/ps
1.5
2.0
Figure 4. samples.
Stress-dilatancy relationships for saturated
Figure 5. Stress-dilatancy relationships for reconstituted samples in terms of (a) mean net stress and (b) average skeleton stress.
220
2.0 1.5 Dilatancy 1.0 0.5 0.0 -0.5 -1.0 0.0 0.5 1.0 q/ps
Deviator stress,q(kPa)
I-0-25 I-0-100 I-100-50 I-100-200
I-0-50 I-0-200 I-100-100
1000 800 600 400 200 0 0

Peak strength envelopes of intact samples
(a)
ICSL of saturated reconstituted samples
Failure envelopes of reconstituted samples
Reconstituted (s = 0 kPa) Reconstituted (s = 100 kPa) Intact (s = 0 kPa) Intact (s = 100 kPa)
200
400
600
800
1000
1.5
2.0
1000
Deviator stress,q(kPa)
Mean net stress, p (kPa)

(b)
Peak strength envelope of pre-yield intact samples
I-0-200
Figure 6. Stress-dilatancy relationships for intact samples.
800
I-0-400
600 400
I-0-200) fall within two different bands. Compared to the post-yield intact samples, a higher stress is required for the pre-yield intact samples to reach the same dilatancy. For q/ps smaller than 1.3, the unsaturated samples (I-100-50, I-100-100 & I-100200) have smaller dilatancy than the saturated samples (I-0-25 & I-0-50) for a given q/ps. Based on the observations in Figures 4 to 6, both bond breakage and suction affect the dilatancy of the fill. With continuous evolution of bond breakage, the stress-dilatancy relationship of post-yield intact state may converge towards that of reconstituted state. There may be two possible contributions of suction on the mechanical behaviour of the fill. Firstly, the additional stabilising forces at the inter-particle contacts induced by suction increase the yield stress of the fill such that a higher stress is required to initiate the bond breakage. Secondly, these stabilising forces also enhance the interlocking between particles, as a result dilative response (or negative dilatancy) is observed at low stress level where bond breakage is not dominated. 5 FAILURE ENVELOPE
I-0-100
Reconstituted (s = 0 kPa) Reconstituted (s = 100 kPa) Intact (s = 0 kPa) Intact (s = 100 kPa)
200 0 0
CSL of saturated reconstituted samples
200
400
600
800
1000
Average skeleton stress, ps (kPa)
Figure 7. Failure envelopes of intact and reconstituted samples in terms of (a) mean net stress and (b) average skeleton stress.
Figure 7 shows the failure envelopes of intact and reconstituted samples. The failure envelopes of intact and unsaturated samples are higher than those of reconstituted and saturated samples, respectively. A non-linear envelope is observed for the saturated intact samples for the range of applied stress used in the study (see Figure 7a). The nonlinear envelope is attributed to the continuous breakdown of the bonding of the intact samples. At low stress (pre-yield), the peak strength envelope of the intact samples lies above the failure line of the saturated reconstituted samples or the intrinsic critical state line (ICSL). The breakdown of bonding is not commenced at
low stress level which may account for the extra shear strength of intact samples. At high stress, a large extent of bonding is damaged after isotropic compression as the post-yield compressibility increases substantially (see Figure 1). As a result the peak strength envelope of intact samples converges towards the ICSL. Figure 7b shows that a linear failure envelope will be used to represent the failure states of either reconstituted or pre-yield intact samples if the stress states are expressed in terms of ps. It seems that the failure states of post-yield intact samples (I-0100, I-0-200 & I-0-400) are well below the failure envelope of pre-yield intact samples. A possible explanation is that the pre-yield intact samples exhibit a negative dilatancy at failure (see Figure 6). Their higher peak strengths are the results of extra strength contributed by the dilatancy (Lo & Wardani 2002). Further study is required to examine this hypothesis. 6 CONCLUSIONS
Suction controlled triaxial tests were conducted on intact and reconstituted samples of an aged compacted fill. The effects of partial saturation and bonding on the dilatancy and shear strength of the fill are investigated and the test results are interpreted based on the frameworks of intrinsic properties of
221
the reconstituted soil and average skeleton stress. The findings of the study are summarised as follows: i. Similar to other bonded soils, a non-linear strength envelope is observed for the intact state. At low stress, the pre-yield intact state exhibits brittle behaviour and a tendency to dilate at large strain. The corresponding peak strength envelope lies above the intrinsic critical state line (ICSL) of saturated reconstituted state. The extra shear strength of pre-yield intact state may be contributed by the post-construction bonding and dilatancy of the fill. A large extent of bonding is damaged after isotropically compressed to high stress. As a result the post-yield intact state exhibits ductile and contractive behaviour and the corresponding peak strength envelope converges towards the ICSL. ii. Pre-yield and post-yield intact states exhibit different stress-dilatancy relationships. It seems that the evolution of bond breakage causes the stress-dilatancy relationship of intact state shifting towards that of reconstitute state. iii. The effect of suction on the mechanical behaviour of the fill is similar to that of bonding, i.e. the yield stress and shear strength increases with increasing suction. The suction provides additional attractive forces at the inter-particle contacts to stabilise the microstructure of the fill, which result in the augmentation of yield stress and shear strength. Further these stabilising forces also enhance the interlocking between particles, as a result a negative dilatancy (or dilative response) is observed at low stress level where bond breakage is not dominated. ACKNOWLEDGEMENT This study was sponsored by the National Natural Science Foundation of China through Grant No. 50878076. REFERENCES
Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Gotechnique 40(3): 329378. Cecconi, M. & Viggiani, G.M.B. 2001. Structural features and mechanical behaviour of a pyroclastic weak rock. International Journal for Numerical and Analytical Methods in Geomechanics 25(15): 15251557. Cotecchia, F., Cafaro, F. & Aresta, B. 2007. Structure and mechanical response of sub-Apennine Blue Clays in relation to their geological and recent loading history. Gotechnique 57(2): 167180. Cui, Y.J. & Delage, P. 1996. Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique 46(2): 291311. Fredlund, D.G., & Morgenstern, N.R. 1977. Stress state variables for unsaturated soils. Journal of Geotechnical Engineering Division ASCE 103(5): 447466.
Gallipoli, D., Gens, A., Sharma, R. & Vaunat, J. 2003. An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Gotechnique 53(1): 123135. Hilf, J.W. 1956. An investigation of pore water pressure in compacted cohesive soils. US Bureau of Reclamation Technical Memorandum 654. Ji, L.T. 2007. Shear strength and nonlinear constitutive model of unsaturated soil. Master Thesis, Hohai University. (In Chinese). Jommi, C. 2000. Remarks on the constitutive modelling of unsaturated soils. In: Tarantino A. and Mancuso C. (eds), Proc. Experimental evidence and theoretical approaches in unsaturated soils: 139153. Jotisankasa, A., Coop, M. & Ridley, A. 2009. The mechanical behaviour of an unsaturated compacted silty clay. Gotechnique 59(5): 415428. Jovicic, V., Coop, M. & Simpson, B. 2006. Interpretation and modelling of deformation characteristics of a stiff North Sea clay. Canadian Geotechnical Journal 43: 341354. Leroueil, S. & Vaughan, P.R. 1990. The general and congruent effects of structure in natural soils and weak rocks. Gotechnique 40(3): 467488. Lo, S.R. & Wardani, S.P.R. 2002. Strength and dilatancy of a silt stabilized by a cement and fly ash mixture. Canadian Geotechnical Journal 39: 7789. Melinda, F., Rahardjo, H., Han, K.K. & Leong, E.C. 2004. Shear strength of compacted soil under infiltration condition. Journal of Geotechnical and Geoenvironmental Engineering 130(8): 807817. Ng, C.W.W. & Chiu, A.C.F. 2001. Behaviour of a loosely compacted unsaturated volcanic soil. Journal of Geotechnical and Geoenvironmental Engineering 127(12): 10271036. Ng, C.W.W. & Chiu, A.C.F. 2003. Laboratory study of loose saturated and unsaturated decomposed granitic soil. Journal of Geotechnical and Geoenvironmental Engineering 129(6): 550559. Rinaldi, V.A. & Capdevila, J.A. 2006. Effect of cement and saturation on the stress-strain behavior of a silty clay. In: Miller, G.A., Zapata, C.E., Houston, S.L. & Fredlund, D.G. (eds.), Proc. 4th International Conference on Unsaturated Soils, Unsaturated Soils 2006 (GSP 147) April 26, 2006, Carefree, Arizona: 1157116. Rowe, P.W. 1963. Stress dilatancy, earth pressures, and slopes. Journal of the Soil Mechanics and Foundations Division, ASCE 89(SM3): 3761. Tarantino, A. 2007. A possible critical state framework for unsaturated compacted soils. Gotechnique 57(4): 385389. Toll, D.G., Ali Rahman, Z. & Gallipoli, D. 2008. Critical state conditions for an unsaturated artificially bonded soil. In: Toll, D.G., Augarde, C.E., Gallipoli, D. & Wheeler, S.J. (eds), Proc. 1st European Conference on Unsaturated Soils, Durham, UK: 435440. Leiden: CRC Press/Balkema. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Gotechnique 53(1): 4154. Wheeler, S.J. & Sivakumar, V. 1995. Elasto-plastic critical state framework for unsaturated soil. Gotechnique 45(1): 3553.
222
An experimental study on the cyclic behavior of natural loess

Y.-J. Cui, A.-M. Tang & J.P. Karam J.M. Terpereau
Ecole des Ponts ParisTech, Marne-la-Valle, France Socit Nationale des Chemins de Fer, Paris, France
ABSTRACT: Instability problems have been observed along the Northern High Speed Line in France. It was suspected that these problems are related to the behavior of loess situated along the line, as collapse, fatigue and liquefaction under the cyclic load induced by the high speed train. In order to investigate the soil behavior in the laboratory corresponding to the above mechanisms, a cyclic triaxial cell was adapted in order to study the behavior of unsaturated soil. This development consists of using a differential pressure transducer for the measurement of the soil volume change in a triaxial double-wall cell. Cyclic triaxial tests were performed on saturated and unsaturated undisturbed soil specimens. The results from tests on initially saturated specimens showed that the soil taken at 2.2 m depth has the lowest resistance to cyclic loading, in relation to its highest porosity and lowest clay fraction. This soil was then studied at unsaturated state with various initial water contents. Unsaturated soil specimens were first subjected to cyclic loadings to decrease their volume. These cyclic loadings were stopped when the soil was considered to become saturated. Afterwards, the back-pressure tubing was saturated with de-aired water and cycles were applied under undrained condition. Significant effect of initial water content was evidenced: the lower the initial water content, the higher the cyclic resistance. This can be explained by the densification of the soil during the initial cyclic loadings. 1 INTRODUCTION saturated conditions. When the problem of liquefaction is of concern, it implies that in situ saturation condition is satisfied. A possible explanation on how a natural soil, initially unsaturated can liquefy, could be explained by the volume change due to cyclic loading under near constant water content conditions, which can lead to soil saturation and therefore the liquefaction. The present work aims at studying the cyclic behavior of natural loess from northern France, at a site close to the railways of northern TGV line, under the train-induced cyclic loading. Emphasis is put on the effect of initial saturation state on the liquefaction resistance by performing special tests in two stages: i) cyclic loading under air-drained and nearly constant water content condition to reach soil saturation, and ii) cyclic loading under undrained condition. 2 MATERIAL
The Northern French TGV (high speed train) railway is subjected to an equivalent 3500 cycles per day and around 1 270 000 cycles per year. Between 1999 and 2004, numerous sinkholes on ground surface after long rainy periods were observed along this line. These holes are about 5 m in diameter and 1 m deep. For the holes situated far from the railway line, the origin was identified and attributed to the cavities that had existed since the First World War and other existing quarries. This represents 60% of the identified cases; the remaining 40% involved the zone of less than 25 m from the railway, for which the collapse could not be explained by cavity presence. In 2003, numerical simulations by finite element modeling were performed by Geodynamics Company in order to assess the cyclic loads in the soil. It was reported that until 5 m away from the railway, the vertical cyclic load varies from 11 to 23 kPa till 4 m depth at a frequency of 20 Hz. The load considerably decreases beyond 5 m. From a mechanical point of view, the collapse of loess could be related to three deformation mechanisms: (i) collapse upon wetting; (ii) fatigue under cyclic loading in drained conditions; (iii) liquefaction when subjected to cyclic loading under
The material studied is loess taken from northern France located at 140 km North of Paris. Sampling was carried out according to the French Standard (AFNOR 1994) in a trench of 1.5 9 m that is 25 m far from the railways. Four different
223
Table 1. Depth (m) 1.2 2.2 3.5 4.9
Geotechnical properties of studied soil. % wL 2m (%) 20 16 16 18 30 28 26 30 Ip 9 9 6 9 d (Mg/m3) 1.52 1.39 1.54 1.55 CaO (%) 5 6 15 9 wnat (%) 18.9 18.1 16.6 23.7 Sr (%) 66 53 55 82
depths (1.2, 2.2, 3.5 and 4.9 m) were considered. The geotechnical properties of the samples are presented in Table 1. The liquid limit (wL) varies from 26 to 30%; the plasticity index (Ip) varies from 6 to 9%; the dry density is rather low (d 1.39 to 1.55 Mg/m3) that correspond to high porosities (n 0.43 0.49) or high void ratio (e 0.76 0.93); the carbonate content is quite high (% CaCO3 5 15%). The grain size distribution analysis showed that the clay fraction (2 m) ranges from 16 to 20%. X-ray diffractometry showed that the clay fraction involves kaolinite, illite and interstratified illite-smectite. The suction measured using filter paper method is not high, ranging from 14 to 34 kPa, even though the natural degree of saturation (Sr) is quite low, comprised between 53 and 82% (initial water content wnat 16.6 23.7%). More details about the soil studied can be found in Karam et al. (2009). 3 EXPERIMENTAL DEVICE
Figure 1.
Experimental set-up of cyclic triaxial test.
The cyclic triaxial cell presented by Cui et al. (2007) was used in the present work. A differential pressure transducer (DPT) was installed in a cyclic triaxial cell for monitoring the total volume change of unsaturated soil specimen. The schematic layout of the system is presented in Figure 1. The vertical pressure applied on soil specimen is controlled by the difference between pressures applied in the two chambers of the vertical loading system. The loading controller, that also controls the confining pressure, monitors these pressures. A pressure transducer is installed below the base pedestal to monitor the vertical stress applied on the specimen. The vertical displacement is monitored by a LVDT (Linear Variable Differential Transformer). All the data including applied pressures, vertical stress and vertical displacement are recorded by the logging system through the control system. Details about the loading control and the vertical displacement measurement of this cyclic triaxial cell were described by De Gennaro et al. (2004). This cyclic cell was modified and adapted to volume change monitoring when testing unsaturated
soils. The principle of double walled triaxial cell was applied by adding an inner cell in plexiglas. As water inside and outside the inner cell is under the same confining pressure, any variation of the water level inside the inner cell, indicated by the DPT changes corresponds to a volume change of the soil specimen. A preliminary calibration is required for data treatment. In cyclic triaxial tests, the measurement of soil volume change requires a very high accuracy. For this reason, a low working pressure, high resolution DPT was used. Moreover, the response time of DPT must be compatible with the frequency at which the cyclic loading is applied during a test. The full-scale pressure output of the DPT is 1250 Pa that corresponds to a water height of 125 mm. The accuracy is 1.25 Pa (0.125 mm water height). The response time under a pressure of 1125 Pa is less than 100 milliseconds. Therefore, theoretically, the DPT is able to work under a cyclic pressure of a frequency lower than 10 Hz. The performance of the developed set-up can be seen through the results presented in Figure 2, obtained using this triaxial cell on the loess at 2.20 m depth at various water contents. In this figure, the deviator stress (q), axial strain (1), and volumetric strain (v) are presented as a function of the cycle number (N). It can be observed that the volume measurement device adopted provides satisfactory response since regular v curves were obtained for all samples. Moreover, good agreement was observed between 1 and v curves. In q N plan, it can be seen that the deviator stress varied from 0 to 20 kPa during the first 700 cycles.
224
250 200 q (kPa) 150 100 50 0 40
EXPERIMENTAL RESULTS
Two types of cyclic triaxial tests were carried out, on initially saturated and unsaturated samples, respectively. Only non alternative compression (one-way compression) loading was applied.
0 500 1000 1500 2000 2500
30
w = 14% w = 21% w = 23% w = 27% w = 10%
4.1 On initially saturated samples Cylindrical soil specimens of 70 mm in diameter and 140 mm high were prepared by cutting and trimming from soil blocks taken at four different depths (1.2 m, 2.2 m, 3.5 m and 4.9 m). After flushing with CO2 gas for about 45 min, the specimens were saturated with de-aired water. 200 kPa back pressure was applied to obtain a satisfactory saturation with a Skemptons coefficient B higher than 0.99. The saturated specimen was then isotropically consolidated under an effective confining pressure of 25 kPa. The volume change due to the consolidation, which was estimated to be small, was not monitored. Force-controlled cyclic triaxial tests were then carried out under undrained condition by applying cyclic loads using a hydraulic piston. The tests were conducted at constant cyclic deviator stress of varying magnitudes at a frequency of 0.05 Hz. The effective confining pressure (3 25 kPa) was maintained constant during cyclic loading. The axial strain (1), the pore water pressure change (u), the deviator stress (q 1 3) and the effective mean pressure (p 1/3 2 3/3) were recorded using a data logging system. In total 12 tests were conducted, 3 on each depth. For the soil of a given depth, an undrained triaxial test upon monotonic loading was first carried out, giving its peak deviator stress qmax. The 3 cyclic triaxial tests were performed under different deviator stress qcyc. Figure 3 presents the results from one test conducted on an initially saturated soil specimen. The figure is organized in four plots: (i) axial strain 1 versus number of cycle N; (ii) deviator stress q versus axial strain 1; (iii) pore water pressure change u versus axial strain 1; and (iv) deviator stress q versus effective net mean stress p. It can be observed in the 1 N plot that 1 increases slightly for the first 3 cycles and then abruptly afterwards. The double amplitude strain (peak-to-peak strain) is about 3% at the fourth cycle and more than 7% at the fifth cycle. The final strain reached 19% in less than 7 cycles. Interestingly, u increases drastically for the first 3 cycles, to reach a value higher than 17 kPa at the end of the third cycle. The final value of u is about 23 kPa, close to the confining pressure of 25 kPa. Examination of q 1 plot shows that q started to decrease significantly when 1 started to develop at a higher rate from the third
1 (%)
20
10
0
0
500
1000
1500
2000
2500
10
v (%)
w = 27% w = 23% w = 21% w = 10%
20
w = 14%
30
500
1000 1500 Number of cycle, N
2000
2500
Figure 2. Cyclic triaxial tests on unsaturated loess at various water contents, f 0.05 Hz, 3 25 kPa (Cui et al. 2007).
And then, its peak value increased 10 kPa every 100 cycles until failure for all the tests except that at 10% water content where after 2000 cycles (q 0 150 kPa), the peak value of deviator stress was increased 10 kPa every 10 cycles until failure at q 240 kPa. The effect of the water content was clearly observed in both 1 N and v N plan. Before failure, at a same cycle number, the axial strain and volumetric strain were larger at higher water content. Moreover, except the case of 21% water content where the failure was not reached, the number of cycle, N, required to reach the failure decreased with the increase of water content. In examining of the effect of the amplitude of q on changes in 1 and v, it can be observed that the strains started generally increasing significantly and then seemed to stabilize under a given cyclic deviator stress. In addition, the higher the water content the larger the first part of quick change. An exception was made by 10% water content at which no effect of deviator stress amplitude was observed: both 1 and v varied linearly as a function of N.
225
Figure 3.
Results on a test at initially saturated state. Figure 4. Cyclic resistance versus number of cycle at the instability point for saturated samples.
cycle; this is obviously the consequence of the specimens section correction. In the p- q plot, in addition to q decrease, a progressive decrease of p was observed, as the result of u increase. If the point corresponding to the abrupt change in 1 rate is defined as the stability loss state of the soil, it can be concluded that for this test, the loss of stability was achieved at about 2% axial strain after 3 cycles (Nloss). Among the 12 tests performed, loss of stability was observed in 10 tests: for the 3 tests on the soil at 3.50 m, stability loss was reached only under qcyc 25 kPa. The results of cyclic resistance from all the tests are plotted versus Nloss in Figure 4. Three regression lines are plotted for the three depths: 1.2, 2.2 and 4.9 m. It is observed that for all the depths, the lower the value of cyc/c, the larger the number of cycles Nloss. Comparison between the four depths showed that the soil at 1.20 m is the most resistant and the soil at 2.20 m is the less stable. 4.2 On initially unsaturated samples
For the tests at initially saturated state, only the soil at 2.20 m was studied. Cylindrical specimens of 140 mm high and 70 mm in diameter were prepared by cutting and trimming from the soil blocs. After introducing it in a cylindrical latex membrane (70 mm in diameter), the soil was wetted by adding water drops using a pipette. The latex membrane was used to avoid the collapse of the soil specimen during wetting. When the desired water content was reached, wetting was stopped and the soil was covered by a plastic film during 24 h for water content homogenization. 3 initial water contents
wi were considered: 30, 32.2 and 33%. The corresponding degrees of saturation Sri are 85, 92 and 94%, respectively. After installation of the soil specimen in the cell, a confining pressure of 25 kPa was applied under drained conditions. As in the previous case on initially saturated specimens, the volume change due to this pressure was estimated to be small and not monitored. Cyclic loading was then applied by increasing progressively the cyclic deviator; it was stopped when the volume change of the soil specimen measured by the differential pressure device (see Figure 1) is equal to the initial air-pore volume in the soil (estimated from the initial degree of saturation). Assuming that the soil water content did not change during the cyclic loading, the suppression of air volume means that the soil became saturated. After this phase, all the draining system (tubing and cell basis) was filled with de-aired water. Undrained cyclic test was then performed without application of back pressure. Figure 5 presents the results from one test (wi 33%) during the first cyclic loading stage. As the q N plot indicates, the samples was loaded under 5 kPa for 30 cycles, 12 kPa for 20 cycles, 22 kPa for 125 cycles, 26 kPa for 27 cycles and 41 kPa for 8 cycles. The 1 N plot shows that this cyclic loading induced an axial strain 1 up to 2.3% and the volume decrease reached about 11.4 cm3 which corresponds to a volumetric strain of 4.5% and a void ratio change from ei 0.93 to ef 0.86. After this first stage of cyclic loading, the cell basis and the drainage tubing were saturated
226
Figure 5. Results from a test at initial unsaturated statesuppression of air-volume.
Figure 6. Results from a test at initial unsaturated stateloading under undrained conditions.
and the second cyclic loading under undrained conditions was started. The results are shown in Figure 6. Cyclic deviator loading was applied with qcyc 27 kPa. The 1 N plot shows that 1 varied slightly and almost linearly with cycle number N in the zone of about 1 2.5%. The variation started to speed up afterward, reaching 37% at N 47. The threshold value of 2.5% for low strain rate is similar to that in the case of initially saturated sample (about 2%). This led us to adopt the same stability loss criterion in terms of changes in 1 rate: for this test, the soil lost stability at 1 2.5%, Nloss 12. The variation of u presents also two distinguished rates: a quick change followed by a slower one. Interestingly, the transition point is at about 2.5% axial strain. Figure 7 presents the cyclic resistance versus cycle number at loss of stability for the soil at 2.20 m at four initial water contents including that in saturated state (34%). Regression lines were added to represent the cyclic resistance curves. It appears clearly that the cyclic resistance strongly depends on the initial water content: the lower the value of wi, the higher the cyclic resistance. In fact, for the tests on initially saturated samples, the samples were in near intact state, and therefore with low density (Table 1). On the contrary, for the tests on initially unsaturated samples, the latter were subjected to a saturation process upon cyclic loading under near constant water content condition. This process induced volume decrease and thus a densification of the samples. As mentioned above,
Figure 7. Results from a test at initial unsaturated stateCyclic resistance versus number of cycle at the instability point.
this densification might increase the soil cyclic resistance. In addition, the lower the initial water content or the lower the initial degree of saturation, the more the densification, and thus the higher the cyclic resistance. Therefore the effect of initial water content should be rather regarded as the effect of initial density.
227
CONCLUSIONS
A differential pressure-based device was developed to monitor the volume change of the soil sample during cyclic triaxial tests. Preliminary tests on the samples of loess taken at 2.2 m depth showed that this device allows good monitoring of the sample volume at a frequency of 0.5 Hz: the curves of volumetric strain are regular and in good agreement with the curves of axial strain measured by LVDT transducer. After the preliminary tests, two other types of tests were conducted, one on initially saturated samples and another on initially unsaturated samples. For the former, soil samples were saturated under 200 kPa back-pressure and then oneway loaded under different deviator stresses that were estimated based on the results from monotonic shear tests. For the latter, soil samples were first subjected to cyclic loading under near constant water content condition with volume monitoring using the differential pressure-based device; when their volume decrease led to the saturation state, they were loaded under undrained condition as the initially saturated samples. The loss of stability was identified from 1-N plot where 1 rate meets rapid change. This rapid change occurs in general at 1.5 to 3% axial strain. This allows the
definition of an instability criterion. According to this criterion, the soil having the highest porosity (at 2.2 m depth) has the lowest cyclic resistance. The effect of initial water content was evidenced on this soil: the lower the initial water content, the higher the cyclic resistance. This effect should be regarded as the effect of the initial densification due to the saturation stage by loading under near constant water content condition. REFERENCES
AFNOR (1994). French Standard NF XP 94202, Gotechnique, Essais in situ, Tome 2. Cui Y.J., Tang A.M., Marcial D., Terpereau J-.M, Marchadier G., and Boulay X. (2007). Use of a differential pressure transducer for the monitoring of soil volume change in cyclic triaxial test on unsaturated soils. Geotechnical Testing Journal, 30(3), 227233. De Gennaro, V., Canou, J., Dupla, J.C., and Benahmed, N. (2004). Influence of loading path on the undrained behaviour of a medium loose sand. Canadian Geotechnical Journal, 41, 166180. Karam, J.P., Cui Y.J., Tang A.M., Terpereau J.M., Marchadier G. (2009). Experimental study on the cyclic resistance of a natural loess from northern France. Soils and Foundations, 49 (3), 421429.
228
Inundation collapse strain and creep measurements for two Campanian pyroclastic soils
P. De Simone
Universit di Napoli Federico II, Napoli, Italy
ABSTRACT: Collapsible soils are characterized by a loose metastable texture causing their collapse to a stable denser state upon wetting. In this paper measurements of collapse strains of two Campanian (Southern Italy) pyroclastic soils by the single oedometer test technique (ASTM D 5333) are presented. The time evolution of strain is analyzed, recognising a capillary rise phase followed by creep. It is shown that both phases can be represented by a power law, the end of capillary rise (EOCR) time separating the two phases being calculated. Pre-inundation and collapse strains are finally expressed in terms of a void index (Burland 1990), using saturation as the reference state. 1 INTRODUCTION b.p.), belonging to the pyroclastic cover of Pizzo dAlvano involved in the Sarno-Quindici May 1998 landslide (de Vita et al. 2006). The specimens of this second soil, sampled at the Pizzo dAlvano site, are indicated in the paper as SVP. 2.2 Methods Tests have been performed on specimens obtained from samples retrieved in-situ by thin walled tube samplers. However in order to minimise the effects of sampling disturbance on the collapse behaviour (Houston 1995) of the AMST B2 soil, some specimens have been trimmed from large undisturbed block samples. Finally, to analyse the relationship between the collapse response of natural and reconstituted specimens, some additional tests have been performed on samples reconstituted by the dry pouring technique (Head 1994). The oedometer tests have been performed increasing the load from 10 to 5000 kPa, according to the usual doubling rule. Each specimen has been inundated under a given vertical stress vin, comprised between 10 and 1200 kPa. A total of 34 tests have been performed, 25 for the CFA flegrean ash and 9 for the vesuvian paleosoil SVP. In particular the CFA tests refer to 7 specimens from block samples (CFA BS), 8 obtained by means of thin walled tube samplers (CFA TS), and 10 reconstituted (CFA Rec), while the SVP are 5 by thin walled tube samplers, and 4 reconstituted (SVP Rec). 3 COLLAPSE STRAIN TIME EVOLUTION
Collapsibility upon wetting of pozzolana, a Neapolitan volcanic ash, is known since ages, having caused many disasters in the metropolitan area of the city of Naples. On the other hand the term pozzolana, which comes from the popular language, is quite vague, the scenario of pyroclastic soils in Neapolitan area and more generally in Campania region (Southern Italy) being more complex. As a matter of fact, different pyroclastic soils result from many different eruptions of volcanoes pertaining to two different districts, namely the Campi Flegrei and the Somma-Vesuvio ones, together with the subsequent soil evolution. 2 MATERIALS & METHODS
2.1 Materials In this paper measurements of collapse and creep strains of two Campanian pyroclastic soils by the single oedometer test technique (ASTM D 5333) are presented. The first soil is a phreatomagmatic base-surge bed from the Agnano Monte Spina eruption (4100 years b.p.), placed in the Campi Flegrei caldera (Dellino et al. 2004) and sampled at the Verdolino site (AMST B2 Ve). The specimens of this soil are indicated in the paper as CFA. The second soil is a residual allophanic material resulting from the weathering of some ashes of Somma-Vesuvio volcano. An andisol, in the language of soil taxonomy (USDA 2006, see e.g. Terribile et al. 2000), a buried soil (paleosoil) sealed by the Avellino pumice eruption (3800 years
A typical time evolution of the collapse strain is shown in Figure 1, where the vertical strain
229
t(m)
0.00 0.1 1 10 100 1000 10000
qk
ua uw wz
(1)
into the mass balance

0.01
qdt nSdz which gives the combined equation

49 - 10kPa 50 - 20kPa 51 - 40kPa 55 - 40kPa 56 - 40kPa 52 - 80kPa 53-150kPa 54-300kPa
(2)
v
0.02
dt
0.03
w nS zdz k (ua uw )
(3)
with the solution t
0.04
Figure 1. CFA TS samples; v log t diagram at collapse.
w nS z 2 k (ua uw ) 2
(4)
v is plotted against the logarithm of time t (in minutes). The diagram refers to the tests performed on CFA TS samples, the stresses at the instant of inundation ranging from 10 (sample 49) to 300 (sample 54) kPa. The observed strain rate was very high in the first minute, then it decreased suddenly the relation between v and log t turning into a linear one. This behaviour is typical of all the experiments performed, the same trend having been observed also with the other material. The process of collapse activated by inundation is thus composed of two successive phases. 3.1 Capillary rise
In Table 1 the corresponding times of capillary rise for tested materials are given, with reference to average values of physical properties, suction values being based on the filter paper technique measurements. An even better agreement has been observed by performing direct measurements of capillary rise time on companion samples. 3.2 End of capillary rise (EOCR) Augustesen et al. (2004) consider an end of primary consolidation instant EOP to separate primary and secondary consolidation. Similarly an end of capillary rise (EOCR) time tEOCR can be introduced to separate the two successive phases of capillary rise and creep. As it can be seen by comparison of Figure 1 with Figure 2, where the vertical strain too is plotted in a log scale, a log v log t representation (t being expressed in minutes) is more suitable to the selection of tEOCR. As a matter of fact each phase can be described by a linear law, namely log v mcr log t ncr log v mcreep log t ncreep (5) (6)
The second part of the process, as suggested by the linear v log t law which remind Buismans (1936) secular settlement, can be attributed to creep (or secondary consolidation). A subsequent comparison of the values of the corresponding secondary compression coefficient C with the ones obtained from the other loading stages (unsaturated and saturated) of the oedometer tests confirms this hypothesis. The first part can be ascribed to an instant settlement associated to the suction reduction produced by inundation. As a matter of fact this fall of suction diffuses upward through the sample together with the saturation front, which travels according to the capillary rise rule. In order to verify this conjecture, a very simple model of capillary rise can be used, namely the one based on the Green-Ampt (1911) wetting front propagation hypothesis (see e.g. Philip 1969). The governing equation is simply obtained in this case by substitution of the Darcys law written in the form (gravity disregarded)
respectively for the capillary rise (5) and creep (6) phases, the time tEOCR being given by the point of interception
Table 1. Time of capillary rise (Green-Ampt approach). Samples CFA BS CFA TS SVP TS n 0.54 0.57 0.73 S 0.41 0.29 0.28 k(m/s) 5E-07 5E-07 1E-07 (kPa) 50 80 200 t(s) 18 8 20
230
t(m)
0.1 0.0001 1 10 100 1000 10000
t(m)
0.1 0.0001 1 10 100 1000 10000
49 - 10kPa 50 - 20kPa 51 - 40kPa 55 - 40kPa 56 - 40kPa 52 - 80kPa 53-150kPa 54-300kPa
0.001
0.001
119 - 20kPa 118 - 40kPa 68 - 40kPa 69 - 80kPa 117 - 80kPa 116-150kPa 74 - 150kPa 75 - 300kPa 76 - 600kPa 77-1200kPa
v
0.01
v
0.01
0.1
0.1
Figure 2. collapse.
CFA TS samples; log v log t diagram at
Figure 4. collapse.
CFA Rec samples; log v log t diagram at
t(m)
0.1 0.0001
0.0001
10
100
1000
10000
0.1 1 10
t(m)
100 1000 10000 59 - 20kPa 60 - 40kPa 61 - 80kPa 62 -150kPa 63 -300kPa
0.001
0.001
v
0.01
0.1
109 - 10kPa 110 - 20kPa 111 - 40kPa 112 - 80kPa 113-150kPa 114-300kPa 115-600kPa
v
0.01 0.1
Figure 3. CFA BS samples; log v log t diagram at collapse.
Figure 5. collapse.
SVP TS samples; log v log t diagram at
log t EOCR
ncreep ncr mcr mcreep
(7)
t(m)
0.0001 0.1 1 10 100 1000 70 - 20kPa 71 - 40kPa 72 - 80kPa 73 - 1200kPa 10000
with the corresponding strain vEOCR given by log v EOCR mcr ncreep mcreep ncr mcr mcreep (8)
0.001
Parameters n and m, which represent respectively values (possibly extrapolated) of logv for t 1 m (logv1m) and of the ratio log v/log t, have been calculated for each test using a least square linear regression line. In figures from 3 to 6 the log v log t plots for the other group of specimens are reported, showing a generally similar behaviour, with the only exception of tests 62 and 63 (SVP TS, Figure 5).
v
0.01 0.1
Figure 6. collapse.
SVP Rec samples; log v log t diagram at
231
Expressions (5) and (6) can be rewritten in terms of tEOCR and vEOCR, deriving the power laws
v v EOCR (t/t EOCR )m
cr
be interpreted as well as a log t linear process, in term of the secondary compression coefficient C defined by (e.g. see Liingaard et al. 2004) C v log t (11)
(9) (10)
v v
EOCR
(t/t
EOCR mcreep
respectively for the capillary rise phase (9), and for creep (10). Numerical values of vEOCR and tEOCR are reported in Figures 7 and 8, while mcr and mcreep are given in Figures 9 and 10. As yet noted before, tests 62 and 63 does not fall into the scheme proposed. In this case the capillary rise phase is practically absent, and the creep phase too behaves very differently, from its start, which is evidently delayed. As a consequence, the corresponding values are omitted. 3.3 Creep
The corresponding numerical values for all the tests performed are reported in Figure 11. As it can be seen, there is a general trend to slightly increase C with the inundation stress vin. In Figure 12 the values of C are plotted together with the ones measured in the same oedometer test before and after inundation. From the picture it is evident that also in terms of creep the transition from an unsaturated to a saturated behaviour is immediate. The values of tests 62 and 63, whose v log t diagrams are shown in Figure 13, confirm the
As shown by Figure 1 (and similar representations for the other tests not reported here), creep can
10
0.1
mcr
0.01
EOCR
1
CFA BS CFA TS
v
0.001
CFA BS CFA TS CFA Rec SVP TS SVP Rec
CFA Rec SVP TS SVP Rec
0.1
0.0001
10
10
100
1000
10000
v in (kPa)
100
1000
10000
v in (kPa)
Figure 7.
6 5 4
Figure 9.
Parameter m for capillary rise phase (mcr).
End of capillary rise strain EOCR.

1
tEOCR (m)
mcreep
3 2 1 0
0.1
10
100
1000
10000
0.01
v in (kPa)
10
100
1000
10000
v in (kPa)
Figure 10. Parameter m for creep phase (mcreep).
Figure 8.
End of capillary rise time tEOCR.
232
0.1
major complexity of the scenario of creep for the material SVP TS (natural samples) at higher stresses. 4 VOID INDEX REPRESENTATION
0.01
C
0.001
CFA BS
0.0001
CFA TS CFA Rec SVP TS SVP Rec
The usual representation of the measurements in a single oedometer inundation collapse test is the graph log v e, from which the value of the collapse potential Ic Ic e (12) 100 100 v 1 e0 can be determined (ASTM D 5333, 3.2.3), e (respectively v) being the variation in void ratio (vertical strain) resulting from inundation. In Figure 14 the values of Ic for all the tests performed are reported. As a sample, in Figure 15 the log v e diagrams for the CFA TS specimens are reported; similar results having been obtained from the other tests. Following Burlands (1990) idea of the normalizing parameter void index Ivo, to compare natural and reconstituted clays, a similar inundation void index Ivc has been introduced to compare saturated and unsaturated states. Assuming saturation as the reference state, the void index Ivc is written e e w100 ew100 e w1000 (13)
0.00001
10
(kPa)
in
v
100
1000
10000
Figure 11.
0.1
Coefficient of secondary compression C.
0.01
C
0.001
SVP TS SVP TS UnSat SVP TS Sat
0.0001
Ivc
10
v in (kPa)
100
1000
10000
Figure 12. Unsaturated, inundation and saturated C. values SVP TS.
where ew100 and ew1000 are the void ratios in saturated condition corresponding to v 100 kPa and 1000 kPa respectively. The resulting graphs for specimens CFA TS are given in Figure 16.
t(m)
0.01 0.00 0.1 1 10 100 1000 10000
10
0.01
v
Ic
0.02
0.03 59 - 20kPa 60 - 40kPa 0.04 61 - 80kPa 62 -150kPa 0.05 63 -300kPa
0.1
0.01
10
v in (kPa)
100
1000
10000
Figure 13.
SVP TS samples; v log t plot.
Figure 14.
Collapse potential Ic .
233
1.5 1.4 1.3 1.2
1.1 1.0 0.9 0.8 10
The effects on the collapse strain of the disturbance associated with the tube sampler technique are evident from comparison between specimens CFA TS and CFA BS, the apparent anomaly of specimens 55 and 56 being a consequence of the greater diameter of their sampler. Following Burland, an inundation void index is finally defined to compare more efficiently the unsaturated and saturated states. REFERENCES
ASTM D 5333 1996. Standard Test Method for Measurement of Collapse Potential of Soils. Augustesen, A. Liingaard, M. Lade, P.V. 2004. Evaluation of Time-Dependent Behavior of Soils. Int.J.Geomech. 4(3): 137156. Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Gotechnique. 40: 329378. Buisman, A.S.K. 1936. Results of long duration settlement tests. Proc. 1st Int,Conf.Soil Mech.Found.Eng. 1;103106. Cambridge Mass. Dellino, P. Isaia, R. La Volpe, L. Orsi, G. 2004. Interaction between particles transported by fallout and surge in the deposits of the Agnano-Monte Spina eruption (Campi Flegrei, Southern Italy). Journal of Volcanology and Geothermal Research. 133: 193210. de Vita, P. Agnello, D. Ambrosino, F. 2006. Landslide susceptibility assessment in ash-fall pyroclastic deposits surrounding Mount Somma-Vesuvius: Application of geophysical surveys for soil thickness mapping. Journal of Applied Geophysics. 59: 126139. Green, W.H. Ampt, G.A. 1911. Studies in soil physics. I. The flow of air and water through soils. J.Agr.Sci. 4; 124. Head, K.H. 1994. Manual of Soil Laboratory Testing. Vol. 2: Permeability, Shear Strength and Compressibility Tests. London. Pentech Press. Houston. S.L. 1995. Foundations and pavements on unsaturated soilsPart one: Collapsible soils. In E.E. Alonso & P. Delage (eds.), Proc. Conf. on Unsaturated Soils. 3: 14211439. Paris. Balkema. Liingaard, M. Augustesen, A. Lade, P.V. 2004. Characterization of Models for Time-Dependent Behavior of Soils. Int.J.Geomech. 4(3):157177. Philip, J.R. 1969. Theory of infiltration. Advances in hydroscience. 5: 215296. Terribile, F. Basile, A. De Mascellis, R. Di Gennaro, A. Vingiani, S. 2000. I suoli delle aree di crisi di Quindici e Sarno: propriet e comportamenti in relazione ai fenomeni franosi del 1998. Quaderni di Geologia Applicata. Pitagora Editrice. 7(1): 6079. USDA, 2006. Keys to Soil Taxonomy. Tenth Ed. United States Department of Agriculture. Natural Resources Conservation Service.
100
v (kPa)
1000
10000
Figure 15.
1
e log v diagrams (CFA TS samples).
-1
Ivc
-2
-3 10 100
v (kPa)
1000
10000
Figure 16. samples).
Inundation void index Ivc (CFA TS
More details on this aspect of the analysis, together with a complete examination of all the tests performed will be given in a forthcoming paper. 5 CONCLUSIONS
The time evolution of strains produced by inundation has been analyzed, recognizing a capillary rise phase followed by creep. An end of capillary rise (EOCR) time is defined and calculated. Creep is analyzed both, as a power law and as a log t process, some more complex behaviours being put forward.
234
Simple shear tests on unsaturated undisturbed specimens containing a weak layer

J.L. Deng
Geosphere Research Institute, Saitama University, Japan
H. Kameya
OYO Corporation, Japan
Y. Tsutsumi & J. Koseki J. Kuwano
Institute of Industrial Science, The University of Tokyo, Japan Geosphere Research Institute, Saitama University, Japan
ABSTRACT: Undisturbed samples were retrieved from a failed slope by the 2004 Niigata-ken Chuetsu Earthquake, Japan in order to evaluate the strength properties of a weak layer (sandy silt, D50 0.053 mm, wL 36.6%, wP 18.5%) which formed the main part of the sliding plane. Simple shear tests were conducted on both unsaturated and saturated specimens, while triaxial compression tests were conducted only on saturated specimens. In saturated tests, shear bands formed along or parallel to the weak layer. The strength parameters obtained for the weak layer of saturated specimens are c 0 and 36.2 in triaxial compression tests and c 0 and 31.7 in simple shear tests. In unsaturated tests, on the other hand, shear bands which formed mainly outside the weak layer had an average inclination of about 45 to the weak layer direction, suggesting that the upper soft rock layer was weaker than the sandwiched weak layer under unsaturated condition. For the upper soft rock of unsaturated specimens, the average strength mobilized on the shear band is c 8.5 kPa and 35.3, which is beyond the failure locus of the weak layer obtained under saturated condition. In other words, the strength of upper soft rock under unsaturated condition is higher than that of weak layer under saturated condition. 1 INTRODUCTION to evaluate the strength of the weak layer which formed the main part of the sliding plane. Under fully saturated condition, a systematic triaxial compression (TC) tests and simple shear tests have been performed by Deng et al. (2009) to assess the effect of full saturation on sliding displacement. In this paper, simple shear tests on unsaturated specimens will be mainly focused, while comparing their results with those obtained with saturated specimens. Hereafter, denotes for effective stress and net stress, respectively, with saturated and unsaturated specimens. 2 INVESTIGATIONS OF A FAILED SLOPE
The 2004 Niigata-ken Chuetsu Earthquake on October 23 in 2004, with a main shock of Mj 6.8, triggered extensive slope failures in Mid Niigata Prefecture, Japan. After the mainshock, many large aftershocks repeatedly struck this area within about two months. According to aerial photo interpretation (MLIT, 2005), 3,791 slopes failed with total breakdown volume of about 100 106 m3 over an area of about 1,310 km2. Besides the main shock and large aftershocks, the rainfall before the earthquake is likely to affect the extensive slope failure as well. This area had been subjected to a continuous heavy rainfall, more than 100 mm/day, leading up to 20th October, so the soil was well saturated and the ground water level was raised from the normal level. With saturated soil and higher ground water level, slope failure is more likely to occur than under normal conditions. Undisturbed samples were retrieved from a failed slope (Fig. 1) and trimmed in laboratory
The failed dip slope is located at Higashi-Takezawa in Yamakoshi Village (currently Nagaoka City), Niigata Prefecture, Japan. It is about 295 m wide, 350 m long with a dip angle of 18 at toe region and 24 at top region of the slope (Figs. 1, 2). Regarding its morphological characteristics, the
235
N Sampling location
Location of the cross section in Fig. 2
Failed slope Elementary school

Figure 1. Failed slope at Higashi-Takezawa site (modified from Kokusai Kogyo, 2004).
Surviving pre-existed sliding mass
Ground water level
Sliding plane
Pre-existed sliding plane ?
24 18 350m Sampling location
Figure 2.
Cross section of the failed slope (modified from SABO Technical Center, 2005).
observations conducted by many researchers (e.g. SABO Technical Center, 2005, Sassa et al., 2005, Chigira & Yagi, 2006, Deng, 2008) support an argument that the earthquake reactivated a preexisting failure plane which then formed most of the present sliding plane. According to the geomorphological properties interpreted from aerial photos and field investigations, a sampling spot was determined which was located on the previous sliding plane and near the current sliding plane (Figs. 2, 3). The properties of samples were expected to be similar to those of the material on the current sliding plane. The observations during sampling conducted by JSCE (2006) in May, 2005 can be summarized as follows: 1. The pre-existed sliding plane was smooth and continuous (Fig. 3). Slickenside or slickenline was not found. 2. On the exposed pre-existed sliding plane, there was strongly weathered siltrock (Fig. 3). At the bottom of this siltrock, a fairly saturated soft sandy silt (D50 0.053 mm, wL 36.6%, wP 18.5%) layer having a thickness of several millimeters was observed. Water was flowing through the observed section (1.5 m wide) from
the sandy silt layer with a maximum flow rate of 100200 cm3/min. Cavities were observed to facilitate the water flow. Due to the water flowing, the soils on the walls of cavities were oxidized. The oxidized soils were considered to be the same soils as those observed on the current sliding. As the ground water level was high at the time of the investigation (Fig. 2), it is expected that at least most of the soft sandy silt layer on the sliding plane had been saturated before the earthquake. 3. Beneath the previous sliding plane, soft fine sandy rock in a lens shape with a maximum thickness of 3 cm was occasionally sandwiched within the weakly weathered siltrock (Fig. 3). However, limitations of this sampling are as follows: 1. The strength of samples may not be a typical strength mobilized on the whole sliding plane. Geologically, the sliding plane should have consisted mostly of the same layer. However, according to Sassa et al. (2005), the sandy silt layer on the west part of the failed slope was much thicker than that at the sampling spot. Such difference in sandy layer thickness may induce a difference in strength.
236
Figure 3.
Sketch showing details of the sandy silt layer and its vicinity (by JSCE, 2006).
2. A single sampling location may not be representative of the whole sliding plane. The sampled blocks were retrieved at the same spot with an area of about 4 m2. This small area may not represent the whole sliding plane with an area of about 105 m2. As most part of the current sliding plane is not exposed, the details of the sliding plane need more investigation. According to the field observation results, the evaluation of the behavior of the thin sandy silt layer on the pre-existed sliding plane is the focus of the experimental work in spite of the above limitations of the sampling. 3 APPARATUS, SPECIMENS AND SIMPLE SHEAR TEST PROCEDURES
3. Instead of a reinforced-membrane or stacked metal rings which are used for ordinary simple shear apparatus, a 0.3-mm-thick membrane was used (Fig. 4). As the membrane is thin and highly deformable (Youngs modulus = 1492 kPa, after Koseki et al., 2005), shear band can form without an additional artificial interference. On the other hand, since the compensative shear stress is not well mobilized at the interface between the specimen and the membrane, the local stress distribution within the specimen would become non-uniform. As the load cell for shear is placed outside the cell (Fig. 4), the effect of piston friction with a maximum value of 2.4 kPa is included in the measured values, while it was neglected in conducting the data analysis. The specimens (width length height 60 60 100 mm, Fig. 5), were retrieved by block sampling method at the location where the previous sliding plane passed through (Figs. 2, 3). In order to get a shear band formed in parallel to the weak layer, the weak layer was arranged to be in the horizontal direction (Fig. 5). The thickness of the weak layer was 14 or 13 mm. In the tests, the horizontal displacement was applied with a maximum shear displacement of 10 or 7 mm. Both tests were unsaturated drained test on undisturbed specimens with a degree of saturation
In order to obtain the strength parameters under unsaturated condition, two simple shear tests were performed on undisturbed specimens. The simple shear apparatus (Fig. 4) is a modified version of conventional simple shear apparatus. Three characteristics of this apparatus are as follows: 1. Two direct-drive (DD) motors are employed to eliminate any time lag on reversal of the loading direction. 2. A cell is used to apply a cell pressure with a maximum value of 1 MPa, while applying a back pressure as well.
237

Figure 4.
Simple shear test apparatus.
v

Soft rock
For all the tests, the rate of loading during shear was kept constant at 0.033 mm/min or 0.055 mm/ min while keeping h0 and v0 constant.
h
Sandy silt layer Soft rock Pre-existed sliding plane
TEST RESULTS AND DISCUSSION
100 mm
Direction of pedestal movement
60 mm 60 mm
Figure 5. Sandwiched specimens in simple shear test.
Sr 96%. After consolidating the specimen isotropically to the prescribed value h0 ( 24 or 49 kPa) with a zero back pressure, drained loading was performed until the vertical stress v reached the prescribed value ( v0 50 or 99 kPa). After a 60-minute drained creep loading by keeping this stress state, drained shearing was applied until the end of test.
Figure 6 shows the results from tests U-1 and U-2. Very limited strain softening was observed in test U-2, i.e. after exceeded 3 mm, the shear stress decreased only 1.4 kPa and then kept constant until the end of the test. In test U-1, the peak stress state was reached after exceeded 9 mm. Figure 7 shows the specimens before and after shearing. For saturated specimen after cyclic shear or monotonic shear, a shear band always formed along or parallel to the sandy silt layer (Deng et al., 2009) showing this layer was weaker than the upper soft rock in the sandy silt layer direction. However, for unsaturated specimen, the sandy silt layer was stronger than the upper soft rock: a crack appeared firstly in the upper soft rock adjacent to the sandy silt layer with an average inclination of about 45 to the sandy silt layer direction, then it was extended and crossed the sandy silt layer. To obtain the strength parameters, the average stress on the shear band was calculated as:
0 2 sin 2 cos 2 v0 h v 0 h 0 2 v 0 h 0 2 cos 2 sin 2
where is the inclination of shear band as shown in Figure 7.
238
60
50
Horizontal shear stress, (kPa)
U-1 U-2 Peak
40
30
20
Test No.
10
' v0
'h0
d/dt
U-1 U-2
(kPa) 50 99
(kPa) mm/min 24 0.03 3 49 0.05 5
0 0 2 4 6 8 10
Horizontal displacement, (mm)
Figure 6. Relationship between horizontal shear stress and horizontal displacement.

Test Before test
Failure loci in simple shear tests and triaxial compression (TC) tests (Deng et al., 2009) on undisturbed specimens are summarized in Figure 8. For the unsaturated specimens, the average strength mobilized on the shear band which is mainly within the upper soft rock is c 8.5 kPa and 35.3. The strength parameters obtained for the sandy silt layer of saturated specimens are c 0 and 36.2 in triaxial compression tests and c 0 and 31.7 in simple shear tests. Thus, the strength of the upper soft rock under unsaturated condition is found to be higher than that of the weak layer (sandy silt) under saturated condition. It should be noted that, under unsaturated
After shearing
Upper soft rock Crack
U-1
Sandy silt layer = 47.0 Lower soft rock
14 mm
Monotonic loading under unsaturated condition
Upper soft rock

Crack in shear band
U-2
= 45.4 Sandy silt layer

13 mm
Lower soft rock
Upper soft rock
Sandy silt layer (shear band)

Cyclic loading under saturated condition (in Deng et al., 2009)
= 0
12 mm
Lower soft rock
(a typical specimen)
Figure 7.
Specimens in simple shear tests.
239
140 120
= tan36.2 '
o
Shear stress on shear band, (kPa)
100 80
strength of upper soft rock under unsaturated condition is higher than that of weak layer (sandy silt) under saturated condition.
= tan31.7o '
= 8.5 + tan35.3o '

60 40 20 0
ACKNOWLEDGEMENTS This study was conducted as a part of the research on Earthquake damage in active-folding areas: Creation of a comprehensive data archive for remedial measures for civil-infrastructure systems that is supported by Special Coordination Funds for Promoting Science and Technology of Japan Ministry of Education, Culture, Sports, Science and Technology. The authors are also grateful to Japan Society for the Promotion of Science (JSPS) for JSPS Postdoctoral Fellowship and Grant-inAid Scientific Research (No. 20-08391) supporting this research. The first author thanks Japan International Cooperation Agency (JICA) for the financial support during his Ph.D. study at the University of Tokyo. REFERENCES
Chigira, M. and Yagi, H. 2006. Geological and geomorphological characteristics of landslides triggered by the 2004 mid Niigata prefecture earthquake in Japan. Engineering Geology 82(4): 202221. Deng, J., Tsutsumi, Y., Kameya, H., Kuwano, J., Kuwano, R. and Koseki, J. 2009. Case study on slope failure at Higashi-Takezawa induced by 2004 Niigata-ken Chuetsu Earthquake (submitted for possible publication in Soils and Foundations). Deng, J. 2008. Case studies on the mechanism of earthquake-induced failure of dip slopes containing a weak layer, Doctor thesis, the University of Tokyo. Japan Society of Civil Engineering 2006. Report on boring and sampling investigation at Chuetsu on creation of a comprehensive data archive, Project Earthquake damage in active-folding areas: Creation of a comprehensive data archive for remedial measures for civilinfrastructure systems, pp. 2_12_47. (in Japanese). Kokusai Kogyo, 2004. http://www.kkc.co.jp/social/disaster/ 200410_niigata_eq/n/09.html. Koseki, J., Yoshida, T. and Sato, T. 2005. Liquefaction properties of Toyoura sand in cyclic torsional shear tests under low confining stress, Soils and Foundations 45(5): 103113. Ministry of Land, Infrastructure and Transport 2005. Information on the Chuetsu Earthquake, http:// www.mlit.go.jp/kisha/kisha05/05/050113_.html (in Japanese). SABO Technical Center 2005. A business report on the examination of countermeasure for landslides at Terano and Higashi-Takezawa in 2004. (in Japanese). Sassa, K., Fukuoka H., Wang, F.W. and Wang, G.H. 2005. Dynamic properties of earthquake-induced large-scale rapid landslides within past landslide masses, Landslides 2(2): 125134.
unsaturated specimens in simple shear test saturated specimens in TC tests (Deng et at., 2009) saturated specimens in simple shear test (Deng et at., 2009)
20
Effective or net normal stress on shear band, ' (kPa)
40
60
80
100
120
140
160
180
200
Figure. 8 Failure loci in simple shear tests and TC tests.
condition, the strength of the sandy silt layer is possibly higher than the upper soft rock, since cracking was initiated in the upper soft rock as mentioned above. The observation on specimens in tests U-1 and U-2 (Fig. 7) and failure locus shown in Figure 8 can provide a piece of evidence to interpret why a small amount of pre-existed sliding mass upon the sampling location (Fig. 2) survived the earthquake: in the earthquake, the sandy silt layer on the sampling location was probably unsaturated so it could be stronger than the adjacent upper soft rock; thus, when the earthquake force was applied on the pre-existed sliding mass, the critical slip surface extended into the upper soft rock instead of always along the pre-existed sliding plane (Fig. 7). However, a detailed stability analysis is needed to provide a more reasonable interpretation. 5 CONCLUSIONS
Simple shear tests on undisturbed specimens containing a thin sandy silt layer were performed under unsaturated condition. The following conclusions can be drawn from the test results: 1. The specimens had a thin sandy silt layer (D50 0.053 mm, wL 36.6%, wP 18.5%) which was 13 or 14 mm thick. However, the shear band with perceptible cracks formed mostly within the upper soft rock, partly passing through the sandy silt layer. 2. The sandy silt layer was stronger than the upper soft rock in the unsaturated tests. Crack appeared firstly in the upper soft rock adjacent to the sandy silt layer with an average inclination of about 45 to the sandy silt layer direction, then passed through the sandy silt layer. 3. For the upper soft rock of unsaturated specimens, the average strength mobilized on the shear band is c 8.5 kPa and 35.3, which is beyond the failure locus of the weak layer under saturated condition. In other words, the
240
Volume change behaviour due to moisture variation in vertisoil from a semiarid region in Pernambuco, Brazil
S.R.M. Ferreira
UFPE, Universidade Federal de Pernambuco, UNICAP, Recife, Brasil
ABSTRACT: Vertisoils are soils that experience change in volume and slickenside are one of the most important morphological features that reflect contraction and fissuring during the drying process, and swelling during wetting. Soils that vary in volume when inundated with water require extra care; regardless of the their use: agriculture, engineering or both. This paper aims at evaluating, through conventional and suction controlled oedometric tests, with various initial humidity and external applied stress, the changes in volume and expansion stress of the soil, due to change in water content in a vertisoil from Petrolndia-PE. The results show that the soil expansion, contraction or collapse, depend on the initial humidity and the external stress applied. The volume variation, in the course of time, due to change in soil humidity, could there be expansion, expansion jointly with collapse, and only collapse or contraction. The conclusion is that the volume changes due to water content variation are associated to the soil initial condition and the conditions that it has been submitted to. 1 INTRODUCTION 2 MATERIALS AND METHODS
One of the main morphological characteristics of the vertisoils is its capability of contracting and fissuring during the drying process and that one of expanding during moistening. As a result, the soil presents slickensides. The volumetric instability phenomenon (contraction and expansion or collapse) of unsaturated soils due to inundation is complex, and is affected by various factors. It depends on the type of soil (origin and formation), the existent climatic determinants, the stress state and so on. The presence of expansive soils in engineering works can cause serious damage (fissures, ruptures and cracks) to the buildings when they are not adequately analyzed during the project and construction phase. This work analyses the characteristics of volume variation due to changes in the moisture content of a vertisoil from Petrolndia CountyPE (520 km away from Recife). The selected place for studies is located in Jatob Basin. The local geology is represented by sediments of the Aliana Formation, which is constituted by siltstones, shales and limestones of brown and reddish colors. The soil from this training has clay or silt features, with dark colors, usually from dark grey to reddish one, Melo (1980). In the field, one has observed desiccation cracks.
The developed program of geotechnical investigation is consisted by two parts. In the first one, in the field, wells investigation with samples collection (undisturbed and dented) and determinations of the soils physical indices were made. In the second one, in the laboratory, characterization assays (soil gradation, consistency limits), characteristic curve, microstructure analysis (O-optic), and chemical ones have been made. The characteristics of volume change associated with the stress state and the moisture variation have been analyzed through simple and conventional oedometer tests and controlled suction, which the undisturbed samples have been statically loaded, in constant humidity, to a determined stress, and then the suction has been progressively inundated or reduced. To analyze the pre-moistening or drying effect, undisturbed samples were molded in rings of 80 cm2 of area and 3.0 cm of height. They have been put in desiccators, with different concentrations of sodium chloride or sulphuric acid, for 10 months and, so they have been molded in rings of 40 cm2 and 2.0 cm of height to make the simple oedometer test. And to analyze the effect of desiccation on stress, undisturbed samples were molded, charged until a determined stress and, later, it was allowed the loss of humidity under
241
stress in the room temperature until the stabilization of the deformations that occurred between 60 and 70 days. 3 RESULTS AND ANALYSES
The soil gradation of the vertisoil from Petrolndia presents discreet variation of the percentage of constituents along its profile. The percentage of (0,002 mm) is superior to 54% and the percentage of sand (grains diameter between 0,02 mm and 2 mm) is inferior to 9%. The relation silt/clay decreases with the depth, indicating that there is a translocation of the finer material, from the surface to sub-superficial horizons, carried by water and percolated by fissures. The vertisoil matrix is characterized by a fine textured, compact, predominantly of silicate clays permeated by micritic calcite crystals that make up much of the silt and fine sand. There are papules in great quantity, sometimes lamellas and probably originated from the filling of flattened channels and pores, destroyed and pushed in the soils matrix, due to the action of the high activity clay (Ta) that occurs in this soil, Figure 1a. Calcitic nodules are typical and common as well as reprecipitation of calcium carbonate (CaCO3) in the pores walls, Figure 1b. In the dry period, one has observed in the field that the fissures, in the soils surface, had thicknesses that varied only a few millimeters from 120 mm, reducing the dimension with the depth, and the extension reached 2.0 meters (observed in the manhole). For water penetration, the largest fissures are more important than a bigger number of slim fissures, because, with the raise of moisture, the soil expands and the slimmest fissures are progressively restricted, while the largest ones can remain open for a longer period of time. In the beginning of the moistening process, the fissuring intensity is as important as the width and the depth
of the individual fissures. With the rains, the soil absorbed water from the surface and from the interior of the fissures, and the soil became more plastic, with the clay particles in process of expansion and the appearance of micro-relief. The surface is constituted of blocks of soils of irregular shapes, being detected, in an area of 100 m2, chosen in an aleatory way in the field, around 190 blocks, each one with average area of 0.53 m2 (Fig. 2a). One has also observed that rain was enough to make some superficial fissures disappear completely, Figure 2b. In the vertisoil, predominated of free calcium carbonate on sodium (Ferreira, 1995), thus presenting a smaller number of cracks increased bandwidth, confirming the details of Ahmad (1983). The soil moisture, measured from the surface to a depth of 3.50 m, varies, in the rainy period, from 42% to 20.90% and, in the dry period, from 14.54% to 21.03% (Fig. 2c). From the depth of 2.50 m on, there has been not, in the two years of observation, significant variation in the moisture content between the dry and rainy periods, which indicates that this is the active zone of change in the level of water. The soil expansion stress, obtained by the method, with a constant volume, increases with the depth in the rainy period, and, in the dry period, it decreases until 2.5 m; from this depth on, it remains basically constant (Fig. 2d). In the same way, one has observed the relation between the expansion stress and the effective stress with the depth (Fig. 2e). Until the depth of 2.5 m, there is considerable influence of the climatic conditionants on the moisture, the expansion stress and the depth which the fissure occurs in. In the field, the fissures were observed until 2.0 m of depth, in the dry period. The behaviour of volume variation due to inundation is analyzed considering the influence of the inundation vertical stress, the expansion stress, the initial humidity and the drying under stress.
Figure 1.
a) Micrograph of papule, fractures and flattened pores. b) Micrograph of calcitic nodule.
242
Figure 2. a) Fissures and micro-reliefs in the dry period. b) Fissures e micro-reliefs in the rainy period, c) Active zone. d) In situ stress and expansion stress and e) Ratio between in situ stress and expansion stress in a vertisoil from Petrolndia.
3.1
Influence of the Consolidation Vertical Stress
The strain versus time curves, after the inundation in the simple oedometer tests, are shown in the Figure 3. The deformation process of expansion and/or collapse is, sometimes, simultaneous. Therefore, what is measured is the resultant, which is a function of the vertical stress, empty spaces and moisture indices (state of tension) which the soil is found in before being inundated with water. This way, to a determined state of stress, the deformation, due to inundation, measured with the time, is only of expansion, to stress under 160 kPa, expansion and collapse in the 240 and 400 kPa (initially, the soil compresses until 8 minutes, then it expands until 240 minutes, to compress afterwards until the deformations stabilization) and only of collapse to the stress that are superior to 640 kPa. The deformation process of expansion or collapse can be divided in three phases, according to the time, due to inundation under stress of constant external consolidation. i. initialto intervals under one minute, in which small deformations in the soil are observed and the water humidifies exclusively the periphery; ii. primaryto intervals between 1 and 300 minutes, the water percolates from the periphery to the centre, humidifying the soil progressively
(a function of hydraulic conductivity) and the deformations occur with higher intensity; iii. secondaryto intervals superior to 300 minutes, the water humidifies the central nucleus, the empty spaces are almost completely filled with water and the deformation velocity decreases (Fig. 3). Roo (2006) considers that the initial deformations are associated with the microstructure, while the primary and secondary deformations are associated with the macrostructure. The variation of the expansion or collapse potential, with the consolidation vertical stress, with the void ratio and with the saturation degree before the soil inundation is shown in Figure 4. To tensions inferior to 312 kPa (Fig. 4a), voids ratios superior to 0,745 (Fig. 4b) and saturation degrees inferior to 63.90% (Fig. 4c) the Petrolndias vertisoil expands when there is a raise in the soil moisture, characterizing the expansion region. To stress superior 312 kPa (Fig. 4a), void ratio inferior to 0.745 (Fig. 4b) and saturation degrees superior to 63.90% (Fig. 4c). Petrolndias vertisoil collapses when there is a raise in the soil moisture characterizing the collapse region. There is a stress (crit 312 kPa), an index of empty spaces (ecrit 0.745) and a degree of saturation in water (Srcrit 63.90%) that is critical, where the soil doesnt expand nor collapse, when inundated. (Fig. 4).
243
25
Collapse or expansion deformation (%)
Expansion or collapse potential (%)
20 15 10 5 0 5 10 15
0,1
10 kPa 80 kPa 240 kPa 640 kPa
EXPANSION
Secondary expansion or collapse
20 15 10 5 0 5 10 SP ou CP = -7.57log + 18.88 r 2 = 0.98
a)
Expansion
Block unit (dry)
Dry
crit = 312 kPa
COLLAPSE
Initial expansion or collapse Primary expansion or collapse
Collapse
Dry
10
100
1000
10000
10
100
1000
10000
Time (minute)
Consolidation vertical stress (kPa)

20
Figure 3. Behaviour of the deformation variation in the period of time under stress with the advance of the dampening front. (a) small alteration in the soil moisture; (b) there is change in the moisture of the soils periphery; (c) only the central nucleus maintains the initial humidity; (d) there is change in the moisture of the whole soil.
15
b)
10 5 0
ecrit = 0.745
Expansion
3.2
Influence of the initial moisture
Collapse
5 10 0,65 SP ou CP = 2083.2e2 - 2901.1e + 1004.7 r 2 = 0.89 0,7 0,75 0,8
The free expansion variation increases (approximately linearly) with the decrease of the initial moisture (addition of the suction) and with the decrease of the vertical stress of consolidation before the inundation (Fig. 5a). The free expansion of the vertisoil, under studies, in moisture content of 17.41% is of high expansivity, according to the Vijayvergiya & Ghazzaly (1973) criteria, which takes in consideration a consolidation stress of 10 kPa. The values of physical indices of the samples, which were previously dampened or dried before being inundated and the values of expansion stress, determined by the charging methods after expansion, with different consolidation and expansion vertical stress (Method 1); and collapse under stress (Method 2) to the initials moistures contents 22.58%, 20.80%, 17.41% and 7.76%, are presented in Table 1. The highest expansion stress values were obtained to the lowest humidity values (7.76%), saturation degree (35.22%) and empty spaces indices (0.57) and higher suction values (117 MPa) and dry apparent specific weight (16.96 kN/m3). The previous humidifying process of the soil causes a reduction in the expansion stress, while the desiccation causes an addition; that shows that, in the field, the climate conditioning factors have considerable influence on the soil expansion stress. The curves of specific volumetric deformation with the soil consolidation vertical stress (v log) with different initial moistures, and inundated in the stress of 10 kPa and 160 kPa, are presented in Figure 5b. To a same initial moisture, the expansion due to inundation decreases with the consolidation vertical stress rise; and to a same
Void ratio before the inundation

20
15
c)
10 5 0 5 10 55
Expansion
Sr crit = 63.90%
Collapse
2 SP ou CP = 0.175Sr - 24.383Sr + 843.43 2 r = 0.84
60
65
70
Degree of saturation before the inundation Sr (%)
Figure 4. Variation between the expansion potential or collapse potential obtained in the simple oedometer tests. a) with the consolidation vertical stress; b) with the void ratio before the inundation; c) with the saturation degree before the inundation.
consolidation vertical stress, the soil expansion is reduced with the rise of initial humidity (Fig. 5b), because the process of previous moistening causes an expansion before the charging. Under the consolidation vertical stress of 160 kPa, the soil inundation causes collapse to the initial moistures of 22.58% and of 20.80%, and expansion to the initial moistures of 17.41% and 7.76% (Fig. 5c). A similar behaviour was found by Presa (1982) to a soil with approximately the same index of empty spaces and different initial moistures, consolidated under the stress of 200 kPa. The initial humidity and the stress to which the soil is submitted have
244
30 a) 25 El = -1.07W + 37.26 r 2 = 0.96
great importance in the behaviour of the volume variation when the soil is inundated. 3.3 Drying under stress
"Free" expansion (%)
20 15 10 5 0
Inundation stress 1 kPa 10 kPa
El = -0.937W+ 26.63 r 2 = 0.98
10 15 20 Initial water content (%)

b) Inundation stress 10 kPa
25
20 Especific volumetric strain (%)
10
10
20
Wo = 22.68% Wo = 20.80% Wo= 17.41 Wo = 7.76%
10
100 1000 Vertical consolidation stress (kPa)
10000
5 Especific volumetric strain (%)
c) Inundation stress160 kPa
10 Wo = 22,68% Wo = 20,80% Wo = 17,41% Wo = 7,76%
15
20
10
100 1000 Vertical consolidation stress (kPa)
10000
Figure 5. Influence of the initial moisture: a) in the free expansion; b) in the compressibility; c) in the expansion or collapse under stress. Table 1.
The process of soil deformation because of desiccation, under stress, with the time, is much slower than during the humidifying process. This is due to the way the humidity is transferred, which during inundation, is processed in the liquid phase, while in the desiccation, it is partially processed in the vapor phase. The effect of the desiccation in the total deformation of the soil is greater to lower stress than to higher tensions (Fig. 6a). The deformations measured during the desiccation result from the addition of three components: immediate compression, deformation due to vertical stress magnitude with the time and the retraction itself, with the time. To stress under 40 kPa, the time required for 50% of the deformations to occur is of, at least, 8 days, while to stress above 320 kPa is of, at most, 2 days. This is explained by the fact that, to consolidation stress under 40 kPa, the immediate compressions and the compressions due to the stress effect are reduced with the time when compared to the ones caused by the desiccation, prevailing over the ones resultant of retraction, which are slower. And to the stress above 320 kPa, the immediate compressions and the ones due to the stress magnitude effect have more importance and occur in a faster way than the ones caused by the desiccation. The region between the curves of charging, at a constant humidity, and the curves of desiccation under stress limits the contraction. (Fig. 6b). The curves of inundation under stress and charging to constant moisture limit two regions: the first is of expansion to stress inferior to 277 kPa, (in Petrolndias vertisoil, there is expansion when the moisture increasesFig. 6b); and the second is of
Influence of the initial humidity in the deformation and stress of expansion. Relation expansion deformationstress s (%) s 11.82 log 29.73 s 7.84 log 18.66 s 8.62 log 16.64 s 6.36 log 11.60 (kPa) r2 0.99 r2 0.99 r2 0.99 r2 0.99 Expansion stress (kPa) Method 1 456 333 153 85 Method 2 328 239 83 67
W (%) 7.76 17.41 20.70 22.60
Initial suction (kPa) 117.000 5.000 700 200
eo 0.597 0.801 0.885 0.936
d (kN/m3) 16.96 15.05 14.37 14.00
Sr (%) 35.22 59.24 63.39 65.43
Winitial moisture in the dry period; eovoid ratio; Srdegree of saturation in water; sexpansion deformation (s 100 H/Hi where: H is the bodys height variation due to inundation and Hi the specimens height before the inundation); stress; ddry apparent specific weight; r2correlation coefficient; Method 1-charging after expansion with different consolidation vertical stress; Method 2-expansion and collapse under stress.
245
a)
Specific volumetric strain (%)
4 10 kPa 6 20 kPa 40 kPa 80 kPa 8 160 kpa 320 kPa 640 kPa 10 1280 kPa
12
The order with which the soil is submitted to stress or inundated influences the expansion stress value. The inundation leads to soils volume increase (expansion) or decrease (collapse) in function of the initial moisture and the vertical stress applied to the soil. Finally, the factors that limit the regions of expansion, collapse and contraction are the curves of inundation, the charging to constant humidity and the desiccation under stress. ACKNOWLEDGMENT To Prof. Mauro C. dos Santos from UFRPE who made possible the identification and microstructural analysis of soils. The investigation described above was performed with the support of Foundation for Support of Science and Technology of the State Pernambuco (FACEPE). REFERENCES
Ahmad, N. Vertisoils. In: WILDINY L.P., SMECK, N.E. & HALL, G.F. Pedogenesis and Soil Taxonomy. Amsterdam, Elsevier, 1983. p. 91123. Ferreira, S.R.M. Colapso e expanso de solos naturais no saturados devidos inundao. Rio de Janeiro, Universidade Federal do Rio de Janeiro, 1995. 379 p. (Tese de Doutorado). Melo, J.G. Estudo hidrolgico da bacia sedimentar do Jatob (PE) recursos explorveis e dispositivos de captao. Recife, Universidade Federal de Pernambuco, 1980. 332p. (Dissertao de Mestrado). Presa, E.P. Deformabilidad de las argillas expansivas bajo succin controlada. Madrid, Universidad Politcnica de Madrid, 1982. 663 p. (Tesis Doctoral). Roo, S.M. Identification and classification of expansive soils IN: Al-Rams, A.A. & Goosen, M.F.A., ed. Taylor & Francis / Balkema, London, 2006. p. 1524. Vijayvergiya, V.N. & Ghazzaly, O.I. Prediction of swelling potential for natural clays. IN: INTERNATIONAL CONFERENCE ON EXPANSIVE SOILS 3rd, 1973. Proceedings. Hayfa, 1973. 1:227236.
10
20
30
40
50
60
70
80
90
Time (day)
25
Expansion
b)
20 15 10 5 0
REGION OF EXPANTION
Curves
Inundation under tension Charging at constant moisture Desiccation under tension
Volumetric strains (%)
REGION OF COLLAPSE REGION OF CONTRACTION
Collapse
5 10 15 1
10
100
1000
10000
Inundation vertical stress (kPa)
Figure 6. Volumetric deformation with the addition of stress, desiccation and inundation: a) volumetric deformation with the simultaneous addition of stress and desiccation; b) expansion, collapse and contraction regions.
collapse to stress superior to 277 kPa. The whole formed by the curves of inundation under stress, charging to constant moisture and desiccation under stress constitutes important information under the influence of the stress trajectory in the behavior of volume variation due to change in the soil moisture. They limit regions of expansion, collapse and contraction. 4 CONCLUSIONS
This work concludes that soils with lower initial moisture and a higher exterior applied vertical stress present bigger expansion due to the raise of moisture.
246
Constant water content triaxial compression tests with a compacted soil

G.B. Georgetti & O.M. Vilar
University of So Paulo, So Carlos, So Paulo, Brazil
ABSTRACT: This paper presents and discusses the test results of a compacted unsaturated soil. Constant water content (CW) triaxial compression tests were performed with soil suction measurement. The soil suction behavior during shearing and its influence on stress-strain curves are presented and compared to the behavior of saturated samples tested under consolidated-drained and consolidated-undrained conditions. The influence of both net confining stress and soil suction on the shear strength envelope is also addressed. Although some non-linearity between shear strength and suction could be observed, a planar failure surface could well be adjusted to the experimental data. 1 INTRODUCTION strength of the non-saturated soil, as well as the development of suction during shearing were investigated using constant water content triaxial compression tests. 2 SOIL PROPERTIES
Shear strength is a fundamental soil property to deal with some typical geotechnical problems, such as stability of slopes and bearing capacity of foundations. It is known to depend on many factors, like range of stress applied and drainage characteristics of the soil, among others. As far as unsaturated soil is concerned, the influence of soil suction has long been recognized (Bishop & Donald 1961, Gulhati & Satija 1981, Fredlund et al. 1978) and various types of tests with suction control or measurement are available. Consolidated-drained (CD) and constant water content (CW) triaxial compression tests are some examples of both unsaturated tests with suction control and measurement. In the former, water and air pressures are kept constant throughout the test, while in CW tests, after imposing suction and consolidation, the specimen is loaded maintaining the air pressure constant and drained and water pressure in an undrained mode. Test results have been interpreted according to an effective stress approach or based on stress state variables, such as net normal stress and suction. When stress state variables are used, an extension of Mohr-Coulomb envelope to encompass soil suction has been introduced (Fredlund et al. 1978), which is represented by a planar surface. Depending on the soil type and suction range, among other factors, the matric suction influence on the shear strength is non-linear, as reported by many authors, e.g. Escario & Sez (1986) and Rhm & Vilar (1995). In this paper, the shear strength of a compacted unsaturated soil is studied. The influence of suction on the stress-strain curves and on the shear
The soil used in the tests is a clayey fine sand of colluvial origin subjected to laterization, a typical pedogenetic process of tropical areas. The soil index properties are shown in Table 1 and according to Unified Soil Classification System (USCS), it is classified as clayey sand (SC). The Standard Proctor compaction test provided a maximum dry density of 1.76 g/cm3 and an optimum water content of 15.3%. 3 TESTS PERFORMED
The experimental program included the measurement of soil water retention properties and triaxial compression tests performed with both saturated and unsaturated soil. Consolidated-drained (CD) and consolidated-undrained tests (CU) were used to characterize the saturated condition, while constant water content (CW) tests were performed
Table 1. s (g/cm3) 2.69 Soil index properties. wL (%) 38 wP (%) 21 Sand (%) 61 Silt (%) 5 Clay (%) 34
247
to characterize the shear strength under unsaturated condition. All the specimens tested were dynamically compacted to an average dry density of 1.68 g/cm3 and water content of 14.3%, corresponding to a compaction degree of 95% and a moisture deviation of 1%. 3.1 Soil water retention curve
After molding the specimens, the water retention characteristics of the soil were obtained following two paths: drainage or desorption, after relaxing suction to zero, and wetting or sorption, after airdrying the sample. Filter paper technique (ASTM 2003) was used and in order to better define the initial portion of the drainage path, the porous plate funnel method was also employed (ASTM 2002). In the drainage path, the compacted specimens were firstly saturated through capillary rising and then allowed to dry until target moisture content. Next, the specimens were put in intimate contact with the filter paper allowing for the equilibrium of potential. In the porous plate funnel, target suctions of 1, 3 and 10 kPa were applied allowing for the equilibrium of water content of the specimens. In the wetting path, the specimens were dried out under laboratory-controlled relative humidity. To reach the target moisture content, some water was added to the samples, which were then put in contact with the filter paper. Filter paper Whatman No. 42 was used, and the suction was computed through Chandler et al. (1992) calibration curves. The experimental data obtained were adjusted with the van Genuchten (1980) equation: 1 w wr (ws wr ) n m [1 ( s ) ] (1)
3.2.1 Saturated soil testing procedure In CD and CU triaxial tests, specimens were saturated using backpressure and B pore water pressure parameter larger than 0.97 was used to indicate saturation. Next, the specimens were isotropically consolidated under effective confining pressure ( 3c) of 40, 100, 150 and 300 kPa in CD tests and 50, 150 and 300 kPa in CU tests. From the consolidation process, the shearing strain rate was calculated as 1.3 103% /s for CD tests and was also used for the CU tests. 3.2.2 Unsaturated soil testing procedure The procedure to test specimens under unsaturated condition initiated with a reduction of soil suction to zero by capillary rising. In the triaxial tests, the soil suction was imposed simultaneously to consolidation by applying net confining stress (3 ua) and matric suction (ua uw) and allowing some water to drain. The drainage valve was then closed, allowing for pore pressure equilibrium. Thus, the initial suction was computed from the difference between air and water pressures. The specimen was then sheared at constant water content, i.e. with air pressure in drained mode and water pressure in undrained mode. The pore water pressure measurement was taken at the base of the specimen, which allowed knowing the soil suction development at any instant. Nevertheless, the volume change was not measured during the shearing process. The strain rate used for CW tests shearing was 3.3 104 % /s. The choice relied on the strain rates used in CW tests by Bishop & Donald (1961) on Braehead silt (4.7 104 % /s) and Gulhati & Satija (1981) on Dhanauri clay (6.7 104 % /s) and also considering that a sandy soil had been studied. Preliminary tests using that strain rate showed pore water pressure stabilization as maximum stress had been reached. Three series of CW tests were performed with (3 ua) of 50, 150 and 300 kPa. In each series, varied suctions were applied to the samples. 4 RESULTS AND ANALYSIS
where ws and wr are saturation and residual water content, respectively, , m and n are form parameters of retention curve, and s is the soil suction. 3.2 Triaxial compression tests
Triaxial compression tests were performed with cylindrical samples of 100 mm height and 50 mm diameter. A data acquisition system and a microcomputer allowed recording data of water pressure, volume change, axial strain and load during consolidation and shearing processes. To test unsaturated soil, the triaxial cell used was similar to the cell described by Fredlund & Rahardjo (1993), with a high air entry ceramic disk of 5 bar (500 kPa) coupled to the cell pedestal. Additionally, the axis-translation technique (Hilf 1956) was employed to turn pore water pressure measurable in a large range of values.
The soil water retention curves are shown in Figure 1, where the experimental points were adjusted using van Genuchtens equation. Some hysteresis can be observed and the curves suggest a bi-modal pore distribution, with an air entry suction of about 6 kPa for the larger voids and 5000 kPa for the smaller pores. To take into account the bi-modal character of the curves, two equations were adjusted to the experimental points of each suction path. Table 2 shows the adjustment parameters of van Genuchtens equation.
248
Figure 1.
Soil water retention curves.
Table 2. Adjusted parameters of van Genuchten equation. Curve Drainage 1 Drainage 2 Wetting 1 Wetting 2 ws (%) wr (%) (kPa-1) m n r
20.0 13.5 0.1982 13.5 0.0 0.0001 19.6 10.1 0.8642 10.7 0.5 0.0001
0.056 20.385 0.81 0.651 2.532 0.98 0.121 3.221 0.99 2.798 2.285 0.99
Figure 2. Results of CD tests: (a) deviator stress versus strain and (b) volumetric strain variation.
Figure 2 depicts the stress-strain and volumetric strain curves of CD tests, while Figure 3 shows the stress-strain curves together with pore water pressure variation with strain of CU tests. A different behavior can be observed from the stressstrain curves. CU tests present peak shear strength for low deformation and a decrease in stress as the strain increases in a strain softening behavior (Fig. 3a). CD tests indicate some discontinuity in the stress-strain curves for strains close to the peak stress of CU tests (Fig. 2a). The stress continues to increase showing a plastic behavior with maximum stress reached at axial strain of about 14% for confining stresses of 150 and 300 kPa. For the lower confining stresses, the deviator stress tends to reach a peak at about 1% of strain and slightly reduces after that point. Regarding volumetric variations of CD and pore water pressures of CU tests, the results have presented a coherent behavior as the tests with the larger confining stresses show volumetric reduction and pore water pressure tending to increase with the confining stress. Tests with lower confining stresses present an opposite behavior. Figure 4 shows the effective saturated shear strength envelope considering the maximum and the peak deviator stress reached in the tests. Effective shear strength envelopes obtained from CD
Figure 3. Results of CU tests: (a) deviator stress versus strain and (b) pore water pressure variation.
249
Figure 4. Shear strength envelopes of (a) CD and (b) CU tests.
and CU tests (Eqs. 2 and 3, respectively) presented reasonably close parameters.
20 tan 31 (kPa) 25 tan 29 (kPa)
(2) (3)
Results of CW triaxial tests under net confining stress of 50 and 300 kPa are shown in Figures 5 and 6. The suction values indicated in the graphs correspond to the suctions at the beginning of the shearing process. Analyzing the results of tests with specimens submitted to net confining stress of 50 kPa (Fig. 5a), one can observe some difference in the curve format as the initial suction increases and also that the curves reached maximum stress at strain lower than 5%. If the deviator stress is normalized through (3 uw) operating during shearing, the shape of stress-strain curves change and peak stress can be observed at axial strain of about 2% (Fig. 5b). Moreover, the suction first decrease and then tends to slightly increase after the peak of stress for the lower initial suctions and decreases monotonically for the larger initial suction (Fig. 5c). As the net confining stress increases, both forms of stress-strain curve representation show a plastic behavior. The shear strength tends to increase with suction in deviator stress versus strain representation, although its influence tends to vanish in the deviator stress normalized through (3 uw) versus strain curve for the net confining stress of 300 kPa (Figs. 6a, b). In Figure 6c it is possible to observe that the matric suction tends to decrease towards a constant value during shearing. Comparing the test results with both unsaturated and saturated specimens, it can be observed that
Figure 5. Results of CW tests under net confining pressure of 50 kPa: (a) deviator stress versus strain, (b) deviator stress normalized through (3 uw) versus strain and (c) suction variation.
CW deviator stress-strain curves are closer in format to the curves obtained in CD tests than in CU tests. When analyzing the stress-strain curves represented using the deviator stress normalized through (3 uw), results of tests performed under net confining stress of 50 kPa indicate some similarity with the behavior of CU curves. Concerning deviator stress behavior and suction development in CW tests, similar results had been found by Rahardjo et al. (2004) while testing a compacted residual soil from the Jurong sedimentary formation. The influence of net confining stress on the shear strength can be analyzed in Figure 7, which compares two specimens tested under the same (ua uw)
250
Figure 7. Results of CW tests under initial suction of 32 kPa: (a) deviator stress versus strain and (b) suction variation.
Figure 6. Results of CW tests under net confining pressure of 300 kPa: (a) deviator stress versus strain, (b) deviator stress normalized through (3 uw) versus strain and (c) suction variation.
Figure 8. Cohesion intercept versus suction to CW tests under net confining stress of 50 kPa.
and different (3 ua). It can be observed that the net confining stress affects not only the shear strength, which had already been expected, but also the pattern of suction development. For the lower net confining stress, after a decrease, the suction tends to increase up to an almost constant value. For the larger net confining stress, after a slight decrease the suction increases, but at a lower rate than the specimen tested with 50 kPa of confining stress. Figure 8 represents the cohesion intercept evolution with suction from CW tests with net confining stress of 50 kPa together with the effective cohesion of CD shear strength envelope. The intercepts were obtained by adjusting a parallel of the CD
envelope to the Mohr circles of CW tests. Some non-linearity can be appreciated in the graph. Figure 9 shows the experimental points fitted with the Mohr-Coulomb extended failure envelope suggested by Fredlund et al. (1978). The coordinates of points were calculated as [(1 3)/2], [(1 3)/2 ua] and (ua uw). Equation 4 presents the obtained failure surface converted to a ( ua) and (ua uw) space, which presented a determination coefficient of 0.99.
19 ( ua )tan 32 ( u a uw )tan 24 (kPa)
(4)
251
A planar failure surface could well be adjusted to the experimental data, although some nonlinearity between shear strength and suction had been observed. ACKNOWLEDGEMENTS The authors would like to acknowledge CAPES (Coordenao de Aperfeioamento de Pessoal de Nvel Superior) for the financial support given to this research. REFERENCES
ASTM. 2002. Standard test methods for determination of the soil water characteristic curve for desorption using a hanging column, pressure extractor, chilled mirror hygrometer, and/or centrifugeD6836. In: Annual Book of ASTM Standards, Vol. 04.09. West Conshohocken: American Society for Testing Materials. ASTM. 2003. Standard test method for measurement of soil potential (suction) using filter paperD5298. In: Annual Book of ASTM Standards, Vol. 04.09. West Conshohocken: American Society for Testing Materials. Bishop, A.W. & Donald, I.B. 1961. The experimental study of partly saturated soil in the triaxial apparatus. In: Soil mechanics and foundation engineering; Proc. 5th intern. conf., Paris, 1961. Paris: Dunod. Chandler, R.J., Crilly, M.S. & Montgomery-Smith, G. 1992. A low-cost method of assessing clay desiccation for low-rise buildings. In Institution of Civil Engineering; Proc., May 1992. Escario, V. & Sez, J. 1986. The shear strength of partly saturated soils. Gotechnique 36(3): 453456. Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics for unsaturated soils. New York: John Wiley & Sons, Inc. Fredlund, D.G., Morgerstern, N.R. & Widger, R.A. 1978. The shear strength of unsaturated soils. Canadian Geotechnical Journal 15(3): 313321. Gulhati, S.K. & Satija, B.S. 1981. Shear strength of partially saturated soils. In Soil mechanics and foundation engineering; Proc. 10th intern. conf., Stockholm, 1519 June 1981. Rotterdam: Balkema. Hilf, J.W. 1956. An investigation of pore-water pressure in compacted cohesive soils. PhD Thesis, Faculty of Graduate School of the University of Colorado, Denver. Rahardjo, H., Heng, O.B. & Choon, L.E. 2004. Shear strength of a compacted residual soil from consolidated drained and constant water content triaxial tests. Canadian Geotechnical Journal 41: 421436. Rhm, S.A. & Vilar, O.M. 1995. Shear strength of an unsaturated sandy soil. In Alonso E.E. & Delage P. (eds.) Unsaturated soils; Proc. 1st intern. conf., Paris, 68 September, 1995. Rotterdam: Balkema. Van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44(5): 892898.
Figure 9.
Extended Mohr-Coulomb failure envelope.
Although some non-linearity was observed in the results, it is interesting to notice that a planar surface could represent the experimental data for the range of net confining stress and suction used in the tests. Another aspect that can be noticed is that the effective cohesion (c ) and the angle of shearing strength () obtained from the planar surface are close to the values of the saturated effective stress envelope. The third shear strength parameter obtained through the planar surface, i.e. the friction angle with respect to suction (b), is lower than , which is in accordance with the theory of unsaturated soil shear strength. 5 CONCLUSION
Triaxial compression tests were performed in order to define the shear strength behavior of a typical compacted lateritic Brazilian soil. CD and CU tests with saturated specimens showed some difference as the behavior of the CD deviator stress-strain curves was similar to a plastic one and CU curves presented a peak of stress. Nevertheless, the volumetric strain and pore water pressure during these tests showed coherence. The unsaturated shear strength was evaluated through CW tests. Deviator stress-strain curves presented a typical plastic behavior, although normalized stress-strain curves had shown a typical strain softening behavior for the lower net confining stress used in the tests. As expected, the soil suction increased the shear strength of the soil. During the shearing process, the suction showed a decrease at the beginning, which then tended to equilibrium.
252
The relationship between the critical state shear strength of unsaturated soils and the soil-water characteristic curve
J.A. Infante Sedano & Sai K. Vanapalli
Civil Engineering Department, University of Ottawa, Ontario, Canada
ABSTRACT: This paper presents comparisons between the measured critical state shear strength of an unsaturated glacial till using the modified ring shear apparatus and the predicted shear strength using the SWCC measured following three different techniques: (i) conventional SWCC without application of any stress on the specimen; (ii) the SWCC measured with an equivalent net normal stress on the specimen that was used during shearing stage (iii) the derived SWCC information of degree of saturation and suction from the specimen at failure conditions (i.e. at critical state) from the modified ring shear tests of the specimen. The best predictions of the shear strength from the SWCC are obtained from the apparent SWCC at critical state. The study also shows that the relationship of the apparent degree of saturation and matric suction of a specimen sheared to critical state is unique and is not influenced by method of testing or compaction water content (i.e. soil structure). 1 INTRODUCTION practical perspective, the soil will be subjected to some loading and there will also be volume change both during the loading and the shearing stages (i.e. compression or dilation). Vanapalli et al. (2003) have reanalyzed Satija (1978) shear strength results on Dhanuari clay from triaxial test results to provide comparisons between the measured and the predicted shear strength values. The SWCC information was derived from the degree of saturation and the matric suction in the soil specimens from triaxial tests at failure conditions. The analyses of the results have shown that the predicted and measured peak strength values were in good agreement at lower matric suctions but showed significant differences at suction values greater than 200 kPa. These differences suggest that the conventional SWCC may not provide reasonable comparisons between the measured and peak strength values over a large suction range. In this paper, an attempt is made to predict the shear strength of unsaturated soils at critical state for a fine-grained soil using the effective shear strength parameters and the SWCC using the semi-empirical model proposed by Vanapalli et al. (1996) and Fredlund et al. (1996). Three different SWCC; namely (i) the conventional SWCC measured without any loading; (ii) the SWCC measured with an equivalent net normal stress on the specimen that was used during shearing stage (iii) the apparent SWCC (the derived SWCC information of degree of saturation and suction from the specimen at failure conditions) from the modified ring shear tests of the specimen.
The mechanical behaviour (i.e. shear strength and volume change) of unsaturated soils is relevant in the design of several geotechnical and geo-environmental structures, which are often in unsaturated conditions. The shear strength of an unsaturated soil can be determined using the modified direct shear or the triaxial shear equipment. However, experimental studies related to the determination of the shear strength of unsaturated soils are time consuming and require extensive laboratory facilities (Escario 1980, Escario & Juc 1989, Gan et al. 1988). Several semi-empirical procedures are therefore proposed to predict the shear strength of unsaturated soils (for example, Vanapalli et al. 1996, Fredlund et al. 1996, berg & Sllfors 1997, Khalili & Khabbaz 1998). The proposed prediction procedures use the effective shear strength parameters (c, ) along with the Soil-Water Characteristic Curve (SWCC) data to predict the peak shear strength of unsaturated soils. The underlying assumption of these methods is that the degree of saturation of the sheared specimens can be reasonably represented by the degree of saturation of the SWCC at an equivalent value of matric suction. The SWCCs are conventionally measured using Tempe cell or the pressure plate apparatus on specimen that are typically 50 mm in diameter and 20 mm in height. The conventional SWCCs are measured without taking into account the influence of loading or the resulting volume change on the SWCC itself. From a
253
DETAILS OF THE TEST PROCEDURES
The paper presents shear strength results obtained from a limited set of tests on specimens of an unsaturated glacial till from Indian Head, Saskatchewan, Canada using the modified ring shear apparatus, developed by Infante Sedano et al. (2007). The different ring shear tests that were conducted include: 1. one specimen compacted at a water content of 15% and sheared under Constant Load and constant Suction (CLS) conditions, 2. one specimen compacted at a water content of 15% and sheared under Constant Volume and constant Suction (CVS) conditions, 3. one specimen compacted at a water content of 19% and sheared under CLS conditions, 4. and one specimen compacted at a water content of 19% and sheared under CVS conditions. Each specimen was subjected to multi-stage shear tests, successively increasing the matric suction to higher stages (Fredlund & Rahardjo 1993). In other words, after completion of each shearing stage, the matric suction was increased to the next level allowing the specimen to reach equilibrium while the amount of water expelled from the specimen was continuously monitored, during all stages of loading and shearing. All the different shear strength tests were consolidated under a net normal stress value of 150 kPa. The compaction water contents of 15% and 19% correspond to the dry and wet side of the optimum of the glacial till compacted under a static stress of 375 kPa. The specimens were compacted in multiple layers within the ring shear testing cell shown in Figure 1. More details of the testing procedures are available in Infante Sedano et al. (2007). In addition to the shear strength tests, the conventional SWCC was measured following ASTM D6836 procedures using a pressure plate apparatus. More details of how the other two different SWCC were measured are detailed in later sections of the paper.
Table 1. Property Plasticity index, Ip (%) Liquid limit, wL (%) Specific gravity, Gs Void ratio, e (after compaction) Effective cohesion, c (kPa) Effective internal friction angle, (deg) Optimum moisture content (%) Properties of the tested soil. Value 15.5 32.5 2.72 0.530.70 10.61 26.3 14.4
Figure 1.
Cross-section of the modified ring shear cell.
2.1
Details of the shear strength tests
In Constant Load with Constant Suction (CLS) tests, both the net normal stress and the matric suction (i.e. the applied air pressure) are maintained constant. This test corresponds to the conventional consolidated drained (CD) tests. The specimen volume changes were monitored on a continuous basis so that the density and void ratio of the specimen could be determined accurately at any stage of the test. The water content of the specimen can change during the shearing process of this test. It is therefore necessary to collect reliable volume measurements to monitor the volume of water displaced to and from the specimen. In Constant Volume with Constant Suction (CVS) tests, the applied air pressure is maintained constant, but the normal load is modified on a continuous basis so that the volume of the specimen remains constant. Since the matric suction is maintained constant, the volume of water entering or leaving the specimen must be continuously monitored to establish the relationship between matric suction and water content of the specimen. As the volume of the specimen remains constant, the degree of saturation will be a direct function of the water content. The CVS tests conducted on the specimen prepared at a compaction water content of 15% were maintained at a constant volume between different shearing stages. This is equivalent to saying that the specimen was initially consolidated to a net normal stress of 150 kPa under zero matric suction (i.e. saturated conditions). After completion of the first stage of testing, the matric suction was increased to the next stage, and the specimen volume was maintained constant during the equilibration stage. Hence, during the final stage the volume of the specimen was the same as in the initial stage while the initial net normal stress was much lower.
254
In contrast, the CVS tests conducted on the specimen prepared at a compaction water content of 19% were consolidated under a net normal stress of 150 kPa between each of the shearing stages. The volume of the specimen at different stages was therefore different from its volume at earlier stages, although it was maintained constant during the shearing process. To explore the influence of shearing on the SWCC, additional specimens (designated, C, for compacted) were measured using conventional methods (i.e. using a pressure plate apparatus following the drying curve). One set of specimens was compacted at a water content wc 15%, and an another set was compacted at a water content, wc 19%. All specimens were saturated after compaction. A third set of specimens (designated NS, for Net Stress) were also compacted at a water content of wc 15%, and were then consolidated under a net normal stress of 150 kPa, for a duration of twenty four hours, after saturation. This procedure was followed to simulate the stress history, prior to shearing, of the specimens used in the modified ring shear apparatus. Figure 2 shows three SWCC determined on specimens compacted under a static stress of 375 kPa, at compaction water contents, wc, of 15% and 19%. Each data point on the curve represents the average degree of saturation of three specimens for each suction value. Figure 3 shows the SWCCs generated from the suction and degree of saturation data obtained experimentally from the modified ring shear apparatus on glacial till compacted at a compaction water content, wc, of 15% and 19%, and are plotted against the SWCC obtained by conventional means on the same soil on specimens prepared at compaction water contents of 15% and 19%. The shear strength envelope was generated from a single
Figure 3. Comparison of SWCC of specimens sheared under CLS, and CVS conditions to SWCC from pressure plate at compaction water content wc 15%, and wc 19%, Conventional (C) and consolidated at a Net Stress (NS) of 150 kPa.
Figure 2. Comparison of SWCCs measured different techniques [Conventional (C), Net Stress (NS) of 150 kPa].
specimen sheared at different matric suctions during multi-stage tests. The best-fit curve was computed through non-linear least squares using the equation presented by Fredlund & Xing (1994). As shown in Figure 3, the data from the modified ring shear tests cannot be fit with SWCCs that reach 100% saturation at a matric suction value of zero. In all cases, a maximum degree of saturation in the range of 85% to 94% was required. This behaviour is similar to the SWCC generated along a wetting curve when water follows preferential paths and air pockets become separated from the continuous air phase thus resulting in entrapped air bubbles that prevent the specimen reaching fully saturated conditions. Outside these pockets, the air phase and the water phase are continuous, and the degree of saturation is greater than the overall degree of saturation which is calculated taking into account the volume of air which is trapped in these pockets. When modeling the SWCC, it is common practice to use the apparent SWCC. The apparent SWCC is a curve which always shows a degree of saturation of 100% when the matric suction value is zero. The apparent SWCC is therefore representative of the SWCC as it exists in the soil outside the pockets of entrapped air. When the volume occupied by these air pockets is not taken in to account, the degree of saturation is higher, reaching close to 100% as the suction values drops to zero. To achieve this, the degree of saturation at any point on the curve may be linearly scaled by dividing it by the maximum degree of saturation at a suction value of zero using the best-fit curve, so that the curve shows full saturation (i.e. 100%) condition. The apparent SWCC curves of the sheared specimens are shown in Figure 4. The most notable feature is that all data of the sheared specimens lie along an approximately unique SWCC, whether
255
strength of unsaturated soils on a models parameters, the modified ring shear test results have been analyzed using the semi-empirical prediction equation proposed by Vanapalli et al. (1996) and Fredlund et al. (1996). This equation has been successfully used to describe the peak shear strength contribution of the same glacial till.
c ( n ua ) tan( ) (ua uw ) (S )(tan )
(1)
Figure 4. Apparent SWCC of specimens sheared under CLS, and CVS conditions, compared to SWCC from pressure plate at compaction water content wc 15%, and wc 19%, Conventional (C) and consolidated at a Net Stress (NS) of 150 kPa.
they were compacted dry or wet of optimum or sheared under constant load or constant volume conditions. Since the SWCC is closely related to the structure of the soil, this suggests that at critical state, a unique structure can be achieved in a given soil irrespective of the mode of failure. Any history-specific (i.e. compaction water content, initial stresses, volume changes) structure of the soil is erased if shear deformations are large enough, and a unique SWCC can be achieved. It should therefore be possible to associate a unique degree of saturation, or even gravimetric water content to a specific value of suction at critical state for a particular soil. Specifically, a given degree of saturation or gravimetric water content could be associated to a unique shear strength value at critical state. Such a behavior cannot be expected for the peak strength as the soil structure would be expected to result in various SWCC depending on the compaction and stress history of the soil (Vanapalli et al. 1999). Another feature of the curve shown in Figure 4 is that it shows a higher degree of saturation for any given matric suction than even the SWCC measured by conventional means on a consolidated specimen. This results in a higher air-entry value for the sheared specimen than the SWCC specimens which implies that models using the degree of saturation to determine the fraction of the matric suction contributing the shear strength of unsaturated soils need to compensated for an unknown difference between the degree of saturation of a sheared specimen and the one determined from the conventional SWCC tests. 3 ANALYSIS OF RESULTS
where: c effective cohesion internal effective friction angle n total normal stress ua pore air pressure uw pore water pressure S degree of saturation a fitting parameter For the purpose of analysis, comparisons will be provided between the measured and predicted values of the component of shear strength contribution due to matric suction. Only the second part of Equation (1), shown in Equation (2), is required for predicting the shear strength contribution due to matric suction:
(ua uw ) (S )(tan
(2)
To illustrate the influence of both the selection of the SWCC and the use of critical state shear
Using the degree of saturation measured when critical state was reached in the stress-strain curves, the value of could be back-calculated for the various experimental results. It was found that a value of at critical state, c 5.0 best-fits the shear strength behaviour of the tested glacial till at critical state. This compares to a value of 2.2 for the same soil under peak strength condition obtained from the versus plasticity index, Ip relationship presented by Vanapalli & Fredlund (2000). This higher value of c implies that the term S will be lower at lower degrees of saturation (and higher matric suctions) than for lower values of . This causes the curve to flatten more at higher matric suctions than using a lower valued parameter such as the one obtained for peak strength. Figure 5 shows the shear strength predictions resulting from using the best-fit SWCC obtained using a pressure-plate apparatus on NS specimens (i.e. consolidated to the net normal stress of the sheared specimens), as well as on conventionally tested specimens, C (i.e. that were not subjected to consolidation). Figure 5 shows that the predictions using c 5.0 are better correlated to the experimental data if the SWCC from NS specimens is used rather than the SWCC obtained from the C specimens. The prediction using the SWCC generated using the C specimens improves if a value of c 2.2 (which is associated to the peak unsaturated shear
256
Figure 5. Measured and predicted shear strength envelope for a specimen prepared at wc 15%, and sheared in CLS conditions.
Figure 6. Measured and predicted shear strength envelope for a specimen prepared at wc 15%, and sheared in CLS conditions.
strength of glacial till) is used instead of c 5.0 which is back-calculated from critical state shear test results (see Figure 6). However, at matric suctions greater 500 kPa, the curve shows greater divergence from the almost horizontal trend shown by the experimental data and from the prediction based on NS specimens and c 5.0. Figure 7 compares the prediction using the SWCC obtained from the pressure plate apparatus using consolidated specimens to the predictions using the SWCC obtained from the sheared specimens. It is apparent from the curves that the SWCC generated using directly the degree of saturation of the sheared specimens does not offer as good a fit, with an overall shape that resembles the prediction curve generated with the pressure-plate on the C specimens. In both cases the curve does not seem to reach a plateau within the range of matric suctions used in the experimental program, while the experimental measurements of the shear strength clearly do. The best fit to the data comes from the apparent SWCC of the sheared test specimens, using a value of c 7.25 obtained by fitting the apparent degree of saturation of the sheared specimens using Equation (2). This suggests that the apparent SWCC provides a better measure of the degree of saturation of a sheared specimen than the SWCC obtained from direct measurements of the degree of saturation of the specimens, since it is independent of the compaction water content and mode of failure (Figure 4). The difference between these predicted values and those generated using the SWCC obtained in the pressure plate apparatus on NS specimens, using a value of c 5.0, are minimal within the range of matric suctions of the experimental program. Since the SWCC generated from NS specimens (at wc 15%) lies relatively closer to the apparent SWCC than those generated from
Figure 7. Measured and predicted shear strength envelope of glacial till using conventional SWCC and modified ring shear test at compaction water content, wc 15%.
C specimens, it should be expected to provide better predictions, than the SWCCs from C specimens, even for sheared specimens that were compacted at wc values other than 15%. This is indeed the case as it can be seen in Figure 8, where the SWCC from NS specimens compacted at a water content of wc 15% yield a better prediction curve than the conventional SWCC obtained using a compaction water content of wc 19% which was the same as the sheared specimens. Even when using a value of c 2.2 (Figure 9), which had yielded better results for the specimen compacted at wc 15%, the SWCC from the NS specimens at wc 15% still provide a much closer fit to the experimental data than the conventional SWCC generated from wc 19% samples. This comparison based on a single set of consolidated specimens cannot be generalized to other soils or other stress histories without additional testing.
257
This is of particular importance for validation of unsaturated soils shear strength models that use the degree of saturation of a soil specimen to predict its behaviour. The results imply that directly using the degree of saturation from shear tests to calibrate such models may not be reliable, and hence a representative SWCC should be carefully chosen for the tested soil.
REFERENCES
Figure 8. Measured and predicted shear strength envelope of glacial till using the conventional SWCC and modified ring shear test results, at compaction water content, wc 19%. Escario, V. 1980. Suction controlled penetration and shear test. In Proceedings of the 4th International Conference on Expansive Soils, Denver, Vol. 2: 781797. Escario, V. & Juca 1989. Shear strength and deformation of partly saturated soils. In Proceedings of the 12th International Conference on Soil Mechanics and Foundation Engineering, Rio de Janeiro, Vol. 2: 4346. Fredlund, D.G. & Harianto Rahardjo. 1993. Soil mechanics for unsaturated soils. Wiley-IEEE. Fredlund D.G. & Xing, A. 1994. Equations for the soilwater characteristic curve. Canadian Geotechnical Journal 31(3): 521532. Fredlund, D.G, Xing, A., Fredlund, M.D. & Barbour, S.L. 1996. Relationship of the unsaturated soil shear strength to the soil-water characteristic curve. Canadian Geotechnical Journal 33(3): 440448. Gan, J.K.M., Fredlund, D.G. 1988. Multi-stage direct shear testing of unsaturated soils. Geotechnical Testing Journal 11(2): 132138. Infante Sedano, J.A., Vanapalli, S.K. & Garga, V.K. 2007. Modified ring shear apparatus for unsaturated soils testing. Geotechnical Testing Journal 30(1): 3947. Khalili, N. & Khabbaz, M.H. 1998. A unique relationship for the determination of the shear strength of unsaturated soils. Gotechnique 48(5): 681687. berg, A.L. & Sllfors, G. 199. Determination of shear strength parameters of unsaturated silts and sands based on the water retention curve. Geotechnical Testing Journal 20(1): 4048. Satija, B.S. 1978. Shear behaviour of partly saturated soils. Ph.D. Thesis, Indian Inst. of Tech., Delhi, India. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction, Canadian Geotechnical Journal, 33(3): 379392. Vanapalli, S.K., Fredlund, D.G. & Pufahl, D.E. 1999. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Gotechnique 49(2): 143159. Vanapalli, S.K. & Fredlund, D.G 2000. Comparison of different procedures to predict unsaturated soil shear strength, In Shackelford CD et al (eds), Advances in Unsaturated Geotechnics:195209. Geotechnical Specialty Publication No. 99. ASCE: New York. Vanapalli, S.K., Garga, V.K. & Infante Sedano, J.A. 2003. A rigorous approach of predicting the shear strength of unsaturated soils. In D. Karube, A. Iizuka, S. Kato, K. Kawai & K. Tateyama (eds), Proceedings of the 2nd Asian Conference on Unsaturated soils: 253258, Osaka 1012 November 2003.
Figure 9. Measured and predicted shear strength envelope of glacial till. The shear strength is predicted using the conventional SWCC and 2.2.
SUMMARY
In this study, four multi-stage axis-translation ring shear tests were conducted. From observations of the available test results, the following conclusions are derived: 1. The choice of the SWCC will have a marked influence on the prediction of the critical state shear strength of unsaturated soils. 2. The SWCC obtained by conventional means can be substantially different from the SWCC obtained from the characteristics of a sheared specimen. In particular, the maximum degree of saturation of the SWCC generated from shear data is lower than 100%, suggesting that entrapment of air occurs during the shearing process. 3. Subjecting the conventional SWCC specimens to a stress history comparable to the stress history of the test specimens before the shearing process can markedly improve the critical state shear strength prediction. 4. The apparent SWCC obtained from sheared specimens provide a better fit to the shear strength of the glacial till at large deformations, both at low and at (relatively) higher matric suctions.
258
Unsaturated shear strength parameters of tropical volcanic residual soils

R. Karlinasari P.P. Rahardjo
Sultan Agung Islamic University, Semarang, Indonesia Parahyangan Catholic University, Bandung, Indonesia
ABSTRACT: The uppermost zone of tropical residual soil is a zone of unsaturated soil with a unique characteristic. The uppermost zone intensively change from unsaturated-saturated phase, because the intensity of sun heat is high and the intensity of rainfall is much higher than its soil permeability. A series of research on the unsaturated shear strength of tropical volcanic residual soil were conducted with Triaxial Consolidated Drained Unsaturated (TXCD-UNSAT) Method. The research is resulted on unsaturated shear strength parameters. These parameters were evaluated with other researchs results on unsaturated shear strength. An analysis based on the two methods of constitutive behaviour of unsaturated shear strength: The Two Independent Stress Variable Method and Effective Stress Method was conducted and the results confirm the published data. Keywords: Unsaturated, Shear Strength Parameters, TXCD-UNSAT, Independent Stress Variable Method, Effective Stress Method 1 INTRODUCTION 2 UNSATURATED SHEAR STRENGTH TEST (TXCD-UNSAT)
This research is a part of a study on the characteristic of tropical volcanic residual soil. To determine the shear strength of the soil a series of triaxial tests were conducted. The triaxial test were Triaxial Consolidated Undrained Test (TXCU) to determine the saturated shear strength parameters and Triaxial Consolidated Drained Unsaturated (TXCD-UNSAT) to determine the unsaturated shear strength parameters. In attempt to measure unsaturated soil parameters, an extensive test is conducted using Modified Consolidation Drained Triaxial Test with sophisticated equipment. A simplified modified Consolidated Drained Triaxial Test were developed at the Geotechnical Laboratory of Parahyangan Catholic University Bandung. A high air entry value porous disc were installed to the modified Triaxial Cell equipment. An addition procedure of matrix suction equalisation were conducted after consolidation stage, resumed with sample compression under constant matrix suction. The results were analyzed based on the two methods of constitutive behaviour of unsaturated shear strength: The Two Independent Stress Variable Method and Effective Stress Method.
The Triaxial Consolidated Drained Test (TXCD) with constant matrix suction is a triaxial test in one series with four stages. The stages are saturation, consolidation, matrix suction equalisation and compression. The triaxial equipment consist of: 1. A triaxial cell with a modified pedestal to conduct unsaturated shear strength test with constant suction. 2. Cell pressure application system. 3. Back pressure application system. 4. Air pressure application system. 5. Pore water pressure measurement system. 6. Volume change measurement system. The diagram of the system is shown in Figure 1 and the pedestal of the modified triaxial test is shown in Figure 2. This test procedure was conducted in a lengthy period of time. One sequel of test need three to four weeks. Table 1 listed the stages, the stress and the matrix suction application. The application of back pressure on sample was conducted by water pressure application from top
259
Table 1. The example of pressure values applied at every stages of TXCD unsaturated test. ua uw ua ua uw
STAGE 1 Consolidation 1.8 0.8 1 Matrix suction equalisation 1.8 0.8 0.4 1 Shearing 1.8 0.8 0.4 1 STAGE 2 Consolidation 2.6 1.2 1.4 Matrix suction equalisation 2.6 1.2 0.4 1.4 Shearing 2.6 1.2 0.4 1.4 Figure 1. soils. Modified triaxial cell to test unsaturated STAGE 3 Consolidation 3 Matrix suction equalisation 3 Shearing 3 1.4 1.6 1.4 0.4 1.6 1.4 0.4 1.6
0.4 0.4
0.8 0.8
1 1
0
Water volume change, Vw (cm3)
-5 -10 -15 -20 -25

0 5
BH02 Cijengkol 0.5-1.0 m (cm )

ua-uw = 0.29 kg/cm2 ua-uw = 0.68 kg/cm2 ua-uw = 0.89 kg/cm2
3
10
15
20
25
Elapsed time, t (hours)

Total volume change, Vt (cm3) 0 -2 -4 -6 -8 0 5 10 15 20 Elapsed time, t (hours) 25 BH02 Cijengkol 0.5-1.0 m
ua-uw = 0.29 kg/cm2 ua-uw = 0.68 kg/cm2 ua-uw = 0.89 kg/cm2
-10
Figure 3. Water volume change and total volume change at matrix suction equalisation stage.
Figure 2. The water compartment at the bottom of ceramic disc triaxial cell pedestal.
and from the bottom of the sample. On saturation and consolidation stages the back pressure were applied via the top cap, and on matrix suction equalisation stage via the water compartment at the bottom of the sample. The application of the air pressure on sample was started to be conducted at matrix suction equalisation stage via the top cap, and will continue to the compression stage.
At the beginning, the sample is saturated, then it is consolidated with cell pressure of 1.8 kg/cm2 and back pressure of 0.8 kg/cm2. After dissipation of pore water pressure achieved, matrix suction was applied. The matrix suction was modelled as the soil condition on nature, where the top of the soil is dry and the bottom of the soil is wet. The moisture tension then happen between the top and the bottom of the sample. The compartment at the bottom of the sample is functioned as the water table in nature. The sample is separated from the water by high air entry value disc. This lay out makes the air cannot escaped from the bottom of the sample, meanwhile the water will be attracted to the top of the sample by capillarity. The back pressure were kept at 0.4 kg/cm2, air pressure from air compressor were applied at the top of the sample at 0.8 kg/cm2,
260
so a matrix suction of 0.4 kg/cm2 was attained. After the water volume change was stabilized, then the compressions were applied (axial compression application). The following figure (Figure 3) shows the water volume change curves and the total volume change curves at matrix suction equaliasation stage. It is interesting that on the second stage curve, a drop of volume change occurred. This phenomenon may be related to the macropore of the soils as mention by Feurharmel, Gehling & Bica (2006).
Shear stress, (kg/cm2)
0.60 0.50 0.40 0.30 0.20 0.10 0.00 0 0.1 0.2
UNSATURATED SHEAR STRENGTH PARAMETERS ANALYSIS
The analysis of the TXCD-Unsat test results were conducted by two methods: The Two Independent Stress Variable Method after Fredlund & Rahardjo, 1993 and Unsaturated Effective Stress Method after Khalili & Khabbaz (1998). 3.1 TXCD-UNSAT results in the two Independent Stress Variable Method
AEV b = 9.51 b = 26.69 BH 02 Neglajaya 0.5 - 1.0 m = 26.5 AEV = 0.30 kg/cm2
0.3 0.4 0.5 0.6 0.7 0.8
Matric suction (ua-uw) (kg/cm2)
Figure 4. The failure envelope of versus (uauw) plane of BH02 Neglajaya 0.51.0 m.
The Independent Stress Variable Method is a method that separate the effect of the nett normal stress and the effect of matrix suction on soils shear strength in a diffeent axis of stresses. The results of Unsaturated Triaxial CD test of BH02 Neglajaya 0.51.0 m on 3D Mohr-Coulomb diagram was shown in diagram of Figure 5. On Figure 4, it is shown the increase of soils shear strength in the raise of matrix suction. It is shown on the diagram that at BH02 Neglajaya 0.51.0 m results, the magnitude of b almost the same as the magnitude of angle, at 0.09 kg/cm2 to 0.29 kg/cm2 matrix suctions (or below the air entry value (AEV) of the soil at 0.3 kg/cm2).
Figure 5.
Mohr Coulomb failure envelop for unsaturated soil samples of BH02 Neglajaya 0.51.0 m.
261
After the AEV the b decrease almost a quater of the b at 0.29 kg/cm2 to 0.69 kg/cm2 matrix suctions. 3.2 TXCD-UNSAT results in Unsaturated Effective Method
The Unsaturated Effective Stress Method analysis started by the determination of efective stress parameters (). Khalili & Khabbaz (1998) in Khalili, Geiser & Blight (2004), collected some formerly research results and determined the relationship between effective stress parameters () and suction ratio as shown in Figure 6 below. Khalili concludes that the effective stress parameter () is a parameter strongly related to soils structures. The suction ratio is the magnitude of matrix suction over suction value at the time that air start to enter the pore in soil structure (sae, suction Air Entry Value (AEV). The relation on Figure 6 has a top limit of y 1.045 (s/sAEV)0.46 and fit bottom limit of y 1.008 (s/sAEV)0.68. Khalili & Khabbaz take the relationship and then define the relationship on drying process as follow: s 0.55 sae 1 if s sae if s Sae
Figure 7. The relation of matrix suction and deviatoric stress on effective method analysis for BH 02 Neglajaya 0.51.0 m sample.
stress q and matrix suction may be defined. The prediction was defined by calculated deviatoric stress q relations as follow: q a ( p net s )M (6 sin ) M (3 sin ) p net 3 ua Figure 7 shows the prediction and the actual deviatoric stress over suction for BH02 Neglajaya 0.51.0 m soil. For the 0.09 kg/cm2 and 0.29 kg/cm2 matrix suctions, the actual results fit the prediction line, meanwhile the actual results of 0.69 kg/cm2 matrix suction is higher than the prediction line. 4 THE COMPARATION OF THE TXCD-UNSAT RESULTS
(1)
(2)
The results of BH02 Neglajaya 0.51.0 m samples, then expressed in failure envelope plane of qp and the diagram shown at Figure 7 below. The diagram shown that the maximum point of unsaturated TXCD stress path from 0.09 kg/cm2 to 0.69 kg/cm2 matrix suction lay on the top of state line for saturated shear strength of M 1.048. By taking a and of saturated soil samples, a prediction of the relationship between deviatoric
Figure 6. Effective stress parameters () versus suction ratio (Khalili & Khabbaz (1998)).
The shear strength of tropical volcanic residual soil based on the TXCD-Unsat in this research have a range of value as follow: a. Based on the Two Independent Stress-State Variable Method the results was resumed in Table 2 below. The maximum value of b is approaching the value of . The magnitude of b drastically decrease on the further increase of suction value. The same results was obtained by Rahardjos research on residual soil of Singapore (Rahardjo, et al, 2004), the resume of the results is shown on Table 2. b. Based on effective parameters of (s/sAEV)0.55 all the results were resumed in Figure 9. As general the point of maximum stress path is lay
262
Figure 8. The failure envelope of qp for a saturated soil and the maximum stress path point of stress from unsaturated soil of BH02 Neglajaya 0.51.0 m. Table 2. Shear strength parameter of Singapore residual soils (Rahardjo, et al, 2004). Granitic Bukit Timah Formation Depth (m) 59 1015 1521 c (kg/cm2) 0.26 0.13 0.12 () 27 35 38 b () 29 to 8 35 to 6 35 to 4 Table 3. Shear strength parameters from TXCDUNSAT tests. Matriks suction (ua uw) (kg/cm2) BH02 Cij 0.51.0 m 0.29 0.68 0.89 BH01 Cil 4.55.0 m 0.40 0.80 1.00 BH02 Neg 0.51.0 m 0.09 0.29 0.69 c (kg/cm2) 0.23 0.35 0.40 0.66 0.68 0.71 0.38 0.48 0.55 () 22 22 22 32 32 32 26.5 26.5 26.5 33 33 33 b () 16.9 14.4
4.88 4.88
26.69 9.51
BH02 Neg 13.013.5 m 0.40 0.93 0.60 1.06 0.90 1.09
33.26 4.545
5
Figure 9. The evolution of critical state line over suction on pq plane, kaolin clay soil (Wheller Sivakumar, 1995) on Khalili, et al, 2004.
THE APPLICATION OF THE UNSATURATED SHEAR STRENGTH ON TROPICAL VOLCANIC RESIDUAL SOILS CONDITION
above the saturated state line. The distance of the points from the saturated state line is indicating a function of the magnitude of suction. There is a specific shapes of the dot line compared to the previous research collected by Khalili at Figure 9.
The uppermost zone of tropical residual soil is a zone of unsaturated soil with a unique characteristic. The uppermost zone intensively change from unsaturated-saturated phase, because the intensity of sun heat is high and the intensity of rainfall is much higher than its soil permeability. From this nature of the uppermost zone which is intensively change from unsaturated-saturated phase, it is
263
Figure 10. Failure envelope on qp plane for saturated soil and the maximum point of stress path for unsaturated soils in this research.
important to have a method that could model this behaviour. Figure 10 shows the failure envelope of qp for a saturated soil samples and the maximum stress path point of stress from unsaturated soil samples on this research, it shows that the maximum stress path point of unsaturated sample lay in some extent above the failure envelope of saturated soil samples. This kind of curve will be very applicable to model the intensively change from unsaturatedsaturated phase on tropical residual soil. 6 CONCLUSION
REFERENCES
Karlinasari, R., 2009, Karakteristik Tanah Residual Tropik Vulkanik Formasi Gunung Api Tua (Qob) Jawa Barat, Disertasi pada Program Doktor Teknik Sipil Universitas Katolik Parahyangan, Bandung. Khalili, N. and Khabbaz, M.H., 1998, A unique relationship for for shear strength determination of unsaturated soils, Geotechnique, 48(5), 681688. Khalili, N., Geiser, F. & Blight, G.E., 2004, Effective stress in unsaturated soils: A critical review with new evidence, International Journal of Geomechanics. ASCE; 4(2), 115126. Melani, I., 2004, Shear Strength Characteristics of A Compacted Kaolin Under Infiltration Conditions, Disertation, School of Civil and Environmental Engineering, Nanyang Technological University. Rahardjo, H., Aung, K.K., Leong, E.C. & Rezaur, R.B., 2004, Characteristic of Residual Soils in Singapore as formed by weathering, Engineering Geology, 73, 157169.
The shear strength parameters obtained from the simplified TXCD-Unsat have been compared to the previously result both from the Two Independent Stress State Method and the Unsaturated Effective Method, and the results confirm the published data.
264
Direct shear test under constant volume state for unsaturated soil
S. Kato, B.S. Kim & S. Shibuya
Graduate School of Engineering, Kobe University, Kobe, Japan
ABSTRACT: This paper examined on the shear behavior of unsaturated soil in Direct Shear Test (DST) under constant volume condition by applying the suction stress obtained from SSM (Suction stress SWRC Method). An opening was set between the upper and the lower shear boxes before the shear process. The relationship between the shear behaviors for saturated and unsaturated soils can be explained by applying the suction stress. In addition, the principal stresses in DST were defined through applying several hypotheses, and the stress paths were analyzed with Cam clay type stress parameters. For the normalized stress plane with the equivalent stress ( pe), in case of saturated soil, the stress path moved along the Hvorslev surface. In case of unsaturated soil, the stress path agreed with the maximum volumetric compression line. Therefore, the suction stress has an important meaning to understand the stress path behaviors under constant volume condition. 1 INTRODUCTION soils in DST under constant pressure condition. They reported that when the suction stress is taken into account with SSM, the shear strength and deformation characteristics could be explained uniquely for saturated and unsaturated soils. In addition, Kim et al. (2010b) derived Cam clay type stress parameters on the basis of several hypotheses, and reported the existence of the critical state in DST under constant pressure condition. The main points of their hypotheses are; i) the Mohr stress circle contacts to the datum line that connects the stress point for the shear plane and the intercept, which equals to the suction stress, for horizontal axis, ii) the parameter b ( (2 3)/ (1 3)) equals to zero, which is similar to the stress state of TC. In this study, in order to examine the shear strength and deformation for saturated and unsaturated soils under constant volume condition, a series of DST were carried out. The stress paths from test results were studied through the application of the suction stress obtained from SSM. Moreover, Cam clay type stress parameters were derived through the application of the same several hypotheses like Kim et al. (2010b). The suction stress obtained from SSM was applied, and then these results were studied in ( p* net ps , q) space. Lastly, the stress paths normalized with the equivalent stress ( pe) were also examined for the state boundary surface. 2 SAMPLE AND TEST PROCEDURE
The theory for shear strength of unsaturated soil has not explained fully results of direct shear test (DST). One of the main reasons is derived from the shortcoming in DST apparatus itself, because the principal stress directions rotate during the shear process. Although there are also other problems, many studies have been made with the direct shear apparatus owing to its advantages (e.g., the drainage distance is shorter than the triaxial test; time saving, simple handling). DST has been carried out under two kinds of test condition, i) the constant pressure condition, ii) the constant volume condition. In the former condition, the vertical stress for the specimen is kept constant during the shear process. In the later condition, the volume of specimen is kept constant during the shear process. Under the constant pressure condition, a series of studies for saturated soil have been carried out so far to revise the mechanism of apparatus and the test procedure for DST (e.g., Shibuya et al., 1997). Some studies for unsaturated soil have tried to explain the shear strengths of DST applying the shear strength theory based on the triaxial test results (Donald 1956; Escario 1980; Escario and Saez 1986; Satija 1978; Gan et al. 1988). However, these studies has not explained sufficiently the results of DST because these studies were just analyzed as the relationship between the vertical stress and shear stress for the shear plane, and Cam clay type stress parameters have not been tried yet. Kim et al. (2010a) proposed SSM (Suction stress-SWRC Method) to study the shear strength and deformation for saturated and unsaturated
The direct shear test apparatus used in this study was a modified type for unsaturated soils as shown
265
Figure 2(a) shows the stress paths for saturated soil. In this figure, a solid line and a dotted line represent the failure line (37.5 degrees) and the maximum volumetric compression point (M.C.P) line (36.06 degrees) obtained under constant pressure condition respectively. It was observed that the stress paths of obtained test results reach to the failure line under constant pressure condition. Figure 2(b) shows the stress-displacement relationships. Although the peak shear strength increases with the increase of vertical stress, each slope (the angle of internal friction) is almost same.
Figure 1. The modified direct shear test apparatus.
60
v = 10kPa
in Figure 1. All data is recorded on the automatic measurement system and the direct shear box was placed inside an air pressure chamber. The pore air pressure was applied to a specimen through a coarse porous-metal of the loading plate. Both a ceramic disk (500 kPa; A.E.V) and a coarse porous-metal of the semicircle shape were installed in the lower part of the shear box. The weathered granite soil that distributes widely inside the western part of Japan was used. The soil particle density of sample is s 2.652 g/cm3, and the liquid limit is NP. The sample that passed through a 2.0 mm opening sieve was used. Specimen of 60 mm in diameter and 20 mm in height was used and has the dry density of about 88% for the maximum dry density (1.931 g/cm3). When setting up the specimen, the lower and the upper shear boxes was sealed with 28 sheets of Teflon (0.1 mm per one sheet) which has an extremely low friction resistance. Accordingly, during the shear process, the Teflon sheets keep up the specimen side without loss of sample from the opening. After the saturation of specimen, the consolidation process was carried out under the specified overburden pressure. In the case of unsaturated soil, air pressure (ua 50 kPa) was additionally applied inside the chamber after the consolidation process. Then, the shear process was started with keeping the volume of specimen constant. The void ratio was obtained by measuring the vertical displacement, and then the degree of saturation and the volumetric water content was derived from the displacement of drained water volume. The horizontal displacement rates used for saturated and unsaturated conditions are 3.33 103 mm/s and 3.33 104 mm/s respectively, which are based on the past studies (Fredlund & Rahardjo, 1993). 3 STRESS PATH AND STRESS DISPLACEMENT RELATIONSHIPS
Saturated soil F.L : ' = 37.5
50
v = 20kPa v = 30kPa(1) v = 30kPa(2) v = 40kPa(1) v = 40kPa(2) v = 50kPa(1) v = 50kPa(2)
Shear stress, (kPa)
40 30 20 10 0
M.C.P line ' = 36.06
10
Vertical stress, v (kPa) (a) v ~
20
30
40
50
60
40
Shear stress, (kPa)
30
Saturated soil Sr = 100%

v = 50kPa(1) v = 40kPa(1)
v = 50kPa(2) v = 40kPa(2) v = 30kPa(2) v = 30kPa(1) v = 20kPa v = 10kPa
20
10
0 60 50
Vertical stress, v (kPa)
40 30 20 10 0
v = 50kPa(1) v = 40kPa(1)
v=50kPa(2) v = 40kPa(2) v = 30kPa(2) v = 30kPa(1)
v = 20kPa v = 10kPa
The effective stress paths for saturated soil under constant volume condition generally reach to the failure line under constant pressure condition.
Horizontal displacement, dh (mm) (b) dh~ v ,
10
Figure 2. DST results under constant volume for saturated soil.
266
Figure 3(a) shows the test results for unsaturated soil. The stress paths move over to the failure line and show larger shear strengths than the failure line for saturated soil. The stress-displacement relationships increase more abruptly than those for saturated soil, and more dilative behaviors are observed in shown Figure 3(b). When the stress paths reached around M.C.P line, the vertical stresses in these results increased after the dilative behavior commonly.
APPLICATIION OF SUCTION STRESS TO STRESS PATHS
80
ua = 50kPa
F.L.: ' = 37.5
Shear stress, (kPa)
60
v = 10kPa v = 20kPa(1) v = 20kPa(2) v = 40kPa(1)
40
v = 40kPa(2)
M.C.P line ' = 36.06

20
The stress state in direct shear test is usually expressed in (, ) space. On the other hand, Fredlund et al. (1978) proposed the failure envelope line for saturated and unsaturated soils in three-dimensional space (Fig. 4). They examined the parameter b which represents the increase of shear strength with the matric suction. Kim et al. (2010a) proposed SSM (Suction stressSWRC Method) to apply the suction stress proposed by Kaube et al. (1996) to DST. SSM is a methodology to estimate the matric suction on the failure plane based on the volumetric water content. In DST, the water content is measured through the drained water volume from the specimen, and the water content can be determined by the matric suction value using SWRC. Then, the suction stress is calculated with the obtained matric suction value and the measured degree of saturation in DST using eq. [1]. They examined the shear behaviors under constant pressure condition in DST. Here, the suction stress is given with the matric suction (s*) which is obtained with SSM as follows (Fig. 4). ps Sr Sr 0 s* 100 Sr 0 (1)
Vertical stress, v (kPa) (a) v ~

v = 10kPa
20
40
60
80
80
v = 40kPa(2)
60
v = 40kPa(1)
40
v = 20kPa(1)
v = 20kPa(2)
20
Unsaturated soil ua = 50kPa

100
v = 40kPa(2)
where Sr the degree of saturation, Sr0 the residual degree of saturation. They reported that as applying the suction stress obtained from SSM, the failure stress states for unsaturated soil corresponds to the failure line for saturated soil. This result implies that the shear strength relationship between saturated and unsaturated soils could be explained through the application of proposed SSM. Accordingly, in this study, SSM was applied to DST results under constant volume condition.
a)
Shear stress, (kPa)
Vertical stress, v (kPa)
80
v = 40kPa(1)
Shear stress, T
b
k(s)
M a c tri
su
io ct
( n,
ua
60
v = 10kPa
'
40
v = 20kPa(2) v = 20kPa(1)
'
s*
O
20
Horizontal displacement, dh (mm) (b) dh v ~,
ps
Net normal stress, (ua)
Figure 3.
Unsaturated soil (ua 50 kPa).
Figure 4. The failure envelope line for saturated and unsaturated soil in 3-dimensional space.
267
80
ua = 50kPa
60
v = 10kPa v = 20kPa(1) v = 20kPa(2) v = 40kPa(1)
F.L : ' = 37.5
40
v = 40kPa(2)
M.C.P line ' = 36.06

20
numbers of Mohrs stress circle. However, if the Mohrs stress circle is defined by one condition, the principal stresses in DST can be decided. Figure 7 represents the concept of the Mohrs stress circles for saturated and unsaturated soils by using several hypotheses proposed by Kim et al., (2010b). They derived Cam clay type stress parameters from the results under constant pressure condition by several hypotheses, and represented the stress paths in ( p* net, q) space. They also reported that M.C.P line corresponds to the critical state line (CSL) in ( p* net, q) space. This CSL is given as follows.
80
(kPa)
20
v + ps (kPa)
40
60
q M pnet 1.466 pnet
(2)
Figure 5. Stress paths for unsaturated soil by applying the suction stress obtained from SSM.
Stress path : Unsaturated state
M.C.P line : Saturated state

k
Figure 8 shows the stress paths for saturated soil under constant volume condition in ( p* net, q) space. In this figure, the solid line shows M.C.P line ( critical state line) under constant pressure condition (see Kim et al. (2010b)). The stress paths of test results reach and move over CSL. It was found that these results correspond to the behavior of stress path under constant pressure condition. Figures 9(a) shows similar results for unsaturated soil. The stress paths reached and move
(kPa) Shear stress,
ps (Suction stress)
+ps
v
ds
ff
Figure 6. Relationship between suction stress circle in and M.C.P line in stress path for unsaturated soil.
Unsaturated condition
C B
c Saturated condition
ff-unsat
Figure 5 shows the stress paths for unsaturated soil after the suction stress obtained from SSM was applied. It was found that the stress paths of one condition (ua 50 kPa) reached to M.C.P line of the constant pressure condition with the progression of shear process. These behaviors can be summarized as shown in Figure 6 graphically. In this figure, the suction stress is given as the intercept of v axis for the tangent line to the stress path, and the slope of this tangent line equals to that for M.C.P line. From this figure, it is understood that if the suction stress obtained from SSM is treated as a stress component, the stress path corresponds to M.C.P line of the constant pressure condition. After all, such a tendency means that the difference of the stress paths between saturated and unsaturated soils is affected by the suction stress. 4 CRITICAL STATE BEHAVIOR IN (P, Q) SPACE
ds
ff-sat
O'
ps
c=0
3f-unsat
3f
A(
1i)
M'
Net normal stress,
1f
1f-unsat
ua(kPa)
Figure 7.
Conception of Mohrs stress circles in DST.

100 Saturated state (CU)
v=10kPa
M(M.C.P line)
80
v=20kPa v=30kPa(1) v=30kPa(2) v=40kPa(1)
60
v=40kPa(2) v=50kPa(1) v=50kPa(2)
40
i rit
ca
ls
e at
lin
20
10
20
30
* Pnet
40
50
60
The stress state for an infinitesimal element is represented with a Mohrs stress circle. Usually, the stress state in DST can be represented as infinite
Figure 8. space.
Stress paths for saturated soil in (p* net, q)
268
200 Unsaturated state (ua=50kPa ; CU)

v=10kPa v=20kPa(1)
2.0 Saturated state (CU)

v=10kPa v=20kPa
150
v=20kPa(2) v=40kPa(1) v=40kPa(2)
1.5
=30kPa(1) v v =30kPa(2) v v =40kPa(1) v v
q / pe'
M(M.C.P line)
v=40kPa(2)
M(M.C.P line) (M.C.P line)
1.0
=50kPa(1) v v v=50kPa(2)
100
H
0.5
e fac sur lev s r vo
50
0.0 0.0
0.2
0.4
0.6
0.8
p*net / pe'
0 20 40
* pnet
60
80
100
120
* , q) space (a) (pnet
200 Unsaturated state (ua=50kPa ; CU)

v=10kPa
v=20kPa(1)
Figure 10. The normalized stress paths for saturated soil in ( p* , q/pe) space. net, /pe
3.0 Unsaturated state (ua=50kPa ; CU) 2.5
v=10kPa v=20kPa(1) v=20kPa(2) v=40kPa(1) v=40kPa(2)
M(M.C.P line)
150
v=20kPa(2) v=40kPa(1) v=40kPa(2)
2.0
M(M.C.P line)
100
q / pe'
1.5
1.0
50
0.5
20
40
60
* + p pnet s
80
100
120
0.0 0.0
0.5
1.0
1.5
2.0
p*net / pe'
* + p , q) space (b) (pnet s
(a) (p*net/pe, q/pe) space

3.0 Unsaturated state (ua=50kPa ; CU) 2.5
v=10kPa v=20kPa(1) v=20kPa(2)
Figure 9. Stress paths for unsaturated soil (ua 50 kPa).
over CSL. If the suction stress is applied to these results, it was found that the stress paths agree with CSL as shown in Fig. 9(b). This tendency is different from that observed under constant pressure condition. From this result, it could be understood that the difference in deformation mode during the shear process, that is, the difference of a dilatancy between two testing conditions affects these behaviors. Therefore, it can be concluded from these test results that CSL can be observed with applying the suction stress in case of the unsaturated soil under constant volume condition. 5 ANALYSIS FOR NORMALIZED STRESS PATHS
2.0
v=40kPa(1) v=40kPa(2)
P C. M.
lin
q / pe'
1.5
1.0
0.5
0.0 0.0
0.2
0.4
0.6
0.8
1.0
1.2
1.4
(p*net + ps) / pe'
(b) (p*net+ps)/pe, q/pe) space

Figure 11. The normalized stress paths for unsaturated soil (ua 50 kPa).
The state boundary surface in DST was examined through normalizing of the stress paths with the equivalent stress (pe), which was decided according to the consolidation curve of saturated specimen under anisotropic stress state. Figure 10 shows the
normalized stress paths for saturated soil in (p* , net,/pe q/pe) space. The stress paths reach and move over M.C.P line to the Hvorslev surface which is shown as the dotted line. This behavior corresponds to that observed under constant pressure condition. Figure 11(a) shows similar results for unsaturated
269
q/pe'
Saturated soil
Critical state line

urf
M.
s lev ors Hv
ace
lin e
C.P
Unsaturated soil
(p*net+ ps)/pe'
Figure 12. Sketch of the normalized stress paths for saturated and unsaturated soils.
soil. The stress paths reach and move over M.C.P line to the Hvorslev surface. Figure 11(b) shows same behavior that the stress paths reach M.C.P line and move along it after the suction stress s obtained from SSM was applied. These behaviors can be explained graphically as shown in Figure 12. The stress paths for saturated soil reach M.C.P line and move over it to the Hvorslev surface. On the other hand, in case of unsaturated soil, it was found that the stress paths move along M.C.P line through applying the suction stress. From this result, it could be concluded that the application of suction stress in DST has an important role to understand the stress path behavior. 6 CONCLUSIONS
several hypotheses, it was found that the stress paths moved along M.C.P line with progression of the shear process. This result represents the existence of CSL in DST under constant volume condition. 4. When the stress paths of the test results were normalized with the equivalent stress (pe), the stress paths for saturated soil showed the existence of the Hvorslev surface, and the results for unsaturated soil moved along M.C.P line through applying the suction stress. Accordingly, it can be concluded that the application of suction stress obtained from SSM has an important meaning to understand the stress path behaviors under constant volume condition. REFERENCES
Donald, I.B., 1956. Shear strength measurements in unsaturated non-cohesive soils with negative pore pressures. Proceedings, 2nd Australian and New Zealand Conference on Soil Mechanics and Foundation Engineering, 200205. Escario, V., 1980. Suction controlled penetration and shear tests. Proceedings, 4th International Conference on Expansive Soils, Denver, Co, Vol. 2: 781797. Escario, V. and Saez, J., 1986. The shear strength of partly saturated soils. Geotechnique 36(13): 453456. Fredlund D.G., and Rahardjo, H., 1993. Soil mechanics for unsaturated soils. John Wiley & sons, New York. Fredlund D.G., Rahardjo, H. and Gan J.K.M., 1987. Non-linearity of strength envelope for unsaturated soils. Proceedings, 6th International Conference on Expansive Soils, New Delhi, India, December 14, 4954. Gan, J.K.M., Fredlund, D.G. and Rahardjo, H., 1988. Determination of the shear strength parameters of an unsaturated soil using the direct shear test. Canadian Geotechnical Journal, 25(8): 500510. Karube, D., Kato, S., Hamada, K. and Honda, M., 1996. The relationship between the mechanical behavior and the state of pore-water in unsaturated soil. Geotechnical Engineering Journal, JSCE, 535(III-34): 8392. (in Japanese). Kim, B.S., Shibuya, S., Park, S.W. and Kato, S., 2010a. Application of suction stress for estimating unsaturated shear strength of soils using direct shear testing under low confining pressure. Canadian Geotechnical Journal. (in press). Kim, B.S., Kato, S. and Shibuya, S., 2010b. Critical state behaviors of unsaturated soil in the direct shear test under constant pressure, Proceedings of 5th International Conference on Unsaturated Soils, 68 September, Barcelona, Spain. (Under contribution). Satija, B.S., 1978. Shear behaviour of partly saturated soils. Ph.D. Thesis, Indian Institute of Technology, Delhi, India. Shibuya, S., Mitachi, T. and Tamate, S., 1997. Interpretation of direct shear box testing of sands as quasisimple shear, Geotechnique 47, No. 4, 769790.
The direct shear test under constant volume condition was carried out for compacted weathered granite specimen for saturated and unsaturated soils. The stress paths of the test results were examined on the critical state behaviors. The obtained results were summarized as follows. 1. The stress paths for saturated soil corresponded to the failure line under constant pressure condition in (v, ) space. In case of unsaturated soil, the stress paths moved over to the failure line and showed larger shear strengths. When the stress path reached around M.C.P line, the vertical stress in these stress paths changed abruptly into the increasing process in the decreasing process. 2. When the suction stress was applied to the test results in (v, ) space, the stress paths agreed with M.C.P line. These behaviors mean that the difference of the stress paths between saturated and unsaturated soils is affected by the suction stress. 3. As the stress paths were expressed with the Cam clay type stress parameters through applying
e sco Ro c rfa su e
270
Effects of matric suction on elastic shear modulus for three unsaturated soils
S. Kawajiri, S. Shibuya & S. Kato T. Kawaguchi
Graduate School of Engineering, Kobe University, Kobe, Japan Department of Civil Engineering, Hakodate National College of Technology, Hokkaido, Japan
ABSTRACT: An oedometer type apparatus has been newly developed for measuring variations of the elastic shear modulus, G of unsaturated soil under constant matric suction, which is equipped with a ceramic disk and a pair of bender elements (BE). Three kinds of soil sample were used to prepare specimens. After consolidation process, specified pore air pressure was applied to desaturate the specimen in step by step. Then the elastic shear modulus of unsaturated soils under each step were measured with the BE method to examine the effects of matric suction. It was found that the normalized shear modulus, G/f(e) for three sample were arranged uniquely against a confining pressure, (2/3v,net s) . This result means that the matric suction has influence in more wide suction range on small-strain behaviors of unsaturated soil as a component of the confining pressure. 1 INTRODUCTION 2 ELASTIC SHEAR MODULUS AND VOID RAITO FUNCTION
Soil stiffness has an important role when analyzing ground deformation in geotechnical engineering practices. It exhibits strong nonlinearity, and its strain-dependency is an important parameter to predict ground movement at small strain level. However, in some practical works, the small strain analysis encounters to unsaturated soils, and usually, it was assumed to be saturated state, which may cause over estimated prediction of the ground movement. At present time, there has been few research related to behaviour of unsaturated soils in small-strain level (e.g. Cho & Santamaria, 2001; Ng & Yung, 2008). In these studies, the samples used were limited, and the stiffness of unsaturated soil and its characteristic in small-strain level is not fully understood. In this study, an oedometer type apparatus, which is equipped with a ceramic disk and a pair of bender elements (BE), has been newly developed, and the effect of matric suction, s on the elastic shear modulus, G for unsaturated soil was examined under one dimensional consolidation state. Through investigations of the effects of matric suction on elastic shear modulus, we can estimate the elastic shear modulus for unsaturated soil with using the appropriate confining stress parameter which takes the effect of matric suction into account.
Hardin & Richart (1963) proposed the next empirical equation to estimate G. G = A f(e) (p/pr)n (OCR)k (1)
where, A = constant having the same dimension as G, f(e) = function of void ratio, OCR = overconsolidation ratio, p = mean effective stress, pr = reference effective stress, and n, k = experimental parameters. Furthermore, Jamiolkowski et al. (1994) have demonstrated that if f(e) is selected properly, then OCR have little influence on G, and G depends only on the current p, which means k equals to 0. If the confining stress parameter is selected appropriately, the relationship of the shear modulus could be arranged uniquely as observed in saturated soil. In this paper, we examined the case that the matric suction applies as a part of the confining stress for small-strain behavior. Because the void ratio, e will influence on the elastic shear modulus, in many research, the obtained elastic shear modulus is normalized by some void ratio function, f(e) to prevent its effect. Here, we adopt the next void ratio function, f(e) proposed by Shibuya & Tanaka (1996). f(e) = e1.5 (2)
271
TEST PROCEDURE
Beroflam cylinder Bender elemant
Silty soil, decomposed soil and fine sand were used for the test. Grain size distribution curves are shown in Figure 1. The physical properties of sample are listed in Table 1. Figure 2 shows the oedometer type apparatus used in this study. A ceramic disk, that has the air entry value of 500 kPa, was mounted inside the pedestal. The specimen size is 60 mm in the diameter and 40 mm in the height. The consolidation ring, which is surrounded by the pressure cell, is set on the pedestal. The pore water inside the specimen is released through the ceramic disk and the drainage line to the pressure released burette system. Therefore, the applied air pressure affects on the specimen as the matric suction. Table 2 shows the summary of test conditions. The specimen of silty soil was prepared from slurry state. After the ceramic disk and the drainage line are saturated, the slurry sample, whose water content is adjusted to 60%, was poured into the consolidation ring. Then the overburden pressure was applied by step by step. After attaining to the prescribed overburden pressure, the air pressure was applied to provide matric suction. On the other hands, the decomposed soil sample was prepared with static compaction. After the ceramic disk and the drainage line are saturated, the air dried sample of prescribed quantity was put into
Percentage finer (%)
Cell
Ceramic disk
study.
Figure 2.
Oedometer type apparatus used in this

Summary of test conditions. Matric suction Overburden pressure v (kPa) s (kPa) 50, 100, 200 10203050100 150200300 400 812153050 100200300 25103050 100
Table 2. Sample Silty soil
Decomposed 50, 100, 200 soil Fine sand 50, 100
100 80 60 40 20 0 0.001
Silty soil Decomposed soil Fine sand
0.01
Grain size (mm)
0.1
Figure 1. Grain size distribution curves of the
samples.
Table 1. Physical properties of samples. dmax (g/cm3) wopt (%)
Sample Silty soil Decomposed soil Fine sand
s (g/cm3) 2.65 2.62 2.62
wL
wP
PI
emax emin 0.934 0.605
NP NP NP NP NP NP 1.94 11.7
the consolidation ring. Then it was compacted statically with the loading plate to attain the relative dry density of 88% for the maximum dry density that was obtained with JGS 0711 A method. Then the specimen was saturated by water immersion for 12 hours. Finally, the specimen of fine sand was prepared by using water-pluviation method so that the initial relative densities of specimen (Dr) would be equal to Dr = 50%. The overburden pressures of 50, 100 and 200 kPa at net stress were applied for consolidation process. After the consolidation process, the air pressure was applied to provide matric suction conditions step by step. When the drainage from the specimen attains equilibrium under the prescribed overburden pressure or matric suction, the elastic shear modulus was measured with the BE method. A pair of bender elements was mounted at the loading plate side and the pedestal side in the apparatus. BE method is a simple technique that can be measure shear wave velocity Vs, and decide the elastic shear modulus with next equation. G = t (Vs)2 where, t = total density of the specimen. (3)
272
Start-to-start method described by Kawaguchi et al (2001) was adopted to decide the travel time of shear wave, whereas tip-to-tip method was adopted to decide the travel distance (Viggiani & Atkinson, 1995). The voltage of the input pulse of 5 volts was applied for BE by using both the sine wave (10, 12 and 15 kHz) and rectangular wave (10 kHz). 4 RESULTS AND DISCUSSIONS
4.1 SWCC for there soil samples Figure 3 shows the SWCC under different overburden pressures. For silty soil sample, the air entry value was from 80 to 100 kPa. From this result, we
understand that in the cases that the matric suction increases from 0 kPa to 100 kPa, from 100 kPa to 200 kPa, and from 200 kPa to 400 kPa, the pore water distributions correspond to the boundary effect stage, the transition stage, and the residual stage, respectively (Vanapalli et al.1996). However, we were not able to identify similar characteristic for decomposed soil and fine sand, because it contains larger soil particle, and its air entry value become smaller than the silty soil. This result clearly shows that SWCC is strongly related grain size distribution. 4.2 Characteristics of Elastic shear modulus of unsaturated soils
100 Degree of saturation, Sr (%) 80 60 40 20 0
suctin history : s = 10kPa 20kPa 30kPa 50kPa 100kPa 150kPa 200kPa 300kPa 400kPa
Silty soil : v,net = 50kPa : v,net = 100kPa : v,net = 200kPa
a)
100
200
300
30kPa
400
Matric suction, s (kPa)

100
suction history : s = 8kPa 12kPa 15kPa 50kPa 100kPa 200kPa 300kPa
Degree of saturation, Sr (%)
Decomposed soil
80 60 40 20 0
b)
Figures 4, 5 and 6 show the relationships of elastic shear modulus to matric suction and overburden pressure for silty soil, decomposed soil and fine sand respectively. Figures 4 and 5 (silty soil and decomposed soil) show similar tendencies that the elastic shear modulus increases with the overburden pressure and the matric suction. In Figure 6, fine sand shows different tendency from silty soil and decomposed soil. These test observations clearly indicate that the elastic shear modulus of unsaturated soils depends on two stress state variable: the matric suction and the overburden pressure (net normal stress). In fine sand, the elastic shear modulus has little changes for increase of the matric suction in comparison with silty soil and decomposed soil. In another words, the effects of the matric suction for the elastic shear modulus are different depending on the soil type.
: v,net = 50kPa : v,net = 100kPa : v,net = 200kPa
150 100 Elastic shear modulus, G (MPa) 50 0
Silty soil
a)
100
200
300
400

suction history : s = 2kPa 100kPa 5kPa 10kPa 50kPa
100 Degree of saturation, Sr (%) 80 60 40 20 0
Fine sand
unsaturated 0 100 200 300 Matric Suction, s (kPa) saturated
:v=50kPa :v=100kPa : v=200kPa 400
150 100 50 0
b)
c)
: v,net = 50kPa : v,net = 100kPa
100
200
300
400
v = 50kPa100kPa200kPa300kPa400kPa 0 100 200 300 Overburden pressure, v (kPa) 400
Figure 3. SWCC in drying path. a) silty soil, b) decomposed soil, c) fine sand.
Figure 4.
Relationship between s and v,net (silty soil).
273
250 200 150 Elastic shear modulus, G (MPa)
Decomposed soil
a)
100 80 60 Elastic shear modulus, G (MPa)
Fine sand
a) : v =50kPa : v =100kPa
100 50 0 250 200 150 100 50 0 0 0 saturated unsaturated 100 200 300 Matric suction, s (kPa)
: v = 50kPa : v =100kPa : v =200kPa 400
40 20 0 100 80 60 40 20 0 0 v= 50kPa100kPa200kPa300kPa400kPa 100 200 300 Overburden pressure, v (kPa) 400 0 unsaturated 20 saturated 40 60 80 Matric suction, s (kPa) 100
b)
b)
v = 50kPa100kPa200kPa300kPa400kPa 100 200 300 Overburden pressure, v (kPa) 400
Figure 5. soil).
Relationship between s and v,net (decomposed
Figure 6.
Relationship between s and v,net (fine sand).
v,net = 100kPa 0.453 0.760 0.795

Void ratio,e
0.450 0.750
a) 100 200 500 ln (v,net+s) (kPa)
Silty soil 60 80 100
b) 100 ln (v,net+s) (kPa)
Decomposed soil 80100 ln G (MPa) 200 0.794
c) 100 ln (v,net+s) (kPa)
Fine sand 300 100 ln G (MPa)
ln G (MPa)
Figure 7.
Relationship among e, (V,net + s) and G. a) silty soil, b) decomposed soil, c) fine sand.
Figures 7 a), b) and c) show the relationships of void ratio, e, for (v,net + s) and the elastic shear modulus, G for the there samples, respectively. The relationships of void ratio for (v,net + s) and the elastic shear modulus are similar each other, and the void ratio decrease almost linearly for semi log scale. These results mean that G is strongly related to (v,net + s). From the results mentioned above, we understand that the elastic shear modulus has relation-
ship with the matric suction and the overburden pressure. Here, we will examine these relationships more detail. Figure 8 show the relationships of the overburden pressure, which takes account of the matric suction, to the normalized shear modulus. From this figure, we can see the relationship of normalized shear modulus, G/f(e) to (2/3v,net + s) is arranged uniquely for all samples. These results could be explained according to the intergranular adhesive from as follows. In the two-spheres model
274
1000 400 G/e 1.5 (MPa) 100 40 10 4 1 10
Silty soil , , : unsaturated : saturated
Fine sand , : unsaturated : saturated
n = 0.44 residual state in fine sand Decomposed soil , , : unsaturated : saturated 20 60 100 , , : v,net = 50kPa , , : v,net = 100kPa , : v,net = 200kPa 200 600 1000
2/3v,net + s (kPa)
Figure 8. Relationship between G/e1.5 and (2/3v,net + s).
with meniscus water, the intergranular adhesive force is given by the next equation. F = r22 (ua uw) 2 r2 T (4)
elastic shear modulus for unsaturated soil. A silty soil, decomposed soil and fine sand were used with different specimen preparation method. After the consolidation process, the specimen was desaturated with applying the matric suction step by step. Then the elastic shear modulus was measured in each unsaturated state. From the obtained test results, we find that both the matric suction and the overburden pressure (net normal stress) affect on the elastic shear modulus. We also find that the relationship of normalized shear modulus, G/f(e) against the sum of confining stress and the matric suction, (2/3v,net + s) shows an unique tendency for all samples. It should be noted that the matric suction has influence on the elastic shear modulus as a component of confining pressure. Because these results shown in this paper are limited, we should accumulate more data to verify these tendencies. REFERENCES
Cho, G.C. & Santamarina, J.C. 2001. Unsaturated particulate materials-Particle level studies, Jour. of Geotech. & Geoenvir. Eng., ASCE, 127(1):8496. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsaturated soils, Can. Geotech. J., 15(3):13321. Hardin, B.O. & Richart, F.E. Jr. 1963. Elastic wave velocities in granular soils, Jour. of SMF Div., ASCE, 89:3365. Jamiolkowski, M. Lancellotta, R. & Lo Presti, D.C.F. 1994. Remarks on the stiffness at small strains of six Italian clay, Pre-failure Deformation of Geomaterials, 2:817836 Rotterdam: Balkema. Karube, D. & Kato, S. 1994. An ideal unsaturated soil and the Bishops soil, Prc. 13th Int. Conf. Soil Mech. Found. Engng., New Delhi, 1:4346. Rotterdam: Balkema. Karube, D., Kato, K., Hamada, K. & Hond, M. 1996. The relationship between the mechanical behavior and the state of porewater in unsaturated soil, Journal of Japanese Society of Civil Engineering, 535:8392. (in Japanese). Kawaguchi, T., Mitachi, T. & Shibuya, S. 2001. Evaluation of shear wave travel time in laboratory bender element test, Proc. of the 15th ICSMGE, 1:155158. Rotterdam: Balkema. Ng, C.W.W. & Yung, S.Y. 2008. Determination of the anisotropic shear stiffness of an unsaturated decomposed soil, Gotechnique, 58(1):2335. Shibuya, S. & Tanaka, H. 1996. Estimate of elastic shear modulus of Holocene soil deposits, Soils and Fundations. 36(4):4555. Vanapalli, S.K., Fredlund, D.G., Pufahl, M.D. & Clifton, A.W. 1996. Model for prediction of shear strength with respect to soil suction, Can. Geotech. Journal, 33(3):379392. Viggiani, G. & Atkinson, J.H. 1995. Interpretation of bender element tests, Gotechnique, 45(1):149154.
where, F = the intergranular adhesive force by the matric suction, r2 = the meniscus radius, (ua uw) = the matric suction, ua = the pore air pressure, ua = the pore water pressure, and T = the surface tension of water. We can understand that based on this equation, the intergranular adhesive force increase with matric suction. However, the spheres hardly move at small strain level with the increase of matric suction. This behavior will explain the test results observed in this study. Consequently, when the matric suction increases, the elastic shear modulus increases with the increase of intergranular adhesive force. Therefore, the matric suction has influence on the characteristic of elastic shear modulus. In the case of fine sand, the matric suction hardly influence on the elastic shear modulus in high suction regions (residual state) as shown in Figure 7. In Figure 9, however, the results of fine sand exist around the other results, and we can understand that test results show an unique relationship. The solid line in Fig. 8 shows equation (1) for the experimental parameter, n of 0.44. However, it is deduced that this parameter depends on SWCC, soil type and particle size, etc. Therefore, we should accumulate more data to verify this experimental parameter value. 5 CONCLUDING REMARKS
In this paper, we studied the elastic shear modulus for unsaturated soil under one dimensional consolidation state by using the bender element method, and examined the effect of matric suction on the
275
Mechanical behaviors and water retention characteristics of micro-porous ceramic particles produced by burning sludge
J. Kawazu, K. Kawai, S. Kanazawa & A. Iizuka
Kobe University, Kobe, Japan
ABSTRACT: Shortage of landfill capacity in areas where industrial waste materials are dumped is a serious environmental problem. In this study, the mechanical characteristics of two kinds of soil samples, produced by burning sludge, were examined. In order to grasp some key fundamental mechanical properties of these soil samples, a series of laboratory tests were performed. It was found that the stress-strain characteristics and the permeability of these soil samples were very similar to those of standard sandy materials. In addition, particles in the soil samples were easily crushable and exhibited high water retentivity. Next, the applicability of the unsaturated elasto-plastic constitutive model of Ohno et al. (2007), in reproducing the mechanical behavior of these soil samples, was examined. The results of this study will help in useful utilization of micro-porous ceramic materials, obtained from sludge burning, in the field of geotechnical engineering. 1 INTRODUCTION to innocuous micro-porous ceramic materials. In this study, fundamental mechanical characteristics of two kinds of soil samples were examined. The samples were produced by burning sludge materials and contained microscopic pores in their particles. In order to grasp some fundamental mechanical properties of the micro-porous ceramic particles, a series of laboratory tests, namely, triaxial compressive shear tests, oedometer tests, permeability tests and water retention tests, were then performed. Finally, the applicability of the unsaturated elasto-plastic constitutive model proposed by Ohno et al (2007) was examined with respect to the mechanical behavior of these soil samples. 2 PHYISICAL PROPERTIES OF MATERIALS
Many industrial processes produce sludge as a waste product. For example, paper sludge (PS), made from condensed and dehydrated fibers and minerals which cannot be recycled to paper, is discharged during recycling of paper. Similarly, water purification plants discharge polluted sludge, made up of condensed suspensoid and sediment of sand and clay. These sludge materials are primarily regarded as industrial waste materials. When such sludge materials are mixed with a small amount of clay and burned at 1300 degree Celsius, they change
Photo 1.
Materials used in this study.
Photo 2.
SEM picture.
Materials used in this study are: i) burnt mixture of PS and clay and; ii) burnt sludge discharged from water purification plant (see Photo 1). These samples are hereafter referred to as white or red samples, respectively. As evident from scanning electron microscopy (SEM) data (see Photo 2) both the samples have innumerable voids in the diameter range 550 m. The specific gravity of soil particles were 2.58 and 2.61 for the white and red samples respectively. The maximum/minimum dry densities measured in a cylinder (5 cm in diameter and 10 cm in height) were 1.45/1.14 g/cm3 and 0.98/0.75 g/cm3 the white and red samples respectively. Figure 1 describes the particle size distribution curve. Geoengineering grain size classification
277
showed the white sample to consist of 10% clay, 10% silt, 63% sand and 17% gravel, while the red sample consisted of 10% clay, 5% silt, 41% sand and 44% gravel. Both samples were found to be well-graded and included a lot of sand. 3 STRESS-STRAIN CHARACTERISTICS
A series of triaxial compression tests (CD tests) was performed to examine the stress-strain characteristics of the soil samples. Figure 2 shows
Percent of finer by weight
100 80 60 40 20 0
white red
10-3
10-2 10-1 1 Diameter (mm)
10
Figure 1.
Particle size distribution curves.

Axial displacement meter Back pressure Load cell
Cell pressure
Porous stone Burette Specimen Differential pressure meter Pore pressure meter Porous stone
Data Logger
Torque motor DA converter
Computer
Figure 2.
Triaxial test apparatus.

1600 -16
the test apparatus used in this study. Dense and loose (relative density, Dr 80% and Dr 30%, respectively) specimens were prepared for both white and red samples. They were set in the triaxial apparatus and distilled water was circulated from the bottom to the top of the specimen for 24 hours to reach saturation. Interestingly, the specimens did not become fully saturated during the tests, due to the existence of many micro voids in the particles thought to prevent saturation. The specimens were then axially compressed under draining condition under various confining pressures (100, 200 and 300 kPa) applied at a rate of 0.1 mm/min. Figure 3 shows how the principal stress difference and volumetric strain change with shear strain for the white samples. Likewise, Figure 4 is for the red samples. It is generally known that strain softening and dilation, due to shear, occurs in sandy soil like these materials. However, both the white and the red samples also show strain hardening and slight dilation under loose conditions. These behaviors resemble that of clayey soil. Figures 5 and 6 show the stress state at the failure point. The closed circle plots indicate the stress state at the peak point of the principal stress difference and the closed triangle plots indicate the stress state at the kink point of the volumetric strain. All interpolated lines intercept the origin of the stress axis. This behavior is also observed for clayey soil. A series of oedometer tests was performed to examine the compressibility of these materials. Figure 7 shows the test apparatus used in the experiment. Dense (Dr 80%) and loose (Dr 30%) specimens were prepared for both the white and the red samples. They were adjusted to a certain water content, w0, and consolidated under a vertical load, from 4.9 to 1256 kPa, in the oedometer. For comparison, air-dried (about zero water content) and soaked specimens were also examined. Figures 8 and 9 show the consolidation
1600
r' : 200kPa
Principal stress difference q (kPa) Principal stress difference q (kPa)
300kPa
-12
1200 -12
100kPa
-16
1200
800
-8
800
-8
400
v
100kPa 300kPa
-4
400
-4
v
0 0
r' : 200kPa
-400 0
Shear strain s(%) (a) Dr 80%
10
4 15
-400 0
Shear strain s(%) (b) Dr 30%
10
4 15
Figure 3.
Stress and strain relation of the white sample under triaxial compression.
278
1600
-16
1600
r': 200kPa
Principal stress difference q (kPa) Principal stress difference q (kPa)
1200 -12
1200
100kPa 300kPa
-16
-12
800
-8
800
-8
400
-4
400
-4
v
v
-400
r' : 200kPa
-800 0
100kPa
-400
Shear strain s(%) (a) Dr 80%
10
300kPa 8
15
-800 0
Shear strain s(%) (b) Dr 30%
10
8 15
Figure 4.
Stress and strain relation of the red sample under triaxial compression.
Principal stress difference (kPa) Principal stress difference (kPa)
1600 1600
(a) white 80%

1200 800 400
(b) white 30%

1200 800 400
at qmax at vmax
200 400 600 800 1000
at qmax at vmax
200 400 600 800 1000
Effective mean stress (kPa)
Figure 5.
Stress states at the failure point on the white sample.

Principal stress difference (kPa)
1600
Principal stress difference (kPa)
1600
(a) red 80%

1200 800 400
(b) red 30%

1200 800 400
at qmax at vmax
200 400 600 800 1000
at qmax at vmax
200 400 600 800 1000
Figure 6. Stress states at the failure point on the red sample.

Computer Data logger
Displacement gauge
Loading cap Guide ring Porous plate Specimen Consolidation ring
Bellofram cylinder
Figure 7.
Oedometer test apparatus.
curves under one-dimensional compression. The air-dried and soaked specimen can be regarded as residual saturated and saturated specimens, respectively. The compressibility would decrease with desaturation according to unsaturated soil mechanics. However, the results of the red sample do not show such a behavior. The materials used in this study are crushable. Especially, particles in the red sample have many voids that impart high crushability. Particle crushability depends on soil moisture. Therefore, the effect of particle crushing due to consolidation pressure is likely to be reflected in the compression curve. Katagiri et al. (1999) conducted oedometer tests on Sirasu soil, which has particles that are easy to crush, and
279
1.3
Void ratio e
1.1 1 0.9 0.8 0.7 10 102
Void ratio e
1.2
(a) white 80%
air-dried w0 =20% w0 =30% soaked
1.3 1.2 1.1 1 0.9 0.8
(b) white 30%
Consolidation pressure p (kPa)
103
0.7
air-dried w0 =20% w0 =30% soaked

10
102 103
Figure 8.
Consolidation curves of the white samples in oedometer test.

2.6
Void ratio e
2.2 2 1.8 1.6 1.4 10 102
Void ratio e
2.4
(a) red 80%
air-dried w0=20% w0=30% w0=40% soaked
2.6 2.4 2.2 2 1.8 1.6
(b) red 30%
103
1.4
air-dried w0=20% w0=30% w0=40% soaked
10
102
103
Figure 9.
Consolidation curves of the red samples in oedometer test.
demonstrated that the process of particle crushing under compression made the compression curve steep. Our test results correspond to their study. 4 PERMEABILITY AND WATER RETENTION CHARACTERISTICS
Before conducting shear tests, a permeability test was carried out for each specimen in the triaxial test apparatus. A water head difference, between the top and the bottom of the specimen, was generated and the resulting flux was measured. Consequently, the coefficients of permeability were estimated as 2.7 104 cm/sec for the dense white sample, 6.9 104 cm/sec for the loose white sample, 9.1 104 cm/sec for the dense red sample, and 1.2 103 cm/sec for the loose red sample. The samples were found to exhibit relatively high permeability given their grain size distribution. Additionally, in order to examine the water retention characteristics, soil column tests were performed. Figure 10 shows the soil column test equipment. The material was spread in a mould and compacted to a relative density of Dr 80%. Four such moulds, each with an inner diameter and height as 20 cm, were piled up as shown in the figure. The moulds were equipped with an ADR censor and tensiometer for measuring the water content and the pore water pressure. In order to increase the degree of saturation of the specimen, distilled water was supplied from the
Figure 10.
Soil column for water retention test.
bottom (to the top) of the specimen. After the water level reached the top surface, the specimens were incubated for several hours, and then the water was slowly drained with the measurement of the water content and the pore water pressure. The water level was finally lowered to 3 m from the bottom of the equipment. The test results are summarized in Figures 11 and 12. Changes in the volumetric water content and the pore water pressure, with time, are shown in these figures. At the beginning of the test, an increase in the volumetric water content and the pore water pressure could be observed, caused by the rise of the water level within the soil column. Sequentially, rundown
280
60 40 20 0 0 2000
1st layer (10 cm from the bottom) 2nd layer (30 cm from the bottom) 3rd layer (50 cm from the bottom) 4th layer (70 cm from the bottom)
Pore water pressure (kPa)
80
10
-10 0 2000 4000 6000 8000 10000
Time (min) (a) Changes in volumetric water content
4000
6000
8000
10000
Time (min) (b) Changes in pore water pressure
Figure 11.
Results of water retention tests (white sample).

150
10
100
50
0 0
2000
4000
6000
8000
Time (min) (a) Changes in volumetric water content
-10 0
2000
4000
6000
8000
Time (min) (b) Changes in pore water pressure
Figure 12. Results of water retention tests (red sample).

150
5
1st layer 2nd layer
Degree of saturation (%)
APPLICABILITY OF UNSATURATED CONSTITUTIVE MODEL
100
50
3rd layer
suction (kPa)
10
Figure 13. Water retention characteristic curves (white sample).

150
Applicability of the unsaturated elasto-plastic constitutive model, proposed by Ohno et al. (2007), to understand the stress-strain behavior of micro-porous ceramic materials, was examined. Ohno et al. indicated that existing unsaturated constitutive models were equivalent in terms of expressing yield function of effective stress and parameter associated with stiffness. Ohno et al. expressed effective stress for unsaturated soil as shown in Equation (1), and proposed a general yield function as shown by equation (3). net ps 1 ps Se s (1) (2)
100 3 layer 50 4 layer

th rd
f , , vp MD ln
p q D vp 0 psat p
(3)
suction (kPa)
Figure 14. Water retention characteristic curves (red sample).
of water level caused decrease of the volumetric water content and the pore water pressure, giving rise to water retention characteristics as shown in Figures 13 and 14. According to these figures, it is found that these materials show high water retentivity even under a high suction range.
In the above equations, ' is the effective stress tensor; net is the net stress tensor, 1 is the second rank unit tensor, is the total stress tensor, s is suction, ps is the suction stress, Se is the effective degree of saturation, vp is the plastic volumetric strain, M is q/p at the critical state and D is the dilatancy coefficient. Increase of yield stress due to desaturation is expressed as the product of yield stress in the saturated state, psat , and a parameter contributing to hardening, . An expression for the parameter (Equation (4)) was introduced by Ohno et al. (2007) as:
n exp 1 Sre ln a
(4)
281
1600
1600
Principal stress difference q (kPa)
1200
Principal stress difference q (kPa)
:Experimental :Simulated
1200
800
800
400
400
0 0
5 10 15
0 0
10
15
Shear strain s (%)

(a) White sample
Shear strain s (%)

(b) Red sample
Figure 15. Changes in principal stress difference under shear.

-4 0
Volumetric strain v (%)
-2
Volumetric strain v (%)
8 0 5 10 15
4 0
10
15
Shear strain s (%)

(a) White sample
Shear strain s (%)

(b) Red sample
Figure 16. Changes in volumetric strain under shear.
Equation (3) is reduced to the Cam-Clay Model upon saturation (Sre 1 and 1). The constitutive model of Ohno et al. was developed from a model for clayey material. Soil materials used in this study can be categorized as sandy material. Yet, as described earlier, their behavior was similar to clayey soil. To understand the phenomenon better, the triaxial compression tests on loose specimen were simulated with input parameters: compression index, 0.037, expansion index, 0.003, and, M 1.74 for the white specimens, and, 0.105, 0.003, and, M 1.72 for the red specimens. A comparison between experimental and simulated data, as shown in Figures 15 and 16, shows that the constitutive model can explains the observed stress and shear strain behaviors. A different approach, such as the subloading surface model, proposed by Hashiguchi (1980), is needed for expressing the occurrence of strain softening after the appearance of the stress peak. Despite its success, the simulated volumetric strain could not reproduce its dependency on the confining stress, as shown in experiments. This is because the effect of the confining stress on dilatancy is expressed only in the first component in the right side of equation (3) and its effect is very small. This calls for the modification of the model, including consideration of the hardening parameter.
CONCLUSIONS
In this study, the mechanical behavior of two kinds of soil, obtained by burning sludge, was examined. It was found that crushability and high water retentibility were two principal characteristic properties of the soils. Although more elaborate studies along with modifications of the constitutive model used in this study are needed for a complete characterization, the results of this study will facilitate the effective utilization of recycled materials like those considered here and help reduce factory waste in the future. REFERENCES
Hashiguchi, K. 1980. Constitutive equations of elastoplastic materials with elastic-plastic transition. Journal of Applied Mechanics, ASME, 47: 266272. Katagiri, M., Katakami, N., Saito, K. & Takayanagi, S. 1999. Relationship between indices of particle crushing and compressibility of Shirasu, Proc. of Sym. on Engineering problem on crushable ground: 8588 (in Japanese). Ohno, S., Kawai, K. & Tachibana, S. 2007. Elasto-plastic constitutive model for unsaturated soil applied effective degree of saturation as a parameter expressing stiffness, Journal of JSCE 63(4): 11321141 (in Japanese).
282
Effect of suction hysteresis on the shear strength of unsaturated soil interfaces

Charbel N. Khoury & Gerald A. Miller
School of Civil Engineering and Environmental Science, University of Oklahoma, Norman, OK, USA
ABSTRACT: The design of soil-structure interfaces is often dominated by the shear strength of the interface, which is affected by seasonal variations (e.g. water table, precipitation) of the soil moisture content. Since unsaturated soils behave differently under drying and wetting (i.e. hysteresis), this research investigates the influence of stress history (i.e. cyclic suction or hysteresis) on the unsaturated interface shearing behavior. To this end, direct shear tests results were conducted on suctioncontrolled rough interfaces under cyclic suction (drying/wetting). Rough steel interface tests were conducted using cohesionless silt. Results indicated that the interface shear strength is dependent on the suction-stress history of the soil. For a given suction, specimens that underwent cyclic suction (i.e. drying followed by wetting) showed higher peak shear strength than those undergoing drying only prior to shearing.
INTRODUCTION AND BACKGROUND
Soil-structure interaction is a common problem in geotechnical engineering that has been widely studied in an attempt to obtain better and more realistic solutions. However, few studies have focused on the interface behavior of these structures in unsaturated soils whose behavior differs considerably from traditional saturated soils. Typical examples of such structures include pile foundations, retaining walls, mechanically stabilized earth walls, reinforced soil slopes, liners for landfills and tunnels. Climatic changes (i.e. water content changes) influence the interface shear strength of these structures. Unsaturated interfaces undergoing drying would behave differently than those undergoing wetting. This behavior is hysteretic in accordance with the hysteresis behavior of soil water characteristic curves (SWCC). Limited studies have examined the effect of hysteresis on the shear strength of unsaturated soils. For example, Thu et al. (2006) conducted constant water content (CW) and consolidated drained (CD) triaxial tests and showed that friction angles with respect to suction (b) from CW differ from those of CD tests; this difference was attributed to the hysteretic behavior of the SWCC. Nishimura and Fredlund (2002) studied the influence of drying and wetting process on the shear
strength of a compacted silty soil and compacted kaolin subjected to high total suction. The compressive strength of the dessicated soil during wetting was less than that during drying, which was related to the contact area of water in the soil during drying and wetting history. Han et al. (1995) conducted direct shear tests to examine the hysteresis effect of drying and wetting on shear strength and showed that wetting had lower shear strength compared to drying under the same matric suction. This difference was attributed to the lower water content in the soil during wetting compared to drying. On the other hand, Galage and Uchimura (2006) reported that the soil at wetting had higher shear strength as compared to the soil at drying under same suction, for sandy-silt soil. Shemsu et al. (2005) studied the cyclic suction loading influence on the shear strength of unsaturated soil. Results showed specimens that underwent cyclic suction loading had higher peak shear strength. However, to the authors knowledge, little or no study that investigation of the effect of drying and wetting has been performed on the shear strength of interfaces. Thus, this study aims at providing experimental data using an interface direct shear test with suction control, where tests are first conducted along the drying curve, and compared with tests conducted at same suction but following wetting after drying.
283
TEST MATERIALS AND SOIL SPECIMENS PREPARATION
A stainless steel plate is used in this study as a counterface. The plate is 25.5 mm thick and 102 mm in diameter with rough surface geometry. The surface roughness (Rn) is defined based on the roughness profile as Rn Rmax/D50, where, Rmax is the maximum peak to valley height, and D50 is the grain size diameter of soil corresponding to fifty percent finer. Rn for the rough steel plate used in this study was approximately 5.4. The soil used in this study is a mixture of two commercially available manufactured soils, Sil-CoSil 250 manufactured by U.S. Silica Company and Glass Beads, Size BT-9, manufactured by Zero Products. The soil mixture consists of 75% ground silica and 25% glass beads by weight. The mixture is non-plastic and has a grain size distribution similar to that of fine sandy silt (48% fine sand, 46% silt, and 6% clay size material). This soil was used for its advantage of obtaining experimental results in a reasonable amount of time compared to natural soils with similar suction range. The soil water characteristic curve (SWCC) of this soil for primary drainage and primary wetting, starting from as-compacted conditions (So 67%) are shown in Figure 1 for net normal stresses of 150 kPa. The soil specimens prepared in this study, for both SWCC and interface testing, are compacted to an initial dry density of 15.4 kN/m3 and at a moisture content of 17.2 1%. The soil is mixed to the desired moisture content and compacted to the required density by moist tamping (i.e. volumebased compaction) directly inside the shear box, of a modified direct shear test, which is assembled by holding the upper and lower half of the shear box together using two screws. The modified direct
shear test device was developed by Miller and Hamid (2007). The device allows for suction controlled tests of unsaturated soil and unsaturated interfaces using the axis translation technique. More details of the apparatus are given by Miller and Hamid (2007).
CYCLIC SUCTION UNSATURATED INTERFACE SHEAR TESTING PROCEDURE
100.0 80.0
Primary Drying Primary Wetting

n-ua= 150 kPa, So = 67%
60.0 40.0 20.0 0.0
0.00
0.05
0.10
0.15
0.20
Figure 1. used.
Soil water characteristic curves of the soil
The procedure to conduct unsaturated direct shear interface tests (DSIT) following the drying path is described by (Khoury et al. 2009, Hatami et al. 2008, Miller and Hamid 2007). In summary, the HAEPD is first saturated with de-aired water prior to testing. A seating load of 1435 kPa is applied on each test specimen in order to stabilize the position of the upper half of the shear box when it is raised to introduce the gap needed before the shearing process. Approximately one hour after the initial seating load, the two screws holding the two halves of the shear box together are removed from the pressure chamber. Then, a gap is created between the two halves of the shear box and the air chamber lid is sealed with bolts after which a target matric suction value (ua uw) following the drying curve is applied to the specimen by increasing the air (ua) and water pressures (uw) simultaneously via the axis translation technique. After equilibrium of matric suction, a target net normal stress (n ua) is applied in increments of 35. After the specimen is consolidated under the target (n ua), it is then subjected to drained shearing while controlling both suction and net normal stress. A shear force is applied to the direct shear device to maintain a displacement rate of 0.005 mm/min up to about 10 mm displacement. Changes in specimen height (measured by LVDTs) and water volume (measured using an automated pump with pressure and volume control) were recorded during all stages of the test. On the other hand, direct shear interface tests (DSIT) with hysteresis are carried out similar to the procedure described above. However, after applying the same initial matric suction (e.g. 50 kPa) and target net normal stress the sample is dried out by increasing suction on the SWCC drying path to a higher value (e.g. 100 kPa) and wetted again to the previous initial suction (as shown in Figure 2) where shearing is conducted to investigate the effect of SWCC hysteresis on the shear strength results. The samples tested on the drying curve are referred as (D) tests while the ones undergoing wetting after drying are referred as (DW).
ua-uw (kPa)
284
100.0 80.0
DSIT tested at a suction on the drying path (D)
0.000 0.002 0.004 v/H0 0.006 0.008

DSIT tested at same suction on wetting path (DW)
(a)
ua-uw (kPa)
60.0 40.0 20.0 0.0
0.010
0.15 0.20
0.00
0.05
0.10
0.012 0 -2 -4 -6 -8 -10 -12 -14
Drying Wetting after Drying

Suction = 50 kPa, (D) test
(b)
Suction = 50 kPa, (DW) test
w %
Figure 2.
Illustration of the DSIT hysteresis test path.
PRESENTATION AND DISCUSSION OF RESULTS
Suction = 100 kPa,
Figure 3a shows a typical plot of vertical displacement (normalized to the specimen height) with time during the equalization phases (i.e. phases involving the application of suction and net normal stress to the specimens) for both drying tests (D) and wetting after drying (DW) at net normal stress of 150 kPa and suction values of 50 kPa. It is observed that the magnitude of specimen compression is practically the same at given suction and net normal stress values. A comparison of change in water content (w%) during equalization periods in the corresponding (D) and (DW) test specimens is shown in Figure 2b. It is observed that the amount of water (w%) drained from the specimen for (D) test at suction 50 kPa is about 8%. However, (DW) test initially drained about the same amount (w 8%) at same suction during drying, then w reached approximately 13% when suction was increased to 100 kPa, and then decreased back to about 11% at suction of 50 kPa on the wetting curve. This hysteresis is similar to the SWCC where for a given suction the water content on the wetting curve is less than that of the drying curve. Shear stress () and volume change responses during shearing of rough unsaturated interface tests subjected to drying (D) and wetting after drying (DW) at same matric suction (50 kPa) and net normal stress (150 kPa) are shown in Figure 4. Based on this figure which is typical of unsaturated rough steel interface behavior for all levels of matric suction and net normal stress, some important observations are summarized as follows: During shearing of both (D) and (DW) tests, peak shear strength is achieved followed by strain softening to a post peak shear stress.
2500 5000 7500 10000 Time (min)
Figure 3. Comparison of consolidation results of rough interface for drying and wetting after drying at similar ua uw of 50 kPa and n ua of 150 kPa.
The interface peak and post peak shear strength from wetting (DW) tests are slightly higher than that of the drying (D) tests. The magnitude of horizontal displacement corresponding to the interface peak shear strength of (DW) tests is smaller than that of the (D) tests. The volume changes during shearing (Figure 4b) indicate that (D) tests slightly compressed initially then started to dilate. However, (DW) tests started dilation just after shearing started. For both tests, dilation apparently started prior to mobilizing the peak shear strength and then decreased once the strain softening process was completed and the shear stress reached the interface post peak shear strength. During shearing, some changes in the specimens water content (w%) are detected in both tests. It is observed that water drained out of the sample for (D) test, but negligible amount of water drained into the sample for (DW) tests. Although both tests dilate, water drainage of (D) tests indicates that there was a tendency for suction to decrease due to increasing pore water pressure. It seems that the breaking of the airwater menisci during shearing created a tendency for increasing pore water pressure. Since those tests are drained suction control, water drained
285
140 120 100 (kPa) 80 60 40 20 0 -0.006 -0.005 -0.004 v/H0 -0.003 -0.002 -0.001 0.000 0.001 0.1 0.0 w (%) -0.1 -0.2 -0.3 -0.4 -0.5 -0.6
140 120 (kPa)

(a)
100 80 60 40 20 0
(b)
0 10 20 30 40 50 60 70 80 90 100110
ua - uw (kPa)
Figure 5. Interface peak shear strength versus suction (ua uw) for (D) and (DW) tests at n ua of 150 kPa.
(c)
10
Horizontal Displ. u (mm)

Figure 4. Influence of suction hysteresis (drying and wetting after drying) on the shear response of rough interfaces at matric suction of 50 kPa net normal stress of 150 kPa.
during shearing to maintain the pore water pressure and suction. This behavior seems consistent with the fact that drying (D) tests which have higher water content than that of the (WD) tests at the same suction. Since various research studies resulted in observations of different shear strength behavior (i.e., some obtained higher strength for wetting while others found that drying strength was higher than wetting, as discussed in the literature), it appears more research is needed to investigate further this hysteretic behavior. Studies (e.g. Han et al. 1995, Nishimura and Fredlund 2002) that showed lower wetting than drying attributed this behavior to the
interparticle forces which are related to the contact area between water and the solid particles. Lower water content (wetting) will have smaller contact area than that of drying. At this time, it seems for this sandy silt material that the suction stress history has a more dominant effect than the solidwater interfacial area on interface shear strength. During shearing, water may be acting as lubricant between the particles and thus resulting in a lower interface shear strength for drying (with higher water content) than that of wetting (with lower water content). A possibility that needs to be further investigated. Peak shear strengths () versus matric suction at net normal stress of 150 kPa are plotted as shown in Figure 5. The interface shear strength of the specimen undergoing wetting after drying (DW) seems higher than that undergoing only drying (D). From Figure 5, it appears that a non-linear representation of peak shear strength with respect to suction is appropriate for (D) tests of the rough interface; however more tests at higher suction are required to better investigate this behavior, especially for (DW) tests. 5 CONCLUSIONS
This study was carried out to investigate the effect of suction hysteresis on the shearing behavior of unsaturated rough interfaces. To this end, a series of suction-controlled direct shear tests were first conducted on the unsaturated rough interfaces following the drying path (D) at desired suction. Other tests were conducted at the same suction but following the wetting curve after drying (DW).
286
The results showed that the interface peak and post peak shear strength of the wetting (DW) tests were slightly higher than those of the drying (D) tests. A behavior that has been observed by some researchers but also contradicts some others as discussed previously. Thus, further research is needed to better understand this behavior. In addition, the peak shear strengths of wetting (DW) tests were mobilized at smaller displacement than that of the drying. During shearing, both tests exhibited a dilation behavior, and then dilation started to decrease once the strain softening process was completed. The interface peak shear strength increased with suction and showed a slight non-linear behavior at higher suction values. REFERENCES
Galage, C.P.K., and Uchimura, T., (2006). Effects of Wetting and Drying on the Unsaturated Shear Strength of a Silty Sand Under Low Suction, Proceedings of the Fourth International Conference of Unsaturated Soils, ASCE, Carefree, Arizona, pp. 12471258. Han, K.K., Rahardjo, H., and Broms, B.B., (1995). Effect of hysteresis on the shear strength of a residual soil, Proceeding of the first international conference on unsaturated soils, Paris, 2, 1995, pp. 499504.
Hatami, K., Khoury, C.N., and Miller, G.A., (2008). Suction-Controlled Testing of Soil-Geotextile Interfaces, The First Pan American Geosynthetics Conference & Exhibition, Cancun, Mexico. Khoury, C.N., Miller, G.A., and Kianoosh, H., (2010). Shear Strength of Unsaturated Soil-Geotextile Interfaces, GeoFlorida, Advances in Analysis, Modeling and Design Conference, West Palm Beach, Florida, in press. Miller, G.A., and Hamid, T.B., (2007). Interface Direct Shear Testing of Unsaturated Soil, Geotechnical Testing Journal, ASTM, Vol. 30, No. 3. Nishimura, T., and Fredlund, D.G., (2002). Hysteresis Effects Resulting from Drying and Wetting under Relatively Dry Conditions, Proceeding of the third international conference on Unsaturated soils, Recife, Brazil, 2002, pp. 301305. Shemsu, K.A., Kiyama, S., Aoyama, S., and Kobayashi, A., (2005). Experimental Study of the Effect of Cyclic Suction Loading on Shearing Behavior of Collapsible Soils, Proceedings of the International Symposium on Advanced Experimental Unsaturated Soil Mechanics, Trento, Italy, 2005, pp. 235241. Thu, T.M., Rahardjo, H., and Leong, E.C., (2006). Effects of Hysteresis on Shear Strength Envelopes from Constant Water Content and Consolidated Drained Triaxial Tests, Proceedings of the Fourth International Conference of Unsaturated Soils, ASCE, Carefree, Arizona, 2006, pp. 12121222.
287
Critical state behaviors of unsaturated soil in the direct shear test under constant pressure
B.S. Kim, S. Kato & S. Shibuya
Graduate School of Engineering, Kobe University, Kobe, Japan
ABSTRACT: In this paper, we examined the direct shear test results of compacted weathered granite specimen for the critical state behavior. In the direct shear test, an opening was set between the upper and the lower shear boxes before the shear process. In the analysis, the principal stresses were decided based on several hypotheses for Mohrs stress circle, and Cam clay type stress parameters were defined. From the test results, the maximum volumetric compression point (M.C.P) line was defined as the critical state line. The volume change behaviors towards the critical state were also observed in comparison with the critical state line obtained by the triaxial test. Moreover, the existence of Hvorslev surface was verified by applying the equivalent stress ( pe ) from the normal consolidation line. Above all, these results show that the critical state behaviors can be recognized in direct shear test. 1 INTRODUCTION line was obtained from the consolidation test, and the triaxial test under drained condition was carried out to decide the critical state line (CSL) for saturated weathered granite specimen. In order to examine the critical state behavior for saturated and unsaturated soils, the mean net principal stress (p* net), the deviator stress (q) derived from the principal stresses, and specific void ratio (v) were used. The obtained results of stress path and deformation were compared with CSL for triaxial test. In addition, the existence of the Hvorslev surface was verified by applying the equivalent stress ( pe ) from the normal consolidation line. 2 TEST PROCEDURE AND HYPOTHESIS FOR PRINCIPAL STRESSES
Casagrande (1938) firstly pointed out the existence of critical state through the direct shear test (DST) results for sand. After that, many studies for a critical state have been made with using the triaxial test apparatus which provides more precious stress state and more accurate volume measurement. Roscoe et al. (1958) and Scofield & Wroth (1968) proposed Cam clay model based on the critical state behavior, which has become a theoretical base to understand mechanical behaviors of soils. In order to examine similar critical state behaviors with the triaxial test results in DST, it should be analyzed with the same stress parameters, (p, q) as Cam clay model. These stress parameters are defined based on the principal stresses which are not decided usually for DST. Therefore, the study for critical state similar to the triaxial test result has not been made yet for DST. Kim et al. (2010) carried out DST under constant pressure condition with a weathered granite specimen for saturated and unsaturated soils. In their analysis, the suction stress obtained from SSM (Suction stress-SWRC Method) was applied. They reported that the suction stress can explain the relationship of shear strength between saturated and unsaturated soils. This study verified the applicability of principal stress parameters, and examined the critical state behavior using the DST results obtained by Kim et al. (2010). Here, in order to decide the principal stresses, several hypotheses for Mohrs stress circle were proposed. Moreover, the normal consolidation
The weathered granite soil that distributes widely in the western part of Japan was used. The soil particle density of sample is s 2.652 g/cm3, and the plasticity limit is NP. Specimens have the initial water content of about 7% and the dry density of about 88% for the maximum dry density (1.931 g/cm3). The size of specimen in DST is 60 mm in diameter and 20 mm in height. The direct shear test apparatus used in this study was a modified type for unsaturated soil (see Kato et al., 2010). After setting up the specimen, the lower and the upper shear boxes was sealed with 28 sheets of Teflon (0.1 mm per one sheet) which has low friction resistance. Accordingly, during the shear process, the Teflon sheets keep up the specimen side without the loss of the sample from the opening.
289
After attaining saturation of the specimen, it was consolidated under the specified overburden pressure. For unsaturated soil, the air pressure (ua 50 kPa and 100 kPa) was additionally applied inside of the chamber after the consolidation process. An opening was set between the upper and the lower shear boxes before the shear process. The shear process was carried out with horizontal displacement rates of 3.33 103 mm/s and 3.33 104 mm/s for saturated and unsaturated soils, respectively. Table 1 shows the comparison between ordinary and proposed standards for the principal stresses in DST. Davis (1968) assumed that the condition of the shear stress on the shear plane has a geometrical relationship as shown in Fig. 1. If this assumption is applied to the result for unsaturated soil, the apparent cohesion (c) induced by the effect of suction can not be considered, and Daviss assumption can not be directly applied to the result for unsaturated soil. Moreover, because a dilatancy equals to zero under constant volume condition in DST, the Daviss assumption shows that the peak point of Mohr circle expresses as the stress state for the shear plane. Although the direction of the principal stress is fixed, this condition does not correspond to the stress state during the shear process in constant volume test. Therefore, in this study, the suction stress (ps) obtained from SSM is applied as the intercept for horizontal axis as shown in Fig. 2. This intercept is connected with a datum line that expresses the stress state on shear plane. Here, because the suction stress equals to zero for saturated soil, this condition agrees with the Daviss assumptions mentioned above.
Table 1. Standards of the principal stresses in direct shear test. Existing standard Proposed standard in this study Suction stress, ps ds TC and 0
3
1 ds
vh
Pp
Mohr circle
3f
V = 45 /2
1f
v
Figure 1. Relationship between and ds by Daviss hypothesis.
Shear stress, T (kPa)
ds
Unsaturated condition
ff
C B
c Saturated condition T ff-unsat ds T ff-sat 1f-unsat
O'
ps
c=0
3f-unsat 3f
A(1i) M
M'
Net normal stress, v ua(kPa)
1f
1f
Figure 2. Mohr circles for unsaturated and saturated soils decided by a standard of the suction stress (ps) from SSM.
Contents Intercept of x axis
The origin (Daviss hypothesis) The angle of friction ds at the peak stress (Daviss point hypothesis) The angle which Dilatancy, consists with a (Daviss certain stress point hypothesis) The angle of the major 0 principal stress relative to the vertical direction, The parameter, b 0 (e.g., b ( b 0.2 0.3 2 3)/( 1 3) in TSS, PSC)
b0
The results obtained using Toyoura sand by Kim et al. (2009) were compared with past studies as shown in Fig. 3. The DST results shows nearly similar angle of internal friction in comparison with the triaxial compression test (TC), torsional simple shear (TSS) test and plane strain compression (PSC) test. It should be noted that DST results was obtained through a new method which keeps the sample loss to a minimum by using Teflon sheet against the opening between the shear boxes. Let us examine a notion of the position of Mohr stress circle based on the datum line. The Daviss assumption sets the position of Mohr circle by applying the dilatancy based on the coaxiality between the principal axis of strain and stress. The experimental background for the Daviss assumption is that the angle of internal friction for DST,
290
50 TSS 45
Isotropic consolidation : ' = 50 kPa , PSC, =0o , TC, =0o : DST d=0.2mm : DST d=0.5mm : DST d=1.0mm : DST d=2.0mm : DST d=3.0mm : DST d=4.0mm
40
ds
35
PSC : =45o~50o Ko - Consolidation test : DST v=200kPa
Ds
30
(Pradhan et al, 1995) : DST v=5kPa

Ds
: TSS v '=98kPa = tan (v')max (TSS)

-1
(Kitazima et al, 1995)
25 0.60
0.65
0.70
0.75
0.80
0.85
0.90
eo
Figure 3. Comparison of the results of DST test on past results (Toyoura sand) (Modified Pradhan et al., 1988b).
1.00
TSS, '=36~45kPa e=0.659~0.708 e=0.755~0.847
Figure 5. Three dimensional plots of according to each parameter (Toyoura sand) (After Lam et al., 1988).
0.95
0.90
0.85
TC, '=100kPa
= 0o
e=0.70 e=0.80
0.80
PSC, '=5~400kPa e=0.666~0.714 e=0770~0.839
=45~50o 60 40 20
0.75
80
()
Figure 4. R / (f 0) versus f (Toyoura sand) (Modified Tatsuoka et al., 1990).
ds becomes smaller than the angle of internal friction for TC, TC. However, the results shown in Fig. 3 show ds equals to TC. This means that the Daviss assumption is not correspond to the stress state in this study, and that the Mohr stress circle contacts to the datum line. Figure 4 shows the angle of internal friction of TC, TSS and PSC according to value (here, means the direction angle of the maximum principal stress for the direction of depositional surface; Tatsuoka et al., 1990). The angle of internal friction of TC at 0 becomes nearly the same value compared to TSS and PSC in range of 4550. If considering both this result and the result of Fig. 3, the stress state for DST in this study may correspond to 0 of TC or 4550 of TSS and PSC. On the other hand, let us examine the value of parameter b ( (2 3)/(1 3)) which relates to
the intermediate principal stress (2). The value of parameter b in the plane strain test (e.g., TSS) has been known in the range of 0.20.3. However, there is not any experimental evidence that the stress state in DST which uses the specimen of a round shape is identical with that of the plane strain test. Pradhan et al. (1988a) also reported that the difference between the intermediate principal stress (2) and the minimum principal stress (3) is very small in TSS using Toyoura sand. Figure 5 shows the variation of value for three parameters b, and . The parameter, equals to (90) related to the direction of depositional surface. The parameter means the direction angle of the intermediate principal stress for the direction of depositional surface. This figure shows that as the parameter b increases, value increases regardless of value. Accordingly, in order to explain the results of DST related to the results of TC in Fig. 3, the parameter b should nearly equal to zero, b0. Therefore, it is assumed in this study that the parameter b equals zero, which means that DST is similar to the stress state of TC, 0. 3 CRITICAL STATE BEHAVIORS IN (p* net, q) SPACE
(f) (f=0,PSC)
It is well known that the specimen in TC becomes homogeneous state up to the peak shear stress point and then non-homogeneous up to the residual state. The stress state at the Maximum Volumetric Compression Point (M.C.P) relates to the residual state and then to the critical state. On the other hand, DST can be defined
291
100
160
Saturated state
v =5kPa
Mpeak
140 120
On the way to residual state, not filnal state.
MM.C.P line
80
v=10kPa v=15kPa v =20kPa v=30kPa v=40kPa
Msat(MCP).
100 80 60 40
60
40
l ica t i r
st
e at
e lin
20
20
0
Saturated state ua = 50kPa ua = 100kPa 0 20 40 60 80 100
10
20
30
40
50
60
200
(a) Saturated soil

Unsaturated state ua = 50kPa
p*net
p*net
Figure 7. Comparison of the stress states for saturated and unsaturated soils around the end of shear process.
Mp(sat.)
150
v =10kPa v =20kPa(1) v=20kPa(2) v=30kPa v =35kPa v =40kPa
100
v=50kPa(1) v=50kPa(2)
Msat(MCP).
50
20
40
60
80
100
p*net
200
(b) Unsaturated soil (ua=50kPa)

v=10kPa v=20kPa(1) v =20kPa(2) v=30kPa v =40kPa(1) v =40kPa(2) v =50kPa
150
Mp(sat.)
Msat(MCP).
as a semi-homogeneous state up to the peak shear stress point because the deformation of a specimen is very small by this point, and then an un-homogeneous up to the residual state. The stress-displacement behavior in DST corresponds to the stress-strain behavior in TC in region of semi- to non-homogeneous states. Both of nonhomogeneous and residual state in DST and TC relate to the critical state. As stated previous section, the principal stresses in DST can be decided based on the several hypotheses shown in Table 1. After this, the mean net stress defined with these principal stresses is expressed as p* net. Figure 6(a) shows stress paths for saturated soil in (p* net, q) space, which curve gently to right direction. The solid lines represent M.C.P line and the peak strength line, respectively. The M.C.P line is defined as the critical state line (CSL) for saturated soil, and given as follows.
* * q M pnet 1.466 pnet
100
(1)
50
20
40
60
80
100
p*net
(c) Unsaturated soil (ua=100kPa)

Figure 6. Stress paths for saturated and unsaturated soil in (p*net, q) space.
Figures 6(b) and 6(c) show the results of unsaturated soil with the peak strength line and M.C.P line for saturated soil. The stress paths for unsaturated soil move up over M.C.P line. Figure 7 shows the comparison of the stress states between saturated and unsaturated soils around the end of shear process. The solid line represents M.C.P line, that is, CSL. The stress states do not correspond to M.C.P line, but it is presumed that they will move towards M.C.P line when the shear process progresses further.
292
COMPARISON OF CRITICAL STATE BEHAVIORS WITH TRIAXIAL TEST RESULT
1.7
CSL (TC)
TC (CD) 1.6
TC (CD) c '=19.62kPa c TC (CD) '=49.05kPa cc TC (CD) cc'=78.48kPa DST(CD) '= 5kPa vv DST(CD) '= 10kPa vv DST(CD) v v'= 15kPa DST(CD) '= 20kPa vv DST(CD) vv '= 30kPa DST(CD) vv '= 40kPa
In order to decide the critical state line, the triaxial test under drained condition was carried out for saturated soil. The shear velocity of 0.1 mm/min was used. Table 2 shows the initial states of specimens. Figure 8 shows the obtained stress-strain relationships. The test results except for the case of the confining pressure of 19.05 kPa show a contractive behavior even when the axial strain exceeds 15%. Figure 9 shows the comparison of the stress paths between TC and DST. Here, the triangle
Table 2. Initial states of specimens for the triaxial test. No. c (kPa) Water content, Dry density, Initial void w (%) d (g/cm3) ratio, e 6.73 6.83 7.37 1.702 1.700 1.700 0.565 0.567 0.569
1.5
n a ti o oli d 8 ons a l c l=0.09 rm No
1.4 Results of direct shear test for saturated state (Sr=100%; H = 2cm) 1.3 0 2 4 6
l i ne
10
ln(p*net)
(a) Saturated soil

1.7
TC1 19.62 TC2 49.05 TC3 78.48
CSL (TC)
DST(CD) v v'= 10kPa
TC (CD) 1.6
DST(CD)
DST(CD) v v'= 20kPa(2) DST(CD) v v'= 30kPa DST(CD) vv'= 35kPa DST(CD) DST(CD) v v'= 50kPa(1) DST(CD) v v'= 50kPa(2)
vv'= 20kPa(1)
vv'= 40kPa
250
Consoildated drained test, CD
1 0
TC (CD) TC (CD)
TC 1 : c=19.62 kPa
200
TC 3 : c=78.48 kPa Volume change rate, v (%)
1.5
-1
Deviator stress, v (%)
150
TC (CD)
cc '=19.62kPa cc '=49.05kPa cc'=78.48kPa
n a ti o ol id ons 8 al c l=0.09 rm No
TC 2 : c =49.05 kPa
-2 -3 -4 -5 Consoildated drained test, CD -6 TC 2 : c=49.05 kPa
100 TC 1 : c =19.62 kPa
1.4 Results of direct shear test for unsaturated state (u a = 50kPa; H=2cm) 1.3 0 2 4 6
50
TC 3 : c=78.48 kPa
li ne
10
15
20
25
10
15
20
25
Shear strain, a (%)
Shear strain, a (%)
(a) a vs. v
(b) a vs. v
10
ln(p*net)
Figure 8.
300
Stress-strain relationships for triaxial test.

1.7
CSL (TC)
DST(CD)
Saturated state
250
v =5kPa v =10kPa v =15kPa v =20kPa v=30kPa v =40kPa TC : c =49.05kPa
Mpeak
TC : c=78.48kPa
TC (CD) 1.6
DST(CD) vv'= 20kPa(1) DST(CD) v v'= 20kPa(2) DST(CD) DST(CD) DST(CD) TC (CD) TC (CD)
'= 10kPa vv
vv'= 30kPa
Msat(MCP).
DST(CD) v v'= 40kPa(2)
vv'= 40kPa(1)
200
vv'= 50kPa cc'=19.62kPa cc'=49.05kPa TC (CD) cc'=78.48kPa
1.5
150
n a ti o ol id 8 ons a l c l=0.09 rm No
TC : c =19.62kPa
100
M.C.P line : Triaxial test
1.4 Results of direct shear test for unsaturated state (u a = 100kPa; H=2cm)
l ine
50 M.C.P of Triaxial test 0 0 50 100 150 200

1.3 0
10
p*net

Figure 10. Comparison of the critical state line between TC and DST.
ln(p*net)
Figure 9.
Comparison of M.C.P lines of TC and DST.
293
v k(s) v
Cr
No
iti ca ls
rm
al
Co
ta
ns
te
ol
M
id a tio n lin e
at
ric
su
ct
io
u ,(
a-
) uw
1.0 Saturated state

=5kPa =5kPa v v A : v=20kPa (d=1.80g/cm3) B : v=20kPa (d=1.90g/cm3)
0.8
=10kPa v=10kPa v =15kPa v=15kPa v

v=20kPa v=20kPa v=30kPa v=30kPa
lin
(C
SL
):
r=
2D
10 0%
(N CL )
MM.C.P line
No Cr iti ca
rm
al
CS
Co ns ol
-U
ln(p), ln(p*net)
ls
ta
te
id
lin
at
io
ur at
ed
(C
lin
SL
s1 sta t
q / pe'
ns
at
s2
s3
0.6
v=40kPa v=40kPa
(N
CL
3D
r=
0.4
10
A B
o Hv
v sl e
sur
e fac
ln(p)
, ln(p *ne )
t
0.2
Figure 11. Sketch of critical state line of DST in 3-dimensional space.
0.0 0.0
0.1
0.2
0.3
0.4
0.5
p*net / pe'
symbol () represents the stress state at M.C.P of TC. In the case of the confining pressure of 49.05 kPa, because the stress-strain relationship did not reach to M.C.P line, it was estimated from the stress-strain relationship. From this result, it was found that the stress state at M.C.P in TC corresponds to M.C.P line of DST. In addition, the normal consolidation line (NCL) was obtained from the oedometer test. Above all, Fig. 10(a) shows the comparison of the behaviors of void ratio between DST and TC for saturated soil in (ln p* net, v) space. The CSL defined by TC has a parallel relationship with NCL. In this figure, though the results of DST show a little dilative behavior after an initial compression, they do not reach to CSL. The reason is presumed that non-uniformity of the deformation develops during the shear process because of no cohesion between soil particles, and this condition induces a deformation behavior which is more brittle than the unsaturated soil. Figures 10(b) and 10(c) show similar test results for the cases of ua 50 kPa and 100 kPa, respectively. In these figures, each stress paths attain to CSL of TC. Accordingly, it is found that CSL of DST corresponds to that of TC. This result means the existence of CSL in DST. Figure 11 represents a sketch of the critical states obtained in (ln p* net, v, s) space. After all, in the two dimensional space, it was defined that CSL exists uniquely for saturated and unsaturated soils. However, it exists as an expanded plane according to matric suction axis in the three dimensional space. 5 STUDY FOR EXISTENCE OF HVORSLEV SURFACE
(a) Saturated soil

3.0 2.5

v=10kPa v=20kPa(1) v=20kPa(2) v=30kPa v=35kPa v=40kPa v=50kPa(1) v=50kPa(2)
2.0
q / p e'
1.5
or Hv
sle
fa ur vs
ce
M sat(MCP).
1.0
0.5
0.0 0.0
0.2
0.4
0.6
0.8
1.0
1.2
p*net / pe'
3.0

=10kPa v v=10kPa =20kPa(1) v v=20kPa(1) =20kPa(2) v v=20kPa(2) =30kPa v v=30kPa =40kPa(1) =40kPa(1) v v =40kPa(2) =40kPa(2) v v =50kPa =50kPa v v
2.5
2.0
q / pe'
1.5
or Hv
e fac ur s v sle
Msat.(MCP)
1.0
0.5
0.0 0.0
0.2
0.4
0.6
0.8
1.0
1.2

Figure 12. Existence of the Hvorslev surface for saturated and unsaturated soils in DST.
p*net / pe'
The Hvorslev surface for the normalized stress paths was examined through applying the
294
equivalent stress (pe ) because the specimen consists in the over-consolidated state in DST. Figure 12(a) shows the normalized stress paths for saturated soil. Stress paths move up with the shear process, and when they reached the dotted line in this figure, they move along this line. Therefore, this dotted line can be regarded as a Hvorslev surface. In addition, the test results for different initial dry densities of 1.8 g/cm3 and 1.9 g/cm3 were plotted, and they also show similar behaviors. From this result, it can be understood that the Hvorslev surface for saturated soil exists uniquely regardless of the initial dry density of specimen. Figures 12(b) and 12(c) show similar test results for unsaturated soil. In these figures, the Hvorslev surface which is deduced from the stress paths was plotted as the dotted line. In the case of unsaturated soil, stress paths move up with the shear process. When stress paths reached to the dotted lines, they move along fixed lines like a previous behavior for saturated soil. Thus, these fixed lines also can be considered as the Hvorslev surfaces. When comparing these Hvorslev surfaces, it is found that they may have a different intercept for the vertical axis according to the suction stress obtained from SSM. 6 SUMMARIES
3. From comparisons of the stress paths in DST for saturated and unsaturated soils and CSL of TC in (ln p* net, v) space, it was found that CSL in DST exists. 4. The existence of the Hvorslev surface was examined using the normalized stress paths, which is given by applying the equivalent stress (pe ). When comparing the obtained Hvorslev surfaces, it was found that they may have different intercepts for the vertical axis according to the suction stress obtained from SSM. REFERENCES
Casagrande, A., 1938. Compaction test and critical density investigation of cohesionless materials for Franklin Falls dam. U.S. Engineer Corps, Boston District. Davis, E.H., 1968. Theories of plasticity and failures of soil masses. Soil Mechanics, selected topics (ed. I.K. Lee), London: Butterworth. Kato, S., Kim, B.S. & Shibuya, S., 2010. Direct shear test under constant volume state for unsaturated soil. Proceeding of 5th International Conference on Unsaturated Soils, 68 September, Barcelona, Spain. (under contribution). Kim, B.S., Kato, S. & Shibuya, S., 2009. Effect of the opening between shear boxes on the shear strength in direct shear test. Memoirs of Construction Engineering Research Institute Foundation, No. 51, pp. 99108. (in Japanese). Kim, B.S., Shibuya, S., Park, S.W. and Kato, S., 2010. Application of suction stress for estimating unsaturated shear strength of soils using direct shear testing under low confining pressure. Canadian Geotechnical Journal. (in press). Pradhan, T.B.S., Tatsuoka, F. & Horii, N., 1988a. Simple shear testing on sand in a torsional shear apparatus. Soils and Foundations, 28(2): 95112. Pradhan, T.B.S., Tatsuoka, F. & Horii, N., 1988b. Strength and deformation characteristics of sand in torsional simple shear. Soils and Foundations, 28(3): 131148. Roscoe, K.H., Schofield, A.N., & Wroth, C.P., 1958. On the yielding of soils. Geotechnique, 8(1): 2253. Schofield, A.N. & Wroth, C.P., 1968. Critical state soil mechanics. McGraw-Hill Book Co., London. Tatsuoka, F., Nakamura, S., Huang, C.C. & Tani, K., 1990. Strength anisotropy and shear band direction in plane strain tests of sand. Soils and Foundations, 30(1): 3554.
In order to examine the critical state behavior for the result under constant pressure condition in DST, the principal stresses were derived based on several hypotheses. In addition, the consolidation test and the triaxial test were carried out for saturated soil. Then CSL defined with TC results was compared with the results of DST. Lastly, the equivalent stress (pe ) was applied to the results of DST to normalize the stress path. These results could be summarized as follows. 1. The stress parameters of Cam clay type were obtained through several hypotheses in DST. It was found that M.C.P line corresponds to CSL in (p* net, q) space. The stress path for unsaturated soil moved over CSL, and reached to the peak stress state. 2. The triaxial test was carried out to decide CSL. It was found that the stress state of M.C.P in TC corresponds to M.C.P line of DST.
295
The decrease of the strength of unsaturated silty sand and the Bishops parameter
J.C. Leal, G. Gallegos & E. Rojas
Facultad de Ingeniera de la Universidad Autnoma de Quertaro, Mxico
ABSTRACT: In 1959 Bishop proposed an effective stress equation for unsaturated soils which generated a lengthy controversy. This equation has an important drawback; there is not a method or equation for the precise determination of the value of parameter . This paper presents the results of a series of suction controlled triaxial and simple compression tests made on silty sand. These tests and the retention curve were made for both the wetting and the drying paths. The triaxial tests results show that the strength of the soil increases to a maximum at certain suction and then decreases when suction is further increased. The strength results are presented in graphics suction-deviator stress and diagrams p-q. For the estimation of parameter , the equations proposed by some researchers have been used. The comparison between experimental and theoretical results shows that some of these relationships do not simulate the strength reduction experimentally observed on the silty sand. 1 INTRODUCTION concerning the value of the cohesive stress for a completely dry soil, and the difference of the cohesive stress at drying and wetting. In the case of clays with large specific surface, the physicochemical forces and the adsorbed water play an important role in its behavior. Is probably for these reasons that the phenomenon of strength reduction has not been experimentally observed on these materials. In that sense, it is necessary to carry out more experimental work on this issue to better understand this phenomenon and finally to improve the expressions related to the strength of these materials. The objective of this paper is to experimentally observe and measure the decrease of strength on unsaturated silty sand, and to compare the value of parameter with the equations proposed by Vanapalli, Khalili and Khabaz (1998) and berg and Sllfours for both trajectories of wetting and drying of silty sand (SM). 2 MATERIALS AND METHODS
Terzaghi, (1936) established the principle of effective stress for saturated soils. The effective stress controls the strength and the changes of volume of the soils. In establishing this principle Terzaghi considered that the solid particles and the water in the soil are incompressible and also that the contact area between two particles can be neglected. The equation of effective stress for saturated soils is written as uw (1)
where effective stress; total stress; and uw interstitial pore pressure. In 1959 Bishop proposed an effective stress equation for unsaturated soils generating a lengthy controversy which is still ongoing. This equation writes ua (ua uw) (2)
where effective stress; ua net stress; (ua uw) suction; and Bishops parameter. The product ( ua) is called here the cohesive stress. The main drawback of this equation is that there is not a method or equation for the precise determination of parameter . Some experimental results indicate that the strength of some soils reaches a maximum and then decreases with further increase in suction (Pereira et al., 2006). However, questions are raised
The soil was classified as silty sand. Its grain-size distribution is shown in figure 1. The coefficients of curvature and uniformity for this material are 1.47 and 7.50 respectively. It was determined that it does not present plasticity. These soil index properties are summarized in Table 1. It is noteworthy that this soil does not show any clay content. The soil was initially dried at room temperature (20C 1C) and all particles larger than
297
Silt Clay 100 90 80 70 60 50 40 30 20 10 0 1E-4 1E-3 0.01
Sand
Gravel
0.1
10
required water content, it was placed in a hermetic containers and their suction was measured using the filter paper method. The samples were stored in a controlled temperature room (20C 1C) according to the directions of the test method D 5298-9403 of the American Society for Testing and Materials (ASTM). The filter paper used in these tests was the Schleicher and Schuell No. 589. This paper was previously calibrated also according to the same standardized ASTM procedure. Other soil specimens were tested in a suction controlled triaxial apparatus to examine the effects of the drying and wetting on the strength of the material. 3 TESTING PROCEDURE: SUCTION CONTROLLED TRIAXIAL TESTS
Percentage passing (%)
Grain size (mm)

Figure 1. Grain-size distribution of soil tested. This soil principally contains silt and sand. Table 1. Index properties and grain sizes of soil. Description % Gravel % Sand % Silt Gs Classification* 0.00 79.00 21.00 2.43 SM
* Unified soil classification system.
2 mm (sieve No. 10) were removed. The remaining soil was divided in two equal parts. One half was washed using a No. 200 sieve; then both, the washed half and the unwashed one were mixed together. A quantity of water was added to obtain a 19.53% moisture content. The mixture was used to make soil specimens by static compaction. All specimens were compacted in five layers. The soil for each layer was weighed (0.70 N) and placed in a cylindrical mould. Then the layer was statically compacted at a maximum stress of 3140 kPa. With these samples an experimental program was used to define the retention curve in wetting and drying, as well as the strength of the soil at different suctions and for both the wetting and the drying path. In order to determine the retention curves, the following procedure was applied: two sets of compacted specimens of silty sand were prepared and all samples in the first set were dried at different water contents so as to obtain the drying path. For the wetting path, all samples of the second set were completely dried in an oven and then they were re-hydrated at different water contents by adding the required amount of water. The water was added by spraying it on the sample which was placed on a scale. Once the specimen absorbed the
All triaxial tests were performed on samples showing the following characteristics: 50 mm in diameter and 100 mm in height, a dry density of 14.89 kN/m3 and a gravimetric water content of 19.53%. Specimens used for the wetting and drying path were prepared in the same way as those used for the retention curves. During the tests, suction was kept constant by circulating an atmosphere of a saline solution of NaCl placed in a container outside the triaxial cell. The circulation of the air was maintained by using a peristaltic pump specifically designed for fluid transfer. The pumping rate was fixed at 15 rpm, corresponding to a flow rate of 8.81 ml/min through a polyvinyl chloride pipe with a internal diameter 3.16 mm. The pump circulated the relative humidity of the container from the bottom to the top of the specimen throughout the shearing process. Figure 2 shows a schematic view of the triaxial apparatus with the suction system. Simple compression tests were performed at the same deformation rate than the suctions controlled triaxial tests. To determine the angle of internal friction of the saturated soil a series of consolidated drained triaxiales test were carried out. The three applied confining stresses were 50, 100 and 150 kPa. To determine the water retention curves filter paper technique was used for the trajectories of wetting and drying. The determination of the retention curve in drying was also performed with the cylinder membrane extractor at suction values between 1.0 kPa to up to 4200 kPa. This device is a container in which the soil samples are placed in initial saturation conditions by immersion and then is pressurized by nitrogen. The nitrogen pressure in the cell is related to the suction applied to the soil. The cylinder membrane extractor is shown in figure 3.
298
900 850 800 750 700 q (KPa) 650 600 550 500 450 400 100000 1000 10
Drying trajectory Humidity trajectory
Peristaltic pump
Soil
Coarse disc
Saline solution NaCl
Net confining stress
0.1
0.001
s (KPa)
Figure 2.
Triaxial apparatus with controlled suction.
Figure 4. Tests triaxiales consolidated and drained (CD) results, residual deviator stress (q) against the suction (s) for both trajectories of drying and wetting.
90
Drying process Humidifying process
Pipe to liberate pressure Manometer Membrane Soil simples in container
80 70 60
q (kPa)
50 40 30 20
Graduated pipe
10
1 10 100 1000 10000 100000
s (kPa)
Figure 3. Cylinder extractor with membrane and tank of pressurized nitrogen. Figure 5. Results of tests of simple compression of a silty sand (SM).
RESULTS AND DISCUSSION
The re-molded samples of silty sand were mounted in the triaxial apparatus with a confining pressure of 150 kPa. These results are shown in figure 4. In this figure, the residual strength of each sample is plotted with its corresponding suction for both the wetting and the drying paths. For both trajectories a maximum strength can be observed. The results of the simple compression tests are included in figure 5. It can be also observed that for both paths a maximum value of resistance
exists. In both the triaxial and the simple compression tests, the largest strength value corresponds to the drying path. From the results of the saturated drained triaxial tests in saturated conditions for different confining pressures, a friction angle of 39.5 was obtained as shown in Figure 6. The way in which water is retained by the soil at different suctions is represented by the retention curves, in which the suction is plotted (s) versus the degree of saturation (Sw). The water retention curves for the trajectories of wetting and drying have been obtained, although the effect of the
299
400 300
(KPa)
Triaxial tests (CD) Saturated soil 1 (kPa) 3 (kPa) 50.0 201.2, 235.8 337.8, 341.4 100.0 501.3, 519.7 150.0
1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2
200 100 0 0 100 200 300 400 500 600 700

(KPa)
Figure 6. Results of triaxial tests (CD) of a silty sand saturated (SM).

100 95 90 85 80 75 70 65 60 55 50 45 40 35 30 25 20 15 10 5 0 0.01 Humidifying process-filter paper. Drying process-paper filter. Drying process-cilinder extractor.
0.1 0
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Sw
Sw(%)
Humidifying process and drying process, eq. (3), Vanapalli (1996). Humidifying process and drying process, eq. (4), Vanapalli (1996). Humidifying process and
1 100 s (kPa) 10000 1000000
drying process, eq. (5), berg and Sllfours (1995). Humidifying process and drying process, eq. (6), Khalili and Khabbaz (1998). Drying process experimental results.
Figure 7. Water retention curves for trajectories of wetting and drying of silty sand (SM).
changes of volume of pores of the soil has not been included. The obtained retention curves are showed in figure 7. The water retention curve in drying shows larger values of suction than the wetting curve. In these curves it can be observed that most water is drained from the pores of the soil at suctions of 1000 kPa. This means that the soil is mostly formed by large pores which need a minor quantity of energy to drain their water, while the small pores need more than 1,000,000 kPa to drain their water. This means that the pore size distribution is a fundamental factor in the shape of the retention curves.
Humidifying process experimental results.
Figure 8. Values of for trajectories of wetting and drying of silty sand (SM).
The theoretical effective stress of the soil at different suctions has been calculated using the values of the parameter proposed by different authors. For both the wetting and drying paths has been computed using the information of the retention curves and of the triaxial tests. The two first equations were proposed by Vanapalli et al. (1996), the third equation was proposed by berg and
300
Sllfours (1995) and finally, the fourth equation belongs to Khalili and Khabbaz (1998). Figure 8 contains the results of and the used equations of the following: (Sw) (Sw Sr)/(1 Sr) Sw (u u ) a w (ua uw )b
0.55
(3) (4) (5) (6)
In Figure 9 the resistance results are presented/ displayed p q, with calculated with the equations (3) to (7). The equation (7) includes the experimental values on which depends. In that figure it can be observed that the value of the effective stress does not remain. The values obtained from p' with the equations (3) to (6) are major. p ua s (7)
where Sw degree of saturation; Sr residual degree of saturation; (ua uw)b air entry value; and fitting parameter. The saturation degree (Sw) was obtained at the end of each triaxial test of controlled suction, the fitting parameter (k) was 1.6, the degree of residual saturation (Sr) and the air entry value was obtained from the curves of drying water retention and humidification of according to the case, the suction (ua uw) was obtained with the paper filter technique when finalizing the triaxial tests. In Figure 8 it can be observed that the results of the equations (3) to (5) of parameter versus saturation degree do not have a significant difference for the drying and humidification trajectories. However, the equation (6) shows a significant difference in its results. In this equation there are changes of curvature that the other equations do not present.
where p effective stress; ua net stress; s suction; and Bishops parameter. The product s is the cohesive stress. 5 CONCLUSIONS
It has been observed that the tested soil reaches a maximum strength for both the wetting and drying paths. This effect is observed in confined and unconfined tests. There are no equations that can predict the behaviour of the parameter and consequently of the shear strength of all soils. For this reason, not only more experiments are necessary, but also the development of theoretical models for unsaturated soils. REFERENCES
ASTM 2004. Section four Construction Volume 04.08 Soil and Rock (I). Designations: D420-D-5611. Fredlund, D.G. & Rahardjo, H. 1993, Soil Mechanics for Unsaturated Soils. New York. A Wiley-Interscience Publications, Inc. Garven, E.A. & Vanapilli, S.K. 2006. Evaluation of empirical procedures for predicting the shear strength of unsaturated soils. Proc. of Fourth International Conference of Unsaturated Soils, Arizona. Kumara, G.Ch.P. & Uchimura, T. 2006. Effects of Wetting and Drying on the Unsaturated Shear Strength of a Silty Sand Under Low Suction. Proc. of Fourth International Conference of Unsaturated Soils, Arizona. Mitchell, J.K. 1993. Fundamentals of Soil Behavior. John Wiley & Sons, Inc., USA. Murray, E.J. 2002. An equation of State for Unsaturated Soils. Canadian Geotechnical Journal, 39 (1). February 2002.00083674. Pereira, A., et al. 2006. A Study on the Shear Envelope of an Unsaturated Colluvium Soil. Proc. of Fourth International Conference of Unsaturated Soils, Arizona. Rojas, E. 2006. Equivalent Stress for Unsaturated soils. Proc. of Fourth International Conference of Unsaturated Soils, Arizona. Whitlow, R. 1999, Fundamentos de Mecnica de suelos. Mxico. Compaa Editorial Continental S.A. de C.V.
900 800 700 600 q (kPa) 500 400 300 200 100 0 0 200 400 p' (kPa)
Figure 9. Diagrams p-q theoretical and experimental results of silty sand (SM).
Eq. (3) Eq. (4) Eq. (5) Eq. (6) Experimental results 600 800 1000
301
Study on gas phase in unsaturated soils

Y. Liu, C.G. Zhao & Q. Fang
School of Civil Engineering, Beijing Jiaotong University, Beijing, China
ABSTRACT: The presence of air in soil makes its properties very complicated. Firstly, this paper listed the divisions of the air-water patterns by different researchers and showed that discontinuous gas exists in unsaturated soils. It plays an important role on the unsaturated soils, because discontinuous gas will change its pressure during soil deformation. Then it was demonstrated that the gas pressure was needed to be used as another stress variable to describe the soil behaviour based on the work expression with the help of thermodynamics and mixture theory. After that a detailed description of gas behaviour in unsaturated soils is given. In order to take account of the gas effect, the plastic change of gas strain is introduced as an internal state variable to reflect the influence of gas phases. In this way, the dissipation caused by gas phase can be included. A simple model for isotropic stress states are given at last, which could couple the three phases of unsaturated soils. 1 INTRODUCTION soils with compressible fluids. However, it is not the case when the gas size is very large. Terzaghi (1944) divided the gas bubbles into two group: gas bubbles which are small compared with the particle size and gas bubbles which are larger than the normal size. Klausner (1990) make an more detailed division, who dose not only consider the bubble size but also consider the bubble shape. For the large gas bubble condition, the change of gas pressure will strongly influence the soil behaviour, which should be treated carefully. This type of soils can be found in marine soils, which has been studied by Nageswaran (1983), Wheeler (1986), Thomas (1987), and so on. In either the internally continuous gas stage or the discontinuous gas with large gas bubble stage, gas is not connecting with outward atmosphere. Therefore, the gas pressure can not remain constant, and its variation will of course affect the soil behaviour. However, most of the present researches are assuming that air remains at atmospheric or constant pressure. This assumption is acceptable in many cases at low moisture content where the gas phase is thought continuous, but can not be used for marine soils or simulating special oil recovery and remediation processes that have recourse to air injection. Loret & Khalili (2000) proposed a threephase model for soils, where air pressure in their paper is a variable, so that the discontinuous gas situation can be included. This paper will first give the work expression for unsaturated soils based on the thermodynamical method to interpret that gas pressure should be treated as a third stress variable. Then gas effect is considered in soils and a constitutive relationship
Unsaturated soils with gas in pore space makes it much more complicated than saturated soils. The existence of gas influence the compressibility of soils, and cause suction make the pressure between gas and water difference. In this case, the pattern of the gas phase in the soils is very important for unsaturated soils. In general, as the degree of saturation increase, there are three different types of unsaturated soils as shown by Wroth & Houlsby (1985), that are continuous gas discontinuous water stage, continuous gas and water stage, and discontinuous gas continuous water stage. Bao et al. (1998) pointed that whether the gas is continuous has great effect on the soil behavior. He divided the type of continuous gas continuous water into gas partially continuous stage and internally continuous stage, which has been verified through suction and air permeability tests. When the moisture is very small, water could only occupy the small voids. As the moisture increased, the water gradually occupy the small passageways and become continuous, this is the gas partially continuous stage. When the moisture is gathered sill more, the water will occupy lager passageways and seal off the boundary voids of soil, render the gas dislinked with the atmosphere, but inside the voids they remain to be interconnectability. This is the internally continuous stage, where the degree of saturation is about 0.85 to 0.9. These two stages are important and difficult for the research of unsaturated soils. Traditionally, when the gas is completely discontinuous, the soils are usually treated as saturated
303
for unsaturated soils included the gas effect is given. At last a constitutive model for isotropic stress states is developed. 2 WORK EXPRESSION FOR UNSATURATED SOILS
matric suction s* ns and the gas pressure ua .The work conjugate strain variables are skeleton strain ij, the degree of saturation Sr and the gas strain a respectively. The gas strain is expressed as follows: d a n(1 Sr )d a / a dVa /V (3)
At present, most of the unsaturated soils models are adopting two independent stress variables (net stress and matric suction) or the effective stress as the stress variables to describe the soil behaviour. If the air pressure is assumed to be constant, the net stress will be simplified into total stress, and the matric suction will be simplified into negative water pressure. In fact, however the air will not always connect with the atmosphere completely, and they maybe occluded and will not remain as atmosphere pressure as showed in the first section. Wheeler (1988) studied on the marine soils containing large gas bubbles. He found that the existence of gas play an important role on the soil and the concept of two independent stress variables may not be applicable to soils where either the water or the gas is discontinuous. Therefore, a more reasonable choice of stress variables should be given. The work expression of unsaturated soils provides a theoretical basis. Based on the balance equations of three phases in soils, making use of the average method and the mixture theory, together with the second of thermodynamics, Zhao et al. (2009) gave an the work expression for unsaturated soil. The interactions between three phases are ignored. This expression is the same as the expression given by Houlsby (1997) but with a more strictly theoretical basis. It could be expressed as follows:
* dW ij d ij nsdSr n 1 Sr ua d ln a
where V is the soil volume and Va is the gas volume. Wheeler (2003) used the modified suction and generalized effective stress as two stress variables to propose a coupled model for unsaturated soils. They interpreted the elastic-plastic processes in unsaturated soils and the coupling movement of yield surfaces and developed a specific constitutive model for isotropic stress state. Their mode could reflect the coupled influence between the plastic change of degree of saturation and the plastic volumetric strains, but the influence of variation of gas pressure was not included. The work expression (1) provided theoretical support that the gas pressure should be taken into account in the model of unsaturated soils. The model proposed by Loret & Khalili (2000) is a good example, in which air pressure is a variable instead of the atmospheric pressure. As a result, the discontinuous gas situation can then be considered. 3 3.1 DESCRIPTION OF GAS PHASE IN UNSATURATED SOILS Behaviour of gas phase
(1)
where ij is the strain tensor; Sr is the degree of saturation; n is the porosity; ua and uw are pore air pressure and pore water pressure, respectively; s ua uw is the matric suction; a is the density of air; and ij* is the generalized effective stress or average soil skeleton stress, which is expressed as:
* ij ij ua ij Sr ua uw ij
(2)
It is shown in equation (1) that the work input of unsaturated soils contains three parts: the first part is the power input by the deformation of soil skeleton, the second part is due to the change of water content, and the third part is caused by the compression of gas phase. As a result, three stress variables need to be used to describe the soil behaviour, i.e., the generalized effective stress ij*, the modified
It has been shown from the above discussion that the gas pressure as well as the generalized effective stress and modified suction are all needed to describe the soil behaviour. Since the last two variables and their corresponding strains have been elaborated by Wheeler et al. (2003), in this section we will pay attention to the behaviour of gas phase. When the gas phase is occluded, its pressure may be changing under isothermal condition for the following reasons: variation of the gas mass, change of the applied load and flow of the fluids. The mass of bubble is likely to increase due to biogenic action or diffusion. The gas pressure will increase with the gas mass. The external loading on soil will result in skeleton deformation, which will change the volume of pore space that gas exists with a consequent change in gas pressure. Any drainage of fluids in soils will also cause deformation of soil matrix and result in deformation of bubble space. Therefore, any deformation of the soil skeleton or the change of moisture content will both cause variation of gas pressure. Conversely, any change of gas volume or pressure will affect the behaviour
304
of soil matrix and the water phase. At the beginning the gas pressure in soil are in equilibrium with the external load. If the external pressure increases and become larger than the gas pressure, the gas space will constrict. The gas volume will continue shrinking until the gas pressure and the external pressure reach a new balance. If the variation is small, it can be recoverable. But if the variation exceeds a critical value, plastic deformation will come out and result in rearrangement of soil structure, and thereby cause plastic change of soil matrix and degree of saturation. In addition, even if the external pressure keeps constant, the soil behaviour could be influenced by the flow of fluids. As we known, plastic change of degree of saturation will occur during the wetting-drying process, which also showed that there is plastic change of gas content. The change is influenced by suction. When suction reach the air entry value, gas will come into the saturated soils, the water filled in the void space will now be replace by gas. Conversely, if the suction reaches a critical value which make the gas pressure smaller than the surface tension, the water will breakthrough into the gas void. In conclusion, plastic change of gas volume may occur in unsaturated soils, which may affect the behaviour of the other two phases. 3.2 Description of the deformation of gas phase
variables into the yield function. This will be shown in the next section. Li (2007) gave a thermodynamics-based constitutive framework for unsaturated soil. However, the behaviour of gas phase in his model is thought to be reversible, and thereby no dissipation. If the gas effect is introduced, the free energy for gas phase can be expressed as: d a n 1 Sr ua d ln a aa (7)
where aa represent the incremental dissipation in pore gas. According to Li (2007), a reflect the evolution of the internal variable of gas phase, which must be mathematically homogeneous with the evolution of the internal rearrangements of gas phase Ha. the thermodynamic force is still function of the state and the pattern of structural rearrangement, a a (a,Ha). Using Legendre transform, the Gibbs function for gas can be obtained: d a a dua aa (8)
* where, a a (ua , Ha ) is the thermodynamic force and is a function of stress and internal variables. Based equation (7), we can get gas strain and its elastic and plastic increment respectively as follows:
In order to include the gas effect into the constitutive model in unsaturated soil, we must take into account of the dissipation caused by gas. Li (2007) pointed that the dissipation in a soil is attributable to its internal structural rearrangements, and can be decomposed to two parts, Hs and Hf, denoting the dissipation in connection with solid phase and pore fluids. If we consider the gas dissipation, the dissipation will have three parts and can be expressed as:
* ,a H s H s H s , H s H s , s* , Sr , H s H s , ua * , ij Hw Hw Hw , Hw Hw , ij * , ij Ha Ha Ha , Ha H a , ij
a
e a
p a
u u , H du M du u u , H H u , H
a ua , Ha
a 2 a a a 2 a a a a a a a a a a a
(9)
(10) (11)
(4) (5) (6)
where Ma is elastic parameter. The plastic change of gas strain is due to the internal structural rearrangement. 4 CONSTITUTIVE MODEL FOR ISOTROPIC STRESS STATES
where Hi, (i s, w, a) denote the internal structural rearrangements of the i phase; and Hs , Hs , Hw , Ha the pattern of and the interactive structural rearrangements. In practice, we would like to use the plastic deformation, such as the plastic strain, the plastic change of degree of saturation and the plastic gas strain, as the internal state variables. The coupled effect can be considered by taking account these
In this section we would like to give a simple coupled constitutive model for isotropic stress states. Since the gas phase will influence the soil behaviour especially when they are occluded, it must be reflected in the constitutive model as well as the liquid phase. According to Wheeler et al. (2003), the yield surfaces for the three phases are given respectively by:
305
* p* p0 * s* s ,( I , D )
(12) (13) (14)
ua uay
liquid phase, it is not a constant, but a variable that influenced by the other two phases. The flow rules are similar to Wheeler et al. (2003). If the coupled effect is considered, then the hardening rule for the three phases can be written as:
* dp0 ksw ksa v d vp dS p d p * p0 w w r a a a (22) * ds * s
* , * , * define the current yield stress; where p0 s uay I means suction increment and D means suction decrease. The elastic and plastic strain increments for solid and liquid phase are given by Wheeler et al. (2003) as follows:
d ve d vp
dp* v p*

kwsv
d vp
w w
dSrp
(23)
(15) duay
*
dp v p*
w ds* s*
(16)
uay
kasv 1 d vp d p a a a
(24)
dSre
(17)
dSrp
w w ds* s*
(18)
where , w are elastic constants for solid and liquid phase respectively. The deformation of gas, like the other two phases, can be decomposed into two parts: an elastic part and a plastic part. When the gas pressure lies in a certain ranges, the deformation is recoverable. We assume that the gas deformation will not affected by the external pressure and pore water pressure in this stage and can be expressed as:
e d a
where ksw, ksa, kws, kas are all coupled parameters. There are some restrictions on these parameters. If the degree of saturation reaches 1.0, then the soil become saturated, and the coupling effect can be neglected. In this case, the parameters are all equal to 0. In addition, if the degree of saturation is very low (for examples, Sr 0.85), the gas phase will be continuous. Then, the gas pressure may become constant and its influence on solid can be neglected, meaning that ksa kas 0. Equation (22)-(24) can be combined to give the expression for the plastic strain increment: d vp
* dp0 k
* v p0
sw
* ds * s
ksa
duay uay
(25)
dSrp
a dua ua
(19)
where a is an elastic constant for gas. If the gas pressure exceeds a limited value, plastic change of gas strain will take place. Then its own plastic change can be given by:
p d a
duay ds* dp* kws *0 1 ksa kas * kws ksa p0 s uay (26)
w w
p da
a a dua
ua
(20)
Since the gas phase must satisfy the Boyle law: d a n 1 Sr dua ua (21)
duay ds* dp* kas *0 kas ksw * 1 ksw kws p0 s uay (27) (28)
a a
1 ksw kws ksa kas
Based on equation (19), (20) and (21), we can deduce that a n(1 Sr). Unlike the solid and
Combined with equation (15), (17), (19), the incremental relationship between stress and strain can be obtained:
306
d v 11 12 dS r 21 22 d a 31 32 where
* 13 dp * 23 ds * 33 dua
(29)
(see figure 2 (a)), which slow down the deformation. At the end of our simulation, the total change of specific volume and the degree of saturation are close to that of Wheeler because of the small gas volume content.
/ vp* w kws w p* a a kas p*
ksw
vs
*
w 1 kas ksa * * s s
a a kas ksw

s*
w w kws ksa * ua a 1 kws ksw * * ua ua

* vua
ksa
(30)
In experiment, the stress variables are usually use the total stress, pore water stress and the pore gas stress. So we need to change the stress variables we used to the normal controllable variables in practice.There are ten soil constants in the model: , w, a, , w, a, ksw, ksa, kws and kas. The model could reflect influence of gas on the soil deformation, and couples the three phases in unsaturated soils. But too many soil contents make it difficult to use. In order to illustrate the model, some numerical examples are given. Consider a sample of unsaturated soil (Wheeler et al. 2003) has the following soil constants: 0.15, 0.02, w 0.12, w 0.02, a 0.005. The coupled parameters are as follows: ksw 0.7, ksa 0.15, kws 0.8 and kas 0. Figure 1 shows the simulation at constant suction of 200 kPa. The gas pressure can not be controlled in our model and is calculated by equation (21). Then we set the variation of water pressure equal to the variation of gas pressure to keep the suction constant. The result calculated by our model is compared with Wheelers model. It can be found that there no difference in elastic deformations. The yield stress in our model is lager because we have considered the gas hardening effect, but this is not found in figure 1 because of the yield occurs at a degree of saturation lower than 0.85. In this stage, the gas effect is not considered. Therefore, the rate of increment of the degree of the saturation are the same as Wheelers model until it reaches 0.85, after which the degree of saturation increase more slowly in our model because we assuming that the gas phase become occluded now. The decrease of specific volume and the increase of degree of saturation are a little smaller than that of Wheeler in the end. The degree of saturation dose not reach 1.0 at the end, because the occluded gas is still in the soil and was not expelled completely
Figure 2 shows the specific gas volume and gas pressure during the above loading process. It can be found from the picture that the change of gas deformation is elastic at first and then plastic strain occurs after the yield gas pressure. In the end the gas volume reaches a steady state (not zero) although the gas pressure may still increase during the loading process. That means the occluded
2.3
2.1 specific volume
1.9
1.7
Wheelers model Model with gas effect
1.5 10 100 net mean stress (kPa) (a) 1000
1.1 Wheelers model model with gas effect
1 degree of saturation
0.9
0.8
0.7
0.6 10 100 net mean stress (kPa) (b) 1000
Figure 1. Model simulation at constant suction: (a) specific volume; (b) degree of saturation.
307
0.5
0.4 specific gas volume
0.3
0.2
0.1
200
400 600 net mean stress (kPa) (a)
800
1000
6000 5000 gas pressure (kPa) 4000 3000 2000 1000 0
A constitutive model for isotropic stress state is developed. The model could reflect the influence of the gas phase on the soil behaviour, and couples the three phases in unsaturated soils. But there are ten soil constants which make the model difficult to use, it is sill need further simplification or modification. Since there is few experiment data on the soil with a high degree of saturation, the numerical examples are given at last to illustrate the model. It can be found that the soil will yield later by taking account of the gas hardening effect. And the degree of saturation will not reach 1.0 because of the existence of occluded gas. Further researches are needed on experiments, because there are few experiments on the unsaturated soils with high degree of saturation. Experiment data are essential to verify the reasonability of the model, and will be helpful to extend the model to more complicated situations. REFERENCES
10
100 net mean stress (kPa) (b)
1000
Figure 2. (a) relationship between specific gas volume and net mean stress; (b) relationship between gas pressure and net mean stress.
gas exists in the soil and can not be expelled completely because there is no continuous passage for gas phase. 5 CONCLUSION
The influence of gas phase in unsaturated soils is discussed in the paper. Experiments data has already proved that gas phase may be discontinuous when the degree of saturation is high. The gas pressure will not remain constant when the gas becomes occluded, and its variation will of course influence the soil behaviour. In this case, the gas pressure was adopted as another stress variable to describe the behaviour of unsaturated soil. The work expression based on a strictly theoretical deduction was introduced to verify that three stress variables (the generalized effective stress ij*, the modified matric suction s* and the modified gas pressure ua*) are needed in constructing the constitutive models for unsaturated soil.
Bao, C.G. & Gong, B.W. & Zhan, L.T. 1998. Properties of unsaturated soils and slope stability of expansive soils. In Pro. 2nd Int. Conf. on Unsaturated Soils: 7198. Beijing, China: International Academic Publishers. Houlsby, G.T. 1997. The work input to an unsaturated granular material. Gotechnique 47(1): 193196. Klausner, Y. 1990. Fundamentals of continuum mechanics of soils. Springer-Verlag. Li, X.S. 2007. Thermodynamics-based constitutive framework for unsaturated soils 1: theory. Gotechnique 57(5): 411422. Loret, B. & Khalili, N. 2000. A three-phase model for unsaturated soils. Int. J. Numer. Anal. Mech. Geomech. 24: 893927. Nageswaran, S. 1983. Effect of gas bubbles on the sea-bed behavior. PhD thesis, Oxford University. Terzaghi, K.1944. Theoretical Soil Mechanics. London: Wiley. Thomas, S.D. 1987. The consolidation behavior of gassy soil. PhD thesis, Oxford University. Wheeler, S.J. The stress-strain behaviour of soils containing gas bubbles. PhD thesis, Oxford University. Wheeler, S.J., Sharma, R.S. & Buisson, M.S. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Gotechnique 53(1): 4154. Wroth, C.P. & Houlsby, G.T. 1985. Soil mechanics Property characterization and analysis procedures, Proc, 11th Int. Conf. Soil Mech. Found Eng. 155. San Francisco. Zhao, C.G., Liu, Y. & Gao, F.P. 2009. Work and energy equations and the principle of generalized effective stress for unsaturated soils. Int. J. Numer. Anal. Meth. Geomech. doi: 10.1002/nag.839.
308
A new unsaturated direct shear apparatus for measuring shear strength of unsaturated soils Part 2: Results and comparison between plane strain and triaxial loading conditions
A. Mirzaii, S.S. Yasrebi & E. Hefzi
Tarbiat Modares University, Tehran, Iran
ABSTRACT: This paper demonstrates the results attained from direct shear tests carried out on compacted soil samples of kaolin-sand mixture. Based on the results, shear strength and volumetric behaviour of understudying soil are represented in terms of net stresses and matrix suction, and shear strength parameters are calculated for plane strain loading condition. The results attained from direct shear tests were then compared with the results obtained from unsaturated triaxial tests carried out on similar compacted kaolin-sand soil samples with similar initial conditions. The comparison outlines a slight difference between the shear strength and volumetric behaviour attained from plane strain and triaxial loading conditions. 1 INTRODUCTION of unsaturated soils by using two independent stress variables of matrix suction and net normal stresses: c ( ua) tan (ua uw) tan b (2)
The mechanical behaviour of unsaturated soils has been of interest during the past decades, and diverse engineering aspects of partially saturated soil media have been investigated by the researchers. As most remarkable, shear strength behaviour of unsaturated soils has been studied in many contributions to demonstrate a general shear strength equation. An early shear strength equation was introduced based on the Mohr-Coloumbs failure criteria and Bishops principle of effective stress in unsaturated soils as: c {( ua) (ua uw)} tan (1)
where is the shear stress, c is the internal effective cohesion, ( ua) is the net stress, (ua uw) is the soil suction, is the effective angle of internal friction, and is the bishop parameter and varies between 0 to 1. Several approaches have been taken to introduce a contribution of bishop parameter as a function of degree of saturation (e.g. Blight 1961, Donald 1961, Bishop and Donald 1951, Houlsby 1997), as a function of suction ratio (e.g. Khalili and Khabbaz, 1998), or as a function of suction and hydraulic hysteresis (e.g. Khalili et al. 2008). Due to the complexity of measurement of Bishops parameter, Fredlund, Morgenstern and Widger (1978) introduced the shear strength
where b is the angle of internal friction with respect to matrix suction and represents linear effect of soil suction on the shear strength behaviour. Since the appearance of early shear strength equations, several studies have been carried out to assess different aspects of shear strength behaviour in unsaturated soils such as nonlinear dependency to soil suction (e.g. Escario and Saez 1986), representation with respect to retention behaviour and saturated shear strength parameters (e.g. Fredlund and Xing 1994, Vanapalli et al 1996), description of strength with respect to critical-state concepts (e.g. Toll 1990, Alonso et al 1991, Wheeler and Sivakumar 1995), and implementation of hydromechanical aspects (e.g. Tarantino & Tombolato 2005, Khalili et al. 2008, Sun et al. 2008) which most of tests were operated in triaxial or plane strain loading conditions equipments, such as triaxial and direct shear apparatuses. Different loading conditions and in consequent different test results caused an insight to compare direct shear box and triaxial tests results for different soils (e.g. Rowe 1969 or Maccarini 1993). Typically in direct shear tests, the non coaxiality between principal shear stress and induced shear strain increments and shear box boundaries causes
309
a non homogenous stress-strain distribution in the soil sample. Additionally, mandatory shear zones cause a deviation in accurate measurement of shear strength parameters, in particular for soils with non occurrence of shear zones in failure (e.g. normally consolidated and lightly over consolidated clays and very loose sands). Thus, direct shear tests is still remained popular in engineering practice owning to its simplicity, its advantages in modeling certain soil interaction parameters, and consistency with the results attained from other tests. This paper represents results attained from a series of direct shear tests on unsaturated samples of kaolin-sand mixtures in saturated and unsaturated conditions. Based on experimental observations, shear strength parameters have been calculated for saturated and unsaturated conditions. In addition, a comparison of plane strain and triaxial loading conditions has been done with regard to the results from a similar study with unsaturated triaxial apparatus. The results emphasize a very slight difference in the value of effective angle of internal friction, while a remarkable deviation exists in the values of effective internal cohesion and apparent cohesion. 2 DIRECT SHEAR TESTS AND RESULTS
160 Net Confining Stress = 200 kPa 120

Shear stress (kPa)
Net Confining Stress = 100 kPa 80
40
Net Confining Stress = 50 kPa
0 0 2 4 6 8 Horizontal displacement (mm) (a) 10 12
200
160
Shear stress (kPa)
120
80
40
A series of saturated and unsaturated direct shear tests have been carried out on compacted soil samples with a novel direct shear apparatus for unsaturated soils designed and built in Tarbiat Modares University. The details of the apparatus features and test procedures have been discussed in detail in part one of following paper. Direct shear tests include 6 constant water content tests on two different initial matrix suctions of 30 kPa and 100 kPa in different net confining stresses of 50 kPa, 100 kPa, and 200 kPa. In these tests, the shearing stages were continued until the samples reached to a steady state and the variation of shear stress, soil volume change, or matrix suction become constant. In addition, 3 conventional saturated direct shear tests were performed to assess shear strength parameters in the saturated condition. 2.1 Horizontal loading and failure After wetting and consolidation stage, soil samples were sheared by application of a horizontal constant load in constant water content condition. The variations of shear stress versus horizontal displacement are plotted in figure 1 for samples with initial matrix suctions of 30 and 100 kPa respectively. As shown, the variation of shear stresses
0 0 2 4 6 8 Horizontal displacement (mm) (b) 10
Figure 1. Variation of shear stress versus horizontal displacement during the shear stage for soil samples with a) initial matrix suction of 30 kPa b) initial matrix suction of 100 kPa.
were leveled off with the increment of horizontal loads and become constant in the horizontal displacements of 4 to 9 mm. During the shear, pore air let to be drained while pore water was in undrained conditions and caused a variation in the matrix suction and the pore spaces during the shear stage. Further investigation of the results accentuates the effect of initial matrix suction and the level of net confining stress on the value failure shear stress. The soil samples with initial matrix suction of 100 kPa appeared to sustain greater shear stress than soil samples with initial matrix suction of 30 kPa. Moreover, higher levels of net confining stress increased samples stiffness and in consequent maximum sustainable shear stresses in failure condition. Table 1 demonstrates the maximum observed shear stresses with their corresponding value of matrix suction.
310
Table 1. Maximum value of shear stress and corresponding matrix suction during the application of horizontal loads.
Matrix Suction (kPa)
75
Net Confining Stress=100 kPa Net Confining Stress=50 kPa
60
Net Confining Stress=200 kPa
Net confining stress * 50 100 200 50 100 200
Initial matrix suction 30 30 30 100 100 100
Shear stress 58 83.3 140.8 79.8 105.5 161.9
Matrix suction 59.9 63.7 55.7 131.5 133.8 141.6
45
30
15
* All stress variables are written in kPa.
0 0 1 2 3 4 5 6 7 8 9 10 11
Horizontal displacement (mm)
2.2
Variation of matrix suction and volume change
(a)
150
Figures 2 and 3 represent the variation of soil matrix suction and volume change during the application of horizontal loads respectively. As illustrated in figure 2, in all tests the value of matrix suction initially tended to have a slight decrease at the beginning of horizontal shearing and then reversely continued to increase and became constant until the samples reached to the failure state. Initial decrease in the level of matrix suction was due to the transfer of a portion of applied horizontal loads to the pore water in the soil voids which consequently decreased the level of matrix suction in the constant water content test condition. In further with respect to the increment of horizontal shear stresses, the specimens fabric became more compacted and the soil aggregates moved near to each other and caused a decrease in the pore water pressure in the constant water content conditions of unsaturated soil samples, and correspondingly the value of matrix suction increased and reached to a constant value in the failure state. The application of horizontal shear loads also altered the volume of soil samples which are plotted in figure 3 as vertical displacements to horizontal. As shown in figure 3, soil samples tended to have a slight downward vertical settlement and became compacted at the beginning of shearing and with the continual of shear load application reversely tended to dilate and had upward vertical settlements. Further investigation of results emphasizes the dependency of dilatancy behaviour of soils on the level of matrix suction and net confining stress. In other words, soil samples with lower values of net stresses tended to dilate more during the shear. A similar behaviour can be observed for the soil samples with higher values of matrix suction. In addition for constant values of net confining
135
Matrix Suction (kPa)
120
105
90 0 1 2 3 4 5 6 7 8 9
(b)
Figure 2. Variation of matrix suction versus horizontal displacement during the shear stage for soil samples with a) initial matrix suction of 30 kPa b) initial matrix suction of 100 kPa.
stresses, dilative behaviour was more significant in samples with higher values of suction. In general, dilatancy tendency is more obvious in soils which are on the dry side of the critical state (e.g. heavily over consolidated clays and dense sands) which dilate, soften and weaken during shear. Thus, for soils on the wet side of the critical state (e.g. normally consolidated clays and loose sands) dilative behaviour can also be observed by other conditions such as the geometrical constraints imposed in a direct shear box. For the understudying soil, the level of compaction of the samples was not in as much as to inherently induce dilatancy during the shear and the apparent observed dilative behaviour was due to the geometrical boundary conditions of the direct shear test. This context can be more cleared by comparison
311
0.3 0.2
Vertical displacement (mm)
Table 2. Plane strain shear strength parameters attained from unsaturated direct shear tests on kaolin-sand mixtures. Initial suction* 30 Net stress 50 29 14.6 29.5 13.9 100 200 100 50 28.8 14.6 53.5 16.5 100 200
0.1 0 0 -0.1 -0.2

10
(Deg) c C b(Deg)
13.2
14.9
17.4
15.4
-0.3
(a)
0.7 0.6
0.5 0.4 0.3 0.2 0.1

Net Confining Stress=200 kPa Net Confining Stress=100 kPa
was assumed to be equal to the saturated condition. As shown in table 2, the value of effective angle of internal friction in unsaturated tests has a similar value in compare to the saturated test results. In addition, soil samples with initial matrix suction of 100 kPa tended to have a more apparent cohesion than samples with initial matrix suction of 30 kPa. The value of angle of internal friction with respect to the soil suction was calculated separately for each value of suction to accentuate nonlinear effect of matrix suction on the shear strength behaviour. 3 COMPARISON OF RESULTS WITH TRIAXIAL LOADING CONDITION
0 -0.1 0 2 4 6 8 10
(b)
Figure 3. Soil vertical settlements variations versus horizontal displacement during the shear stage for soil samples with a) initial matrix suction of 30 kPa b) initial matrix suction of 100 kPa.
of similar studies and results attained from triaxial tests which is described in the following sections of the article. 2.3 Plane strain shear strength parameters
Based on the preceding results, table 1, and equation 2, different shear strength parameters were estimated for understudying soil in the plain strain loading condition. Three conventional saturated drained direct shear tests were carried out to assess saturated shear strength parameters, and the value of effective angle of internal friction and effective internal cohesion were estimated to be 29.2 degree and 14.6 kPa respectively. Unsaturated shear strength parameters were also calculated and are shown in table 2. For this condition, the value of effective internal cohesion
Data from a series of unsaturated triaxial tests on a similar soil were used to compare results from direct shear loading tests with triaxial loading condition. Triaxial tests were performed with Mirzaii et al. in similar initial conditions to direct shear tests. Triaxial tests were carried out with a doubledwall unsaturated triaxial cell on similar kaolinsand mixture with similar initial compaction properties. After the compaction procedure, specimens were brought to the suction levels of 30 and 100 kPa with the application of axis translation technique. The soil samples were consolidated in net confining stresses of 50, 100, and 200 kPa and then were sheared in constant water content condition. Further details of test programs and procedures are described more in detail in the Mirzaii et al. The variations of deviator stress versus axial strain are plotted in figure 3 for unsaturated triaxial tests. Corresponding values of phase state and stress state for the maximum shear strength are also shown in table 3 for triaxial tests. The variation of matrix suction, volume change, and shear strength parameters and compared between plane strain and triaxial loading conditions in straight forward.
312
600
Net Confining Stress= 200 kPa
120 100 Matrix suction (kPa) Net Stress= 50 kPa Net Stress= 100 kPa
450 Dev. Stress (kPa)
80 60
Net Confining Stress= 100 kPa 300 Net Confining Stress= 50 kPa 150
Net Stress= 200 kPa 40 20 0 0 5 10 15 (a) 200 Net stress= 50 kPa Net stress= 100 kPa Net stress= 200 kPa 120 20 25 30 35
0 0 5 10 15 (a) 750 Net Confining Stress= 200 kPa Matrix suction (kPa) 20 25 30 35 Axial Strain %
Axial Strain %
160
600
Dev. Stress (kPa)
450 Net Confining Stress= 100 kPa 300 Net Confining Stress= 50 kPa
80
40
0 150 0 5 10 15 20 25 30 35 Axial strain % 0 (b) 0 5 10 15 20 25 30 35
Axial Strain % (b)
Figure 4. Variation of deviator stress versus axial strains in triaxial tests during the shear stage for soil samples with a) initial matrix suction of 30 kPa b) initial matrix suction of 100 kPa. Table 3. Phase state and stress state variables attained from triaxial tests corresponding to the maximum deviator stress in shear stage. Initial suction* 30 Net stress Sr% ua uw p ua q 50 100 200 100 50 100 200 73.8 147.9 397 586
Figure 5. Variation of matrix suction versus axial strains in triaxial tests during the shear stage for soil samples with a) initial matrix suction of 30 kPa b) initial matrix suction of 100 kPa.
70.2 75.9 94.3 54.8 63 95.7 92.7 69.8 142.4 160.7 126.5 205.4 379.9 124 217 203.7 315.3 538.3 225 353
3.1
Variation of matrix suction and volume changes
Figure 5 represents the variation of matrix suction in triaxial tests during the shear, and a similar behaviour to direct shear tests is observed. In similar, the pore water pressure tended to increase at the beginning of shear stage in triaxial tests while reversely tended to decrease with continual of shear loads to a constant value in the steady
state. This behaviour caused an initial decrease in the level of matrix suction and reversely created an increase of matrix suction with the rest of the test until the soil suction reached to a constant value in large strains. Thus, volume changes in triaxial tests were in contrast to the results of direct shear tests in shearing stage. The volume changes of unsaturated specimens are plotted versus axial strains for triaxial constant water content tests in figure 6 for the shear stage. The pore air volume tended to decrease in the shearing stage of constant water content triaxial tests and consequently caused a decrease in the total volume of soil samples. This compressive behaviour is of contrast with the occurrence of dilative behaviour in direct shear tests. Typically, dilatancy is common for heavily over consolidated cohesive soils or for very dense granular geomaterials. For the case of understudying soil, the level of compaction of the samples was not in as much as to inherently behave dilative in shear. The apparent observed dilative behaviour was due to the geometrical constraints in the boundaries of direct shear box which generated a mandatory horizontal shear zone in the middle plane of specimens for the generation of failure in soil samples
313
0 0
10
15
20
25
30
35
Table 4. Triaxial shear strength parameters attained from unsaturated direct shear tests on kaolin-sand mixtures. Initial suction* Net stress 30 50 31.8 3.8 25.7 12.9 100 200 100 50 32 3.8 35.6 12.6 100 200
-2
Vol. Strain %
Net Stress= 50 kPa -4 Net Stress= 100 kPa -6 Net Stress= 200 kPa -8
(Deg) c C b(Deg)
13.3
17.4
11.2
12.1
Axial Strain %
(a)
0 0 NetCW02-S100-50 stress= 50 kPa -2 5 10 15 20 25 30 35
-4
Net stress= 100 kPa CW02-S100-100
-6
Net CW02-S100-200 stress= 200 kPa -8
Axial Strain %
shear box cases the value of plane strain cohesion is more than triaxial cohesion. Further investigation of results shows a similar pattern in the comparison of unsaturated plane strain and triaxial shear strength parameters. The values of effective angle of internal friction were similar for unsaturated triaxial and plane strain conditions, while a reasonable difference appeared in the value of apparent cohesion in two different loading conditions which consequently affected the values of basic friction angles. 4 CONCLUSIONS
Vol. Strain %
(b)
Figure 6. Changes in volumetric strains versus axial strains in triaxial tests during the shear stage for soil samples with a) initial matrix suction of 30 kPa b) initial matrix suction of 100 kPa.
which consequently caused dilative volume change behaviour. 3.2 Shear strength parameters
Based on the results and considering equation 2, shear strength parameters were calculated for understudying soil in the triaxial loading space and are shown in table 4. Triaxial saturated shear parameters of effective cohesion and effective angle of internal friction were estimated to be 3.8 kPa and 31.8 degrees respectively by operation of a series of drained saturated triaxial tests. A comparison of saturated shear strength parameters in triaxial and plane strain condition shows a very slight difference in the value of effective internal angle of friction while there is a significant difference in the value of effective internal cohesion. Typically, direct shear tests are more successful in representation of tensile and cohesive behaviour of soils than conventional triaxial tests and in most cases due to the boundary condition of direct
In this paper, results from a series of unsaturated direct shear apparatus were presented, and shear strength parameters were calculated for unsaturated plane strain loading conditions. The results were then compared with data from a series of unsaturated constant water content triaxial tests on similar soil with similar initial conditions. Direct shear tests included 6 constant water content conditions in two initial matrix suctions of 30 and 100 kPa at net confining stresses of 50, 100, and 200 kPa. At the end of consolidation stages, soil samples were horizontally sheared until they reached to the failure state. The results highlighted the effect of initial matrix suction and the level of net confining stress on the value maximum sustainable shear force in the shearing stage. Due to the undrained condition of water in the pore spaces, the value of matrix suction varied in the shear stage and initially tended to decrease and reversely increased in further and reached to a constant value. In spite of the low level of compaction of soil specimens, geometrical shapes of shear boxes caused dilative volume change behaviour in horizontal shear stage of unsaturated direct shear tests. Based on the result, plane strain shear strength parameters were calculated for understudying soil. Data from a series of unsaturated triaxial tests were used to compare mechanical behaviour of
314
understudying soil in plane strain and triaxial loading conditions in particular for the variation of matrix suction, volumetric behaviour, and shear strength parameters. As shown, a similar behaviour was seen for the variation of soil matrix suction in the shear stage. Thus, compressive volume changes in triaxial tests were in contrast to dilative behaviour of soil samples in direct shear tests. Further investigation of the results and test procedures emphasized the effect of geometrical constraints and mandatory horizontal shear zone of direct shear tests on distinct volumetric behaviour of understudying soil in triaxial and plane strain conditions. Comparison of shear strength parameters in triaxial and plane strain condition showed a very slight difference in the value of effective internal angle of friction, while plane strain values of effective and apparent cohesion were seen to be more than corresponding values in triaxial tests. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Bishop, A.W. 1959. The principle of effective stress. Teknisk Ukeblad I Samarbeide Med Teknikk, Oslo, Norway, 106(39), 859863. Bishop, A.W. & Donald, I.B. 1951. The experimental study of partly saturated soil in the triaxial apparatus. In Proceedings of the 5th International Conference Soil Mechanics and Foundation Engineering, Paris, Vol. 1, pp. 1321. Blight, G.E. 1961. Strength and Consolidation Characteristics of Compacted Soils. Ph.D. dissertation, University of London, London. Donald, I.B. 1961. The Mechanical Properties of Saturated and Partly Saturated Soils with Special Reference to Negative Pore Water Pressure. Ph.D. dissertation, University of London, London. Escario, V. & Saez, J. 1986. The shear strength of partly saturated soils. Some aspects of effective stress in saturated and partly saturated soil. Gotechnique 36(3): 453456.
Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsatuarted soils. Can. Geotech. J. 15(3): 177197, 313321. Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater characteristic curve. Can. Geotech. J. 31(4), 533546. Hefzi, E. 2009. Evaluation of shear strength behaviour of unsaturated soils by direct shear apparatus. MSc Thesis. Tarbiat Modares University, Tehran. Houlsby, G.T. 1997. The work input to an unsaturated granular material. Gotechnique 47(1): 193196. Khalili, N., Habte, M.A. & Zargarbashi, S. 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Computers and Geotechnics 35: 872889. Khalili, N. & Khabbaz, M.H. 1998. A unique relationship for for the determination of the shear strength of unsaturated soils. Gotechnique 48(2): 6817. Maccarini, M. 1993. A comparison of direct shear box tests with triaxial compression tests for a residual soil. Geotechnical and geological engineering 11(2): 6980. Mirzaii, A., Yasrebi, S.S. & Gatmiri, B. Critical-state behaviour of an unsaturated clayey sand. Gotechnique revised paper submitted for publication. Rowe, P.W. 1969. The relation between the shear strength of sands in triaxial compression, plane strain and direct shear. Gotechnique 19(1): 7586. Sun, D., Sheng, D., Xiang, L. & Sloan, S.W. 2008. Elastoplastic prediction of hydro-mechanical behaviour of unsaturated soils under undrained conditions. Computers and Geotechnics 35: 845852. Tarantino, A. & Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique 55(4): 307317. Toll, D.G. 1990. A framework for unsaturated soil behaviour. Gotechnique 40(1): 3144. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction. Can. Geotech. J. 33: 379392. Wheeler, S.J. & Sivakumar, V. 1995. An elasto-plastic critical state framework for unsaturated soil. Gotechnique 45(1): 3553.
315
A constitutive relationship for collapsible soils in terms of stress state variables

J.D. Nelson
Engineering Analytics, Inc., Fort Collins, Colorado, USA
C.N. Hatton K.C. Chao
URS Corporation, Denver, Colorado, USA Engineering Analytics, Inc., Fort Collins, Colorado, USA
ABSTRACT: Soils that are stable when dry, but undergo a decrease in volume with an increase in water content are commonly termed, collapsible soils. This paper presents the results of an investigation of the fundamental mechanism of collapse in unsaturated soils as it relates to stress state variables. A critical value of soil suction was determined at which collapse initiates. It was observed that the critical value of soil suction is that at which the soil begins to imbibe water as exhibited on the soil water characteristic curve (SWCC) curve for imbibition conditions. 1 INTRODUCTION method. The volume change of the sample was measured using axial and radial transducers. Soil samples were consolidated under a constant suction at different values of confining stress. The suction was then relaxed in increments while maintaining a constant confining pressure. The increments of suction change were selected so as to maintain a high resolution of the change in void ratio. A critical value of soil suction was determined at which collapse initiated. It was seen that the critical value of soil suction was approximately that at which the soil begins to imbibe water as exhibited on the SWCC curve for imbibition conditions. 2 2.1 FACTORS AFFECTING COLLAPSE Soil type
Soils that are stable when dry, but undergo a decrease in volume with an increase in water content are commonly termed collapsible soils. Collapsible soils generally have a quasi-stable matrix that derives its strength from either negative pore water pressure (soil suction) or cementation of some nature between the particles. In clayey soils, the stability is commonly the result of clay bonds formed by the clay particles that coat the sand or silt grains, or by flocs of clay between the particles (Barden, et al., 1973). The soil used in this investigation was a sandy silty clay. Consequently, the bonding mechanism was most likely due to clay bonds. It was not a cemented soil. This paper investigates the fundamental mechanism of collapse in unsaturated soils in terms of soil suction. In general, the soil water characteristic curve (SWCC) of a soil defines the residual saturation and displacement pressure, hd, of the soil. It is hypothesized that these characteristics relate to the collapse phenomena. This paper presents results from a carefully controlled research investigation conducted at Colorado State University (Hatton, 1988). A silty clay obtained locally was used in the investigation. The soil was compacted to a low density at a low water content to produce collapse potential. The soil was isotropically consolidated in a triaxial cell. Soil suction was controlled using the axis translation
Almost any soil given the proper conditions can have a potential for collapse. Dudley (1970) and Houston, et al. (1988a) discussed the many groups of naturally deposited soils with a collapsible structure. They were categorized by three methods of deposition: wind, water, or residual soils. Wind deposited soils typically have a porous, open structure with large open and interconnected voids. Water deposited collapsible soils are the result of rapid deposition of large volumes. In general, these soils are composed of structurally competent materials including sand, gravel, cobbles,
317
and boulders. However, they have a weak matrix of low density that allows them to collapse with the addition of water. Residual soils typically have an open porous structure resulting from the weathering of elements of the matrix. 2.2 Fabric
soil is typical of the Front Range Area. Soil properties are listed in Table 1. Clay constituents were determined using X-ray diffraction techniques. 3.2 Sample preparation
The term fabric describes the arrangement of particles, particle groups, and pore spaces. In a collapsing soil, the fabric governs the behavior of the soil, and particle arrangement gives the soil its ability to collapse. Clay bonds are a common type of bond. There are varieties of clay bonds, one being the result of clay covering the individual sand grains, and another being flocs of clay formed at sand and silt particle contacts. Barden et al. (1973) suggested that the soil gains its strength from electro-chemical properties of the clay as well as suction effects. The addition of water relaxes the suction and causes the clay flocs to weaken resulting in collapse. 2.3 Critical water content
Samples of uniform density and structure were obtained by utilizing an undercompaction method as described by Ladd (1978). In this procedure, a lower compactive effort was initially applied to the lower layers and increased in each subsequent layer such that the degree of compaction was the same throughout the final sample. The procedure is described in detail in Hatton (1988). 3.3 Soil water characteristic curves
Kane (1967) indicated that collapse occurs at a specific critical water content. A soil with a water content above the critical level would not exhibit potential for collapse. Kane (1967) stated that soil maintained the same deformation characteristics even above the critical water level. Holtz & Hilf (1961) proposed that loess having a water content below 20% would resist consolidation when dry, but would exhibit collapse behavior with the addition of water and load. However, a soil with natural water content above 20% would consolidate with the application of load. Booth (1975 & 1977) found that the critical water content varied with soil type and initial density. Booth indicated that the critical water content is 2% greater than the optimum water content. Research by Houston (1988b) indicated that a percentage of full collapse results from partial wetting, and presents percentages of full collapse that correlate with degrees of saturation. According to Houston (1988b), 100% of the fully wetted collapse potential can result from a degree of saturation of 90%. 3 3.1 TESTING PROCEDURES Soil
The soil water characteristic curves (SWCCs) were determined using pressure plate and pressure membrane apparatus for values of suction above 2 kPa. The soil samples were compacted into molds 25 mm deep and 55 mm in diameter to a wet density of 15.5 kN/m3. Details of the test procedures and apparatus are presented in Hatton (1988). The imbibition curve was determined in the same manner as the desaturation curve. However, the SWCCs were not determined at values of suction below 2 kPa for the imbibition cycle. The displacement pressure, hd, was found to be below the resolution of the pressure plate/membrane equipment. Therefore, a more direct method adopted from McWhorter & Sunada (1977) was used for suction values below 2 kPa. The apparatus is shown in Figure 1. A capillary tube was used to measure the change in water content of the soil. Suction was increased at increments of 10 mm of water and allowed to equilibrate. Below the displacement pressure, very small changes in water content were observed for each incremental change in suction. A large and immediate change was observed when the suction was sufficient to initiate desaturation, i.e., when the suction reached the displacement pressure. This method was also applied to samples subjected to different values of applied net normal stress. The displacement pressure was measured for values of net normal stress of 10, 20, 40, and 80 kPa. Typical soil water characteristic curves for drainage and imbibition are shown in Figure 2. The observed desaturation displacement pressures for zero effective stress ranged from 0.79 kPa to 1.57 kPa (80 to 160 mm of water). The displacement pressure varied linearly with the applied net normal stress as shown in Figure 3. 3.4 Sample consolidation
Soil samples were obtained locally from test pits dug at the Agronomy Research Farm at Colorado State University in Fort Collins, Colorado. The soil is an Aeolian deposit consisting of sandy silty clay. This
Collapse characteristics were measured using isotropically consolidated samples in a triaxial cell.
318
Table 1. Soil properties. Standard proctor Atterberg limits USCS classification CL LL (%) 30 PL (%) 15 Specific gravity Gs 2.7 Compaction properties wopt (%) 14.5 max kN/m3 18 Gradation Sand (%) 38 Silt (%) 24 Clay (%) 38 Mineral Smectite (%) 70 Illite (%) 10 Kaolinite (%) 20 Quartz (%) 1
PORVIC SAMPLE CAPILLARY TUBE
INPUT HOUSE AIR PRESSURE INPUT HOUSE VACUUM
POSITIVE/NEGATIVE PRESSURE REGULATOR
MONOMETER (SHOWING NEGATIVE PRESSURE)
Figure 1.
10,000
Apparatus for measurement of displacement pressure, hd.

30
Imbibition
Pressure-Saturation Data Displacement Pressure (kPa) 1-D Consolation Data 20
1,000 Capillary Suction (kPa)
Desaturation
100
10
10
0.1 10 20 30 40 50 Volumetric Water Content
0 -20 0 20 40 60 80 100 Effective Stress (kPa)
Figure 2.
Soil Water Characteristic Curves (SWCC).
Figure 3. Variation of displacement pressure (hd) with applied confining stress.
The base of the cell was equipped with four inlets. Two entered the cell, and the other two entered the sample pedestal. The inlets to the pedestal were used to provide pore water pressure to the sample and allowed flushing of the diffused air from
beneath the porous stone. One of the cell inlets was used to provide the confining pressure, and the other was used to supply air pressure to the top cap on the sample. A recess was machined in the
319
top of the pedestal to allow for the placement of a high air entry stone. Sealed flexible tubing connected the air supply from the cell base to the top cap. The air then passed through the small port in the top cap to the porous stone located on top of the specimen. In this way, the system provided for separate control of the pore water pressure at the bottom of the stone and pore air pressure at the top cap of the sample. Only isotropic states of stress were applied on the samples. The strain tensor was measured by determining axial and radial deformation. The axial deformation was measured through the loading piston. Radial deformation was measured by means of inductance type transducers mounted on the side of the specimen (Hatton, 1988). After the soil specimen had been prepared the cell pressure and pore air pressure were increased in 5 kPa increments until the desired cell pressure was attained. The sample was allowed to come to equilibrium. When zero inflow or outflow of water was observed, the sample was consolidated to the desired confining net normal stress in increments over a period of several days. Once the desired consolidation stress was attained, the soil suction was reduced in increments. While the soil suction was being reduced, the confining pressure was also reduced so as to maintain a constant net normal stress. The collapse of the sample was considered finished at the point when the sample stopped imbibing water. The test was completed once suction had been reduced to zero kPa and the soil stopped imbibing water. Table 2 lists the initial conditions of the soil. 3.5 Stress paths
define the constitutive plane during collapse. Test 1 saturated the sample at a low confining pressure; this test defined the intersection of the constitutive surface with the void ratio, e vs. (-ua) plane at zero suction. Test 2 was consolidated to an initial suction of 300 kPa. The net normal stress was then increased in measurements to 400 kPa after which the suction was reduced to zero in increments. This defined the outer limits of the constitutive surface for this investigation. The remaining four tests were conducted so as to define the constitutive plane. For Test 3 the initial suction was reduced to 150 kPa and then consolidated to 100 kPa. The suction was then reduced in increments to 75, 50, 25, 10, 5, and 0 kPa. It was observed that most of the collapse occurred below 75 kPa, beginning at or around 50 kPa. Tests 4 and 5 were conducted in a manner similar to Test 3 except that the samples were initially consolidated at a suction of 75 kPa. In Test 6 the sample was consolidated at a suction of 75 kPa and after the suction had been reduced to zero it was loaded to a net normal stress of 400 kPa.
Table 2.
Initial test conditions. Initial consolidation stresses
Test no. 1 2 3 4 5 6
Suction (kPa) 0 300 150 75 75 75
Net normal stress (kPa) 0 400 100 200 400 30
The stress paths followed during the testing are shown in Figure 4. Six tests were conducted to
1.0
0.75 6
(k Pa
1 3 100
Suction (kPa) 200
2 300 400
0.5
0
St re m al or Ne tN
20 0
ss
Figure 4.
Stress paths followed for each test.
30
10
320
RELATIONSHIP BETWEEN SUCTION AND COLLAPSE
The projections of the stress paths onto the suction-void ratio plane are shown in Figure 5. Test 1 was basically a consolidation test at zero suction and plots as a line on the vertical axis. In Test 2 the increments by which the suction was reduced were too large and the line from zero to 75 kPa suction is not well defined. In all other tests, the suction was reduced in increments from 75 kPa to 50, 25, 10, 5, and 0 kPa. The projection of the stress paths onto the void ratio-net normal stress plane are shown in Figure 6. From Figures 5 and 6, the following observations are made. Regardless of the stress path followed, the void ratio at the final conditions of zero suction on ( ua) 400 kPa, was the same for all tests. Also, except for Test 6, which was conducted at a very low applied stress, the amount of collapse did not appear to vary greatly with applied stress. Also, the stiffness of the soil, as shown by the slopes of the stress paths in Figure 6, appear to be nearly the same regardless of the applied suction. Also, the slope did not appear to be affected by whether the collapse had occurred before or after the application of the confining stress. This would
indicate that the soil stiffness is influenced primarily by the fabric of the soil matrix and not by the clay bonds. In Tests 3 and 4, the suction was increased after the final stress state conditions had been reached. As seen in Figure 6 there was no increase in the void ratio showing that the collapse process is irreversible. This, of course, was expected. Figure 7 shows the percent collapse as a function of the suction for all tests. The data generally appears to follow the same trend. During Test 6 some expansion occurred as the suction was relaxed. This is probably due to expansion of the montmorillonitic content of the clay fraction. At higher values of applied stress that amount of expansion was restricted by the applied stress. After the heave had occurred, collapse occurred as the suction was reduced to zero. Results from Matyas & Radhakrishna (1968) and Barden et al. (1969) are also presented in Figure 7. The results of this investigation agree fairly well with the trends shown by previous investigations. From Figure 7 it is seen that in Tests 4 and 5 the collapse was initiated at a suction close to 50 kPa. In Test 3 some collapse occurred beginning around 75 kPa. In Test 2 the increment of suction decrease was too large to determine at what suction the
1.0
1.0 6
0.9 3 Void Ratio (e) 0.8 1 2 Void Ratio (e)
0.9
0.8 4 0.7 5
0.7 3 0.6
0.6 5
0.5 0 100 Suction (kPa) 200 300
0.5 0 100 Suction (kPa) 200 300
(a) Tests 1, 2, and 3
(b) Tests 4, 5, and 6
Figure 5.
Projections of stress paths on e vs. h plane.

1.0 0.9 0.8 Void Ratio (e) 0.7 0.6 0.5 0.4 10 100 Effective Stress (kPa) 1000 1 3 1.0 0.9 0.8 Void Ratio (e) 0.7 4 0.6 0.5 0.4 10 100 Effective Stress (kPa) 1000 5 6
(a) Tests 1, 2, and 3
(b) Tests 4, 5, and 6
Figure 6.
Projections of stress paths on e vs. ( ua) plane.
321
125 Barden et al (1969) 100 Percent of Total Collapse
75
Test 2, 400 kPa Test 3, 100 kPa Test 4, 200 kPa Test 5, 400 kPa Test 6, 30 kPa Matyas and Radhakrishna (1968)
50
25
The stiffness of the soil did not change appreciably after the collapse had occurred. The collapse phenomenon was seen to be irreversible. The percent of full collapse potential that occurs is a function of the suction. Collapse was initiated as soon as the suction had been reduced to the point at which the sample began to imbibe water and continued until the suction had been reduced to zero. REFERENCES
Barden, L., Madedor, A.O. & Sides, G.R. 1969. Volume Change Characteristics of Unsaturated Clay. Journal of Soil Mechanics and Foundations. ASCE. 95:3351. Barden, L., McGown, A. & Collins, K. 1973. The Collapse Mechanism in a Partially Saturated Soil. Engineering Geology. 7(1):4960. Amsterdam. Booth, A.R. 1975. The Factors Influencing Collapse Settlement in Compacted Soils. Proc. 6th Regional Conference for Africa Soil Mechanics and Foundation Engineering. Durban. 1:5763. Booth, A.R. 1977. Collapse Settlement in Compacted Soils. National Institute for Transport and Road Research. Bulletin 13, Report: 324. Dudley, J.H. 1970. Review of Collapsing Soils. Journal of Soil Mechanics and Foundations. ASCE. 96(SM3): 925947. Hatton, C.N. 1988. Constitutive Relationships for Collapse in a Remolded Soil, M. S. Thesis, Colorado State University (J.D. Nelson, advisor). Holtz, W.G. & Hilf, J.W. 1961. Settlement of Soil Foundation Due to Saturation. Proc. 5th International Conference for Soil Mechanics and Foundation Engineering. Paris. 673679. Houston, S.L., Houston, W.M. & Spadola, D.J. 1988a. Prediction of Field Collapse of Soil due to Wetting. Journal of Geotechnical Engineering. ASCE. January. 114(1):4058. Houston, S.L. 1988b. Pavement Problems Caused by Collapsible Subgrades. Journal of Transportation Engineering. November. 114(6). Paper No. 22902. Kane, H. 1967. Consolidation of Two Loessial Soils. Highway Reasearch Record. No. 284. 2637. Ladd, R.S. 1978. Preparing Test Specimens Using Undercompaction. ASTM Geotechnical Journal. March. 1:1623. Matyas, E.L. & Radhakrishna, H.S. 1968. H.S. Volume Changes Characteristics of Partially Saturated Soils. Geotechnique. 18:432448. McWhorter, D. & Sunada, D.K. 1977. Groundwater Hydrology and Hydraulics. Fort Collins: Water Resources Publications.
50
100
150
200
250
300
Suction (kPa)
Figure 7. Effect of suction vs. percent of total collapse.
collapse was initiated. Nevertheless, in Tests 3 through 5 it appears that the collapse generally occurred at the same general value of suction, and was not affected by applied stress. The value at which collapse initiates appears to be about the same as the suction at which water content increased for the imbibitions SWCC as shown in Figure 2. Thus, collapse was initiated as soon as the water content began to increase above the residual saturation, i.e., at the point where the soil began to imbibe water. Test 6 was conducted at a very low value of applied stress and collapse did not initiate until a low value of suction was reached. This indicates that the value of suction at which collapse is initiated is, in fact, a function of the applied stress. The stress at which the effect of suction was stress dependent was somewhere between 30 kPa and 100 kPa. Also of interest in Figure 7 is the fact that the degree of collapse depends on the suction, and total collapse is not achieved until the suction reaches zero. 5 CONCLUSIONS
A carefully controlled investigation was conducted into the relationship between suction, applied stress, and collapse. From the results the following conclusion can be drawn. For various stress paths the final void reaction was the same. That is, the total decrease in volume resulting from collapse and compression of the soil was the same.
322
Direct shear strength of compacted bentonite under different suctions

T. Nishimura H. Rahardjo J. Koseki
Ashikaga Institute of Technology, Tochigi, Japan Nanyang Technological University, Singapore Institute of Industrial Science, University of Tokyo, Japan
ABSTRACT: This study focuses on both soil-water characteristic curve and shear strength of compacted bentonite. Soil-water characteristic curve was measured using vapor pressure technique and soil suction of the compacted bentonite was measured by a chilled mirror dewpoint potentiometer in high suction ranges. Hysteresis between drying path and wetting path was found to be significant. Direct shear tests under constant volume were conducted on compacted bentonite of different suctions and under constant volume swelling conditions. This study demonstrates the influence of suction on stress path of shear stress versus vertical stress. Calculated shear strength parameters (effective angle of internal friction and apparent cohesion) decrease with reduction of suction and when swelling occurs.
INTRODUCTION
2 2.1
TESTING PROGRAM Soil materials
This study focuses on shear strength of compacted bentonite at different suction values. The soil used in this test program is sodium bentonite mixed with silica sand. Soil-water characteristic curve was measured using vapor pressure technique which can provide a relationship between gravimetric water content and high soil suctions. Soil suctions of compacted bentonite were measured directly using a chilled mirror dewpoint potentiometer. This study provides comparison between the magnitude of suction controlled by the vapor pressure technique with and the measured value from the chilled mirror dewpoint potentiometer. Direct shear tests were carried out under constant volume conditions. As a result, the stress path or the relationship between shear stress and vertical stress during the direct shear tests under constant volume can be measured. The constant volume direct shear tests were performed on compacted bentonite at various normal stresses and suctions higher than 2.8 MPa. In addition, direct shear tests were also conducted on compacted bentonite under constant volume conditions and the results were compared with those obtained from compacted bentonite at high suctions.
Sodium bentonite was used for this test program. Silica sand was mixed into the bentonite with a ratio of 30% by dry weight. The soil specimen was statically compacted in the shear box under a compression stress of 3.5 MPa. The soil specimen had a dry density of 1.6 g/cm3. 2.2 A chilled mirror dewpoint potentiometer
Soil suctions of compacted bentonite were measured directly using a chilled mirror dewpoint potentiometer (WP4-T of DECAGON DEVICES) as shown in Photo 1. WP4-T is an instrument for measuring water potential. The water potential is measured in MegaPascal and pF units. The measured water potential is recorded as soil suction in MPa for this study. The measuring range and accuracy of WP4-T is summarized in Table 1. 2.3 Direct shear equipment
Figure 1 shows the shear box for the direct shear apparatus. The shear box consists of upper box
323
Photo 1. A chilled mirror dewpoint potentiometer (WP4-T). Table 1. Range and accuracy of a chilled mirror dewpoint potentiometer. Range (MPa) 0 to 10 10 to 300 Accuracy (MPa or %) 0.1 MPa 1%
suction. Vanapaplli et al., (1999) indicated that a suction value over than 1500 kPa can be considered as high suction. The low suction (1500 kPa) can be controlled in a pressure plate apparatus using a porous plate or a micro porous membrane. The high suction can be controlled using a vapor pressure technique. High suction values can be achieved by controlling relative humidity (RH) using several different salt solutions in a desiccator (Oteo-Mazo et al., 1995, Delage et al., 1998, Vanapalli et al., 1999). The salt solution method was used in this study to achieve high suction values as summarized in Table 2. The glass desiccator was then placed in a temperature controlled chamber for at least 14 days to allow the specimens to achieve an equilibrium condition with respect to the controlled suction. 2.5 Measurement of soil-water characteristic curve for bentonite The compacted bentotite specimen with an initial water content of 4.4% was placed in glass desiccators with different salt solutions. The seven different salt solutions summarized in Table 2 were used in the test program. The gravimetric water content of each soil specimens was measured after achieving the equilibrium condition. As a result, the relationship between measured water content and soil suction in high soil suction range can be obtained. The salt solution used in vapor pressure technique controlled relative humidity in the glass desiccator. The suction corresponding to relative humidity (RH) can be calculated using the relationship between relative humidity and suction (i.e., Lord Kelvins equation) (Fredlund and Rahardjo, 1993). 135022 ln(RH) (1)
where is soil suction or total suction (kPa) and RH is relative humidity (%).
Figure 1. Shear box used for direct shear test under constant volume.
and lower box. Five rings were placed between upper box and lower box. The rings were used to minimize frictions between the boxes. The shear box is placed in a chamber where water can be supplied to submerge the specimen. 2.4 Vapor pressure technique
Table 2. Relationship between relative humidity (RH) and suction values for different salt solutions at 20C (Delage et al. (1998)). Salt solutions (Chemical symbol) K2S04 KNO3 NH4H2PO4 NaCl Mg(NO3)2 6H2O MgCl2 6H2O LiCl Relative humidity (%) 98 95 93.1 75 54 33 11 Suction (kPa) 2,830 6,940 9,800 39,000 83,400 148,000 296,000
Soil suction has a range from 0 kPa to 106 kPa which consists of matric suction and osmotic suction (Fredlund, 2006). Measurement technique for soil suction is dependent on the magnitude of soil
324
The value of suction in compacted bentonite specimen at equilibrium condition corresponded to relative humidity generated by the salt solution as summarized in Table 2. The controlled suction value was then measured using the chilled mirror dewpoint potentiometer. Measurement of suction value was carried out for at least 30 minutes. 2.6 Direct shear test
Glass desiccator Direct shear box
Soil specimen
The specimens were statically compacted at an initial water content of 4.4%. The soil suction of the specimen was measured using the chilled mirror dewpoint potentiometer which gave a suction value of 110 MPa. Three different series of direct shear tests under constant volume were carried out on compacted bentonite specimens. In the first series of tests, soil specimen was allowed to swell by introducing water to the specimen. Photo 2 shows the process where the soil specimen was allowed to swell when water was introduced to the specimen. The upper portion of the rod was connected to a load cell. The load cell was fixed to the frame of the chamber. During swelling, the swelling pressure of the compacted bentonite was measured at least once a month. In this testing program free swell was not allowed. Subsequently, the shear box was set up in the direct shear apparatus. The applied vertical stresses to the soil specimen were 200 kPa, 500 kPa, 1000 kPa and 1500 kPa for investigating the effect of vertical stress on shear strength of compacted bentonite.
Salt solution
Figure 2. Vapor pressure technique.
Pressure gauge
Load cell senser
De-air water Shear box Data roger
In the second series of tests, the shear box with the statically compacted specimen was placed in the glass desiccator to achieve a desired suction value by controlling relative humidity as explained in Section 2.4 (see Fig. 2). The desired suction value was 2.8 MPa corresponding to RH of 98% as achieved using. Potassium Sulfate (K2SO4). Typically, a period of month was required to achieve the equilibration condition. Subsequently, the shear box was transferred to the direct shear test apparatus. The four different vertical stresses were applied to the soil specimens. In the third series of tests, the soil specimen compacted at an initial water content of 4.4%. As mentioned above, the specimen had a suction value of 110 MPa. The shear box was transferred to the direct shear test apparatus. The four different vertical stresses were applied to the specimen before the shearing process. All of three different series of direct shear tests had different vertical stresses (i.e., 200 kPa, 500 kPa, 1000 kPa and 1500 kPa). After ensuring equilibrium conditions, the specimen was sheared at a rate of 0.05 mm per minute. The soil specimen was sheared to an approximately 6 mm horizontal displacement or until a peak value of shear stress was observed. The total volume of soil specimen was maintained constant during the shearing of the specimen by increasing the vertical stress on the specimens. The constant volume condition was checked by monitoring vertical displacement gauge placed on the specimen. It was ensured that the maximum vertical movement of the specimen was not more than 0.01 mm throughout the shearing stage. 3 TEST RESULTS
Chamber
3.1 Measurements of suction of salt solution using the chill mirror dewpoint potentiometer The different salt solutions were put into separate cups for the chilled mirror dewpoint potentiometer measurements and the suction measurements
Photo 2. A compacted bentonite specimen during swelling in the chamber.
325
were conducted for a period of 30 minutes. Figure 3 shows the relationship between measured suction by the chilled dewpoint potentiometer and the theoretical suction values summarized in Table 2. The measured suction values appear to agree closely with the theoretical suction values. Meanwhile, the suction of bentonite specimen was controlled by equilibrating the specimen with the suction corresponding to the relative humidity in the glass desiccator. Relationship between the measured suction of bentonite specimen and the suction controlled by the salt solution is shown in Figure 4. The controlled suction by the salt solution is plotted in the horizontal axis. The relationship
shows good agreement until the suction value of 83 MPa and the data sets start to deviate from the diagonal line for suction values beyond 83 MPa. 3.2 Soil-water characteristic curve in high suction
Relationship between gravimetric water content and suction of the compacted bentonite is shown in Figure 5 as a soil-water characteristic curve. The suction plotted on the horizontal axis had a range from 2.8 MPa to 305 MPa. The soil-water characteristic curve consists of a drying path and a wetting path. The bentonite had a gravimetric water content of 14% at a suction value of 305 MPa. Hysteresis between drying and wetting curves appears to be significant. 3.3 Shear stress versus horizontal displacement relationship The relationship between shear stress and horizontal displacement for compacted bentonite of different suctions and the swelling bentonite at constant volume are presented in Figures 6, 7, 8. Four solid black circles in Figure 6 describe the maximum shear stress value correspond to each vertical stress. The solid black circles in Figures 7 and 8 are explained on section 3.4. The measured gravimetric water contents of the compacted bentonite specimens used in three different test series are as follows; the submerged bentonite was 44.9%, the bentonite with suction of 2.8 MPa was 15.8% and the bentonite with suction of 110 MPa was 4.4%. The shear stress versus horizontal displacement obtained from the first test series is shown in Figure 6. The shear stress of the swelled bentonite increased smoothly at the beginning of shearing process. The effect of vertical stress on
200 Gravimetric water content % 150 100 50 0 1 10 Suction MPa 100 1000 Drying path Wetting path
Figure 3. Comparison between controlled suction from salt solutions and measured suctions using chilled mirror dewpoint potentiometer.
Figure 4. Comparison between suction in bentonite specimen as controlled by salt solutions and suction measurement of bentonite specimen using chilled mirror dewpoint potentiometer.
Figure 5. Soil-water characteristic curve of compacted bentonite in high suction range.
326
shear strength is observed for the entire range of horizontal displacement. Figure 7 shows stress-strain curve for the bentonite with an initial suction of 2.8 MPa. It shows a non-linear increase in shear stresses with respect to horizontal displacement, similar to the characteristics observed in the swelled bentonite. Due to the effect of high suction value, the shear strength increased rapidly with increasing horizontal displacements. The stress-stain curve for the initial suction of 110 MPa is shown in Figure 8. The shear stress increased slowly at the beginning of shearing and achieved a maximum shear stress that is higher compared with both the swelled bentonite and the bentonite at suction of 2.8 MPa. The effect of vertical stress on shear strength remains significant
regardless the high magnitude of suction. The shear strength parameters of the compacted bentonite were determined in accordance with the Mohr-Columb failure theory. 3.4 Failure envelope
Shear stresses were plotted against vertical stresses for the three different test series. Figure 9 shows variations of vertical stress for swelled bentonite. Shear stress increased while vertical stress remained to be the same as the initial vertical stress with the exception of the vertical stress of 200 kPa. Subsequently, the bentonite specimen achieved the failure condition. The relationship between shear stress and vertical stress for bentonite with the suction of 2.8 MPa is shown in Figure 10. The shear stress increased while maintaining a constant value of vertical stress before the shear stress reached the failure
Figure 6. Shear stress versus horizontal displacement relationship for swelling bentonite under constant volume. Figure 8. Shear stress versus horizontal displacement for bentonite at the suction of 110 MPa.
300 250 Shear stress kPa 200 150 100 50 0 0 500 1000 1500 2000 Vertical stress kPa Swelled
Figure 7. Shear stress versus horizontal displacement for bentonite at the suction of 2.8 MPa.
Figure 9.
Stress paths for swelled bentonite.
327
Table 3. Determined strength parameters. Swelled Effective angle of internal friction (degrees) Apparent cohesion (kPa) 2.6 20 Suction of 2.8 MPa 9.9 51 Suction of 110 MPa 10.5 70
in the strength parameters. In other words, swelling caused considerable reductions in shear strength.
Figure 10. 2.8 MPa. Stress paths for bentonite at suction of
CONCLUSIONS
Failure envelopes of compacted bentonite at high suctions were obtained from direct shear tests. High suctions were controlled using vapor pressure technique. In addition, the soil-water characteristic curve of compacted bentonite was also determined. Hysteresis was found to be significant even at high suction range. It appears that the compacted bentonite had at least a gravimetric water content of 14% at 305 MPa suction. The results indicate that swelling caused significant reduction in shear strength parameters of compacted bentonite. ACKNOWLEDGEMENTS
Figure 11. 110 MPa.
Stress paths for bentonite at suction of
This research work was supported by the Ashikaga Institute of Technology. Also, department of Civil Engineering in Ashikaga Institute of Technology support the research. REFERENCES
Delage, P., Howat, M.D. & Cui, Y.J. 1998. The relationship between suction and swelling properties in a heavily compacted unsaturated clay. Engineering Geology. 50. pp. 3148. Fredlund, D.G. 2006. Unsaturated soil mechanics in engineering practice. Journal of Geotechnical and Geoenvironmental Engineering. Vol. 132. No. 3. pp. 286321. Fredlund, D.G. & Rahardjo, H. 1993. Chapter 4, Measurement of Soil Suction, Soil Mechanics for Unsaturated Soils. A Wiley-InterScience Publication, JOHN WILEY & SONS, INC. pp. 64106. Vanapalli, S.K., Fredlund, D.G. & Pufahl, D.E. 1999. The influence of soil structure and stress history on the soil-water charcteristics of a compacted till. Geotechnique. Vol. 49. No. 2. pp. 143159.
line. Four black circle symbols refer to the minimum vertical stress value during shear. Each point corresponds to the four points on stress-horizontal displacement curves shown in Figure 7. Figure 11 shows the variations of shear stress against vertical stress for bentonite with the suction of 110 MPa. Except for vertical stresses of 1000 kPa and 1500 kPa, the shear stresses increased along the failure line after the vertical stress achieving a minimum value. Three failure envelopes were defined using the stress paths. Strength parameters were determined from three failure envelopes and summarized in Table 3. When the bentonite had a suction of 110 MPa, two strength parameters (effective angle of internal friction and apparent cohesion) are higher than those obtained from other cases. A decrement in suction caused the reduction
328
Investigation of the behavior of an unsaturated sand using a cyclic direct shear device
T. Nishimura I. Shahrour H.B. Bian
Ashikaga Institute of Technology, Japan Universit de Lille, France University of Metz, France
ABSTRACT: Several investigators have widely used cyclic triaxial apparatus to assess the liquefaction resistance for sands. Some studies used a triaxial apparatus to appear the liquefaction resistance of sands. Cyclic triaxial tests used for the unsaturated sand conducted out recently. This study focuses on cyclic shear properties of Hostun sand under unsaturated condition using modified direct shear apparatus. The matric suction of soil specimens were controlled from 0 kPa to 20 kPa used pressure membrane technique. The mobilized shear stress and excess pore-water pressure of unsaturated Hostun sand were measured under cyclic loading. 1 1.1 INTRODUCTION Back ground test for compacted, unsaturated silty soil using the modified triaxial test apparatus. Nishimura (2009) measured the changes of pore-water pressure during cyclic loading. As above mentioned, a few experimental researches were reported dynamic properties of different soil type using cyclic triaxial test apparatus. 1.2 Purpose of this study
Cyclic shear tests are conventionally used to evaluate the liquefaction potential for saturated sands under undrained loading conditions. Saturated cyclic triaxial tests were accepted to interpret the liquefaction for saturated sands. Experimental researches regard to unsaturated sand cyclic behavior was reported with controlling of matric suction. In addition, the cyclic direct shear apparatus is possible to measure the mobilized shear stresses. A height of soil specimen for the cyclic direct shear test is generally less than that for the cyclic triaxial test. It is effective that equilibrium time become to be fast in the controlling matric suction. Regard to dynamic properties of saturated sands, many experimental researches had been mentioned in the past. Yoshimi et al. (1989) described the relationship between degree of saturation and B-value for Toyoura sand before cyclic loading. They described that decreasing of degree of saturation was effective to increase the liquefaction resistance for unsaturated Toyoura sand. It is widely accepted that the liquefaction resistance of unsaturated sands significantly increased with decreasing in degree of saturation. The modified cyclic triaxial test apparatus for unsaturated soils able to measure the matric suction. Nishimura (2009) conducted out cyclic triaxial
This study demonstrates the cyclic shear properties of unsaturated Hostun sand using the modified direct shear apparatus. The Hostun sand was sheared in the range from 0 kPa to 20 kPa in matric suction under constant normal stress. The matric suction in the unsaturated Hostun sand was controlled using pressure membrane technique in the process of application of normal stress (i.e., before cyclic shearing). The excess pore-water pressure of unsaturated Hostun sand was measured under undrained condition. The mobilized shear stress of the unsaturated Hostun sand decreased with the cyclic number. When the cycle number exceeded fifth, the mobilized shear was close to be 10 kPa. It was interest that the residual shear resistance of unsaturated Hostun sand was not influenced by the initial matric suction. Consequently, the residual shear strength of the unsaturated Hostun sand subjected to cyclic loading independent to the initial matric suction.
329
2 2.1
LITERATURES Shear strength for unsaturated soils
Unsaturated soil mechanics is required to solve dynamic behavior of unsaturated soils subjected to cyclic loading or earthquake motion. The solutions of dynamic behavior of unsaturated soils are not established, and the geotechnical engineers require evaluation the experimental test results and predictions. The shear strength of unsaturated soil has been formulated in terms of two independent stress state variables (i.e., net normal stress and matric suction). The shear strength equation for unsaturated soil proposed by Fredlund et al. (1978) has been accepted widely in geotechnical engineers. Fredlund et al. (1978) proposed shear strength equation for an unsaturated soils as equation (1): f [c (n ua) tan ] [(ua uw) tan b] (1)
soil-water characteristic curve. The effective degree of saturation is defined in equation (2) which has t similar role with defined by Bishops effective stress for unsaturated soils. Sre (Sr0 Sr ) (Sr0 Srr ) (2)
where Sre is effective degree of saturation, Sr0 is degree of saturation at zero matric suction, Srr is degree of saturation at residual suction and Sr is degree of saturation at any matric suction. 2.3 Prediction model using soil-water characteristic curve
where f is shear strength of an unsaturated soil, c is effective cohesion of the soil, (n ua) is net normal stress, is effective angle of shearing resistance for a saturated soil. (ua uw) is matric suction, b is angle of shearing resistance relative to an increasing in matric suction. The matric suction is defined as the difference between pore-air pressure and pore-water pressure. Applications of matric suction need time consuming and cost in the variety experimental tests. Many geotechnical engineers performed to predict the shear strength using the soil-water characteristic curve, which the matric suction had a limited ranges. 2.2 Soil-water characteristic curve The soil-water characteristic curve is relationship between the soil suction and the soil moisture (either water content or degree of saturation). The soil-water characteristic curve is key feature in the geotechnical engineering, agriculture, soil science. The term is variety depend on the each research areas. In generally, the soil-water characteristic curves were measured in pressure technique using porous plate. The pressure plate technique occupies time consuming, and often causes to induce errors or incorrect data sets. Nishimura et al. (2009) was successful to control relative low matric suction in short experimental times using the pressure membrane technique. Several reports presented many mathematic models for the soil-water characteristic curve on limited soil suction ranges. Two or three fitting parameters govern to simulate satisfactory. The effective degree of saturation is conceptual and interpretative tool which obtained from the
Several models have been suggested to realize the relationship between shear strength and soil suction for unsaturated soils. The mathematical models was formulated by Fredlund et al. (1996). Vanapalli et al. (1996), Khalili et al. (1998) and Oberg et al. (1997), and predicted the unsaturated soil shear strength using the soil-water characteristic curve. 3 3.1 TEST PROCEDURE Soil material
This test program used Hostun sand. The grain size distribution is shown in Fig. 1. Table 1 shows the components obtained from the grain size distribution. The soil-water characteristic curve for Hostun sand was determined using pressure membrane technique. The curve is shown in Fig. 2. Square points means the drying pass through (i.e., change of degree of saturation when the matric
Figure 1.
Grain size distribution for Hostun sand.
330
Table 1.
Properties of Hostun sand. Components
Sand % Silt % Clay %
97.7 2 0
Photo 1. Modified direct shear apparatus for cyclic loading.
Figure 2. Soil-water characteristic curve for Hostun sand.
suction increases). Triangle points mean the wetting pass trough (i.e., matric suction decreases). The soil-water characteristic curve extended up to the matric suction of 1000 kPa using the mathematic model suggested by Fredlund & Xing (1994). 3.2 Direct shear test apparatus
This test program used the modified direct shear test apparatus, which consisted of a cell chamber, direct shear box, pore-pressure supply system, drainage system and cyclic horizontal displacement control system. The whole apparatus is shown in Photo 1. The direct shear box was placed in the air chamber as shown in Photo 2. The direct shear box consist upper box and lower box. The micro porous membrane was installed into the lower box. The micro porous membrane had a properties mentioned in Table 2. Particularly, the coefficient of permeability (i.e., hydraulic conductivity) was 1.67 109 cm/sec. Photo 3 shows the lower box with the micro porous membrane. A diameter of direct shear box was 60 mm. The test apparatus is possible to remain a constant normal stress during shear. In cyclic process, test apparatus creates a cyclic horizontal displacement using AD transfer system. The lower shear box moved smoothly at range of 1.0 mm in alternation. Desired shear speed was 1.0 mm per minutes. A loading sensor fixed at the upper direct shear box
Photo 2.
Direct shear box in the air chamber.
Table 2. Properties of micro porous membrane used in this test program. Thickness (m) 140 Air entry value (kPa) 250 Pore diameter (m) 0.45
Material Polyether sulfone
measured the mobilized shear forces on cyclic shear process. 3.3 Test procedure for cyclic direct shear test
Saturated micro porous membrane was placed on the porous stone in the lower direct shear box. A water supply compartment under the porous stone was
331
Photo 3. Lowe portion of the direct shear box with micro porous membrane.
Figure 3. Decrement of shear stress of saturated Hostun sand subjected to cyclic loading.
saturated fully. After the direct shear box was built up, the box was submergence in the de-aired water. Hostun sand was stored into the shear box within the de-aired water. Hostun sand specimen was fully saturated, and a normal stress of 50 kPa was loaded. All of saturated Hostun sand specimens with the normal stress of 50 kPa had a void ratio of 0.72. Subsequently, the shear box was transferred to the cyclic direct shear test apparatus. The different matric suctions were applied to the saturated Hostun sand specimens using pressure membrane technique. The desired matric suctions had a range from 0 kPa to 20 kPa (i.e., 0, 4 and 20 kPa). After ensuring equilibrium conditions, saturated or unsaturated Hostun sand was loaded due to cyclic horizontal displacement interactions. Cyclic shearing was applied up to the cyclic number of 15. The height of sand specimen was measured during cyclic shearing. The excess pore-water pressures were measured under undrained conditions. The air pressure applied in the chamber was maintained constant corresponding to the desired matric suctions. 4 4.1 TEST RESULTS Change of shear stress due to cyclic loading
Figure 4. Decrement of shear stress of unsaturated Hostun sand at matric suction of 4 kPa subjected to cyclic loading.
The three different matric suctions were applied to saturated Hostun sand specimens under the constant normal stress of 50 kPa. The relationship between cyclic number and shear stress was shown in Figs. 3 to 5. All Hostun sand specimens were described highly shear stress regardless of magnitude of matric suction when virgin horizontal displacement was applied. Subsequently, shear stresses decreased considerable with cyclic numbers. When cyclic number was over fifth, the Hostun sand specimens achieved to the residual shear stress condition. The remained shear resistance was approximately 10 kPa at number of 15.
Figure 5. Decrement of shear stress of unsaturated Hostun sand at matric suction of 30 kPa subjected to cyclic loading.
All of Hostun sand specimens reached to the residual condition regime. This tendency was observed regardless of both saturated and unsaturated conditions.
332
4.2
Volume change under cyclic loading
Vertical displacement of all Hostun sand specimens was measured during cyclic shear. The relationship between vertical displacement and number of cyclic are shown in Figs. 6 to 8. Positive vertical displacement values mean the mount of shrinkage for Hostun sand specimens. All of Hostun sand specimens described the shrinkage at commence of interaction shear. Saturated Hostun sand and unsaturated Hostun sand with matric suction of 4 kPa induced shrinkage, and mount of vertical displacement approached to 1.0 mm. However, few shrinkages were measured to matric suction of 20 kPa. The mount of shrinkage was less than 0.5 mm at residual condition. It was evidenced that effect of matric suction influence to the dilation of unsaturated Hostun sand.
Figure 8. Change of vertical displacement of unsaturated Hostun sand at matric suction of 20 kPa subjected to cyclic loading.
Figure 6. Change of vertical displacement of saturated Hostun sand subjected to cyclic loading.
Figure 9. Change of excess pore-water pressure of saturated Hostun sand subjected to cyclic loading.
Figure 7. Change of vertical displacement of unsaturated Hostun sand at matric suction of 4 kPa subjected to cyclic loading.
Figure 10. Change of excess pore-water pressure of unsaturated Hostun sand at matric suction of 4 kPa subjected to cyclic loading.
333
Hostun sand approached to be steady in residual conditions, when cyclic number was fifth. Finally, measured residual shear resistance was approximately 10 kPa. 2. The saturated and unsaturated Hostun sand occurred shrinkage at commence of cyclic loading under constant normal stress. Mount of shrinkage decreased with increase of matric suction. 3. The excess pore-water pressures were measured under undrained condition for the unsaturated Hostun sand. The excess pore-water pressures described as positive values during cyclic shear. Occurrence of positive pore-water pressure was less than 0.5 kPa.
Figure 11. Change of excess pore-water pressure of unsaturated Hostun sand at matric suction of 20 kPa subjected to cyclic loading.
ACKNOWLEDGEMENT This empirical test research was supported in 2008 by Collaborative Research Center, Ashikaga Institute of Technology Japan. REFERENCES
Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsaturated soils, Canadian Geotechnical Journal, Vol. 15, No. 3, pp. 313321. Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater characteristic curve, Canadian Geotechnical Journal, 31, pp. 521532. Fredlund, D.G., Xing, A., Fredlund, M.D. & Barbour, S.L. 1996. The relationship of the unsaturated soil shear strength to the soil-water characteristic curve, Canadian Geotechnical Journal, 32, pp. 440448. Khalili, N. & Khabbaz, M.H. 1998. Unique relationship for the determination of the shear strength of unsaturated soils, Geotechnique, 48(5), pp. 681687. Nishimura, T. 2009. Measuring excess pore-water pressure of an unsaturated silty soil in cyclic shear test, IS-Tokyo, pp. 14711476. Nishimura, T., Toyota, H. & Koseki, J. 2009. Evaluation of apparent cohesion of an unsaturated soil using difference shear test, 4th Asia-Pacific Conference on Unsaturated Soils, 2325 November 2009, Newcastle, Australia, pp. 109114. Oberg, A. & Sallfours, G. 1997. Determination of shear strength parameters of unsaturated silts and sands based on the water retention curve, Geotechnical Testing Journal, 20, pp. 4048. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength envelope with respect to soil suction, Canadian Geotechnical Journal, 33, pp. 379392. Yoshimi, Y., Tanaka, K. & Tokimatsu, K. 1989. Liquefaction resistance of a partially saturated sand, SOILS AND FOUNDATIONS, Vol. 29, No. 3, pp. 157162.
4.3
Excess pore-water pressure under undrained condition
Relationship between excess pore-water pressures and cyclic numbers are shown in Figs. 9 to 11. The pore-water pressures of Hostun sand were measured under undrained condition, and the pore-air pressures were maintained a constant. Change of pore-water pressures correspond to change of matric suction. For saturated Hostun sand specimen, measured excess pore-water pressures were remained negative values during shear, and the pressure had approximately minus 0.5 kPa. In case of unsaturated Hostun sand, excess pore-water pressures were described as positive. At matric suction of 4 kPa, measured positive porewater pressure was 0.5 kPa at least. However, the Hostun sand with matric suction of 20 kPa had significant low subjected to cyclic loading. Occurrence of positive pore-water pressure was observed regardless of magnitude of matric suction. 5 CONCLUSIONS
This study conducted out the cyclic direct shear test for both saturated and unsaturated Hostun sand. Relatively low matric suctions were applied to saturated Hostun sand specimens using pressure membrane technique. The pressure membrane technique was effective method instead of the pressure plate technique. During cyclic shear, the normal stress of 50 kPa was maintained. The amplitude applied to Hostun sand specimens was 1.0 mm under a shear horizontal speed was 1.0 mm per minute. Results obtained from this test program leaded some interprets as follows: 1. Decrement in shear resistance ensured at few cyclic numbers. Both saturated and unsaturated
334
Pore pressure under undrained condition for compacted unsaturated silty soil having a low matric suction
T. Nishimura J. Koseki
Ashikaga Institute of Technology, Japan Institute of Industrial Science, University of Tokyo, Tokyo, Japan
ABSTRACT: Experimental studies, such as performing triaxial compression tests while controlling the matric suction using axis translation technique, are useful in establishing the mathematical modeling of unsaturated soils. In many relevant studies, the triaxial tests are processed under drained and exhausted condition, in which the matric suction is kept constant during shearing stage. Since few experimental studies have been conducted under undrained and unexhausted condition, shear properties of unsaturated soil under different drainage and exhaust conditions are investigated in this study. In this test program, the triaxial compression tests are conducted on a compacted unsaturated soil, where an initial matric suction of 20 kPa was applied to soil specimens using the pressure membrane technique instead of a ceramic disk. Change in the matric suction is observed as results of measurement of both the pore-air pressure and the pore-water pressure. Effective degree of saturation was employed in evaluating the effective stresses in unsaturated soil instead of in Bishops equation. 1 INTRODUCTIONS The effective stress path during shearing under undrained-unexhausted condition is obtained, which is defined as relationships between mean effective principal stress and deviator stress. 2 2.1 LITERATURE SURVEY Shear strength of unsaturated soils
Several different types of soils in nature are more commonly found in a state of unsaturated condition, particularly in arid and semi-arid regions. Also, expansive soils, collapsible soils, residual soils and compacted soils are conventionally categorized as problematic soils. These problematic soils are frequently in a state of unsaturated condition. The influence of suction is generally ignored in conventional geotechnical engineering which adopts the conservative assumption that the soils are in a state of saturated condition (Vanapalli, 2009). This study presents shear properties of an unsaturated soil under undrained and unexhausted condition. In this test program the triaxial compression tests are performed on a compacted unsaturated soil. An initial matric suction of 20 kPa was applied to soil specimens using the pressure membrane technique instead of a ceramic disk. Also, two different shear strain rates are employed in the shearing stages. Change in the matric suctions is observed as results of measurement of both the pore-air pressure and the porewater pressure. Effective degree of saturation was employed in evaluating the effective stresses in unsaturated soil instead of in Bishops equation. The soil-water characteristic curve was referred to in evaluating the effective degree of saturation.
The effective stress equation of saturated soils proposed by Terzaghi (1936) laid a foundation toward the establishment of geotechnical engineering. Two stress variables, namely net normal stress and matric suction are required for interpreting the shear properties of unsaturated soils (Fredlund & Rahardjo, 1993a). Bishop (1959) proposed the effective stress equation using two stress variables for unsaturated soils including a parameter, . It can be used for interpreting the shear strength of unsaturated soils extending the same philosophy as Terzaghi (1943). f [c (n ua) tan ] [(ua uw) tan b] (1)
where, f is shear strength of unsaturated soil, c is effective cohesion, is effective internal friction angle, (n ua) is net normal stress, (ua uw) is matric suction, b is parameter depended on the degree of saturation.
335
Many investigators attempted to realize the influence of net normal stress and matric suction for understanding the unsaturated soils engineering problems (Bishop & Donald (1961), Jennings & Burland (1962), Matyas & Radhakrishna (1968)). Fredlund et al (1978) suggested unsaturated soil shear strength equation including shear contribution due to matric suction, b instead of a parameter, suggested by Bishop (1959). The axis-translation technique proposed by Hilf (1956) pushed forward to unsaturated soil experimental tests. The air entry value of ceramic disk should be greater than the matric suction of the soil specimen. Several conventional triaxial tests for measuring the shear strength of unsaturated soils were proposed by Bishop & Donald (1961). For the unsaturated loose silt, a consolidated drained test was performed, that matric suction was maintained constant during shear. Previously, the results of consolidated drained test on unsaturated soils were presented by Bishop & Donald (1961), Satija (1978), Gulhati & Satija (1981), Escario (1980), Ho & Fredlund (1982), Fredlund & Rahardjo (1987). Further studies by Ho & Fredlund (1982) described planar failure envelope adequately defined failure conditions. Recent experimental evidence using a higher range of matric suction (i.e., 0 to 500 kPa) indicates a significant non-linear in the failure envelope (Gan (1986), Escario & Saez (1986)). Nishimura (2004) also reported a non-linear in the shear stress versus matric suction, relationship. These results were obtained using modified direct shear test on silty soil. Fredlund, Rahardjo & Gan (1987) published the experimental evidences and procedures with respect to the non-linear of the failure envelope for unsaturated soils. In addition, they attempted to address the question of whether all testing procedures yield unique shear strength parameters. 3 3.1 TEST PROCEDURE Soil materials
100 90 80 70 Passing (%) 60 50 40 30 20 10 0

0.001
0.01 0.1 1 10
Size (mm)
Figure 1.
Grain size distribution for the soil material.
Figure 2.
The basement of SWCC test apparatus.
This test program used a non plastic silty soil having a grain size distribution curve as shown in Fig. 1. SWCC test and unsaturated triaxial test were performed on statically compacted specimens with a mould water content of 10%. The soil specimen had a void ratio of 0.89, dry unit weight of 13.7 kN/m3 and a degree of saturation of 29.7%. 3.2 SWCC test apparatus
apparatus is shown in Fig. 2. The SWCC test apparatus is capable of employing both the pressure membrane technique and the pressure plate technique. Both a micro porous membrane and a ceramic porous disk were installed in the basement of the apparatus as shown in Fig. 2, on which a steal mould was built in an acryl air chamber. External air pressure that corresponds to the matric suction is applied to the acryl air chamber. A nylon tube was connected to the water supply compartment, through which the water expelled from the specimen was accumulated into a glass double burette that was placed outside the chamber. The amount of drainage water was measured using a differential pressure sensor. The sensor was installed at bottom of the double glass burette. Vertical displacement of soil specimen was directly measured with a displacement sensor for evaluating the total volume change of the specimen. Change in the degree of saturation was evaluated based on measurement of both the drainage water and the vertical displacement. 3.3 Modified triaxial compression apparatus for unsaturated soil A conventional double-cell type triaxial compression test apparatus was employed in this study. Its schematic figure is shown in Fig. 3. The apparatus
Determination of the soil-water characteristic curve was made using a newly developed SWCC test apparatus. The basement of the SWCC test
336
Figure 4. The modified pedestal for pressure membrane technique. Figure 3. The triaxial compression test apparatus.
is capable of measuring volume change of the specimen, pore-air pressure and pore-water pressure. The pore-air pressure is supplied through the top of the specimen. The pore-water pressure is measured outside the chamber using a pressure sensor. Volume change of soil specimens is recorded using a pre-calibrated gap sensor installed in the inner cell. Since the pressure membrane technique is found successful in controlling or measuring relatively lower matric suction (i.e. from 0 kPa to 20 kPa) (Nishimura et al (2009)), a new pedestal was developed for employing the pressure membrane technique. The newly developed pedestal is shown in Fig. 4, on which a micro porous membrane was fixed by means of a fixing plate and six screw nails. The micro porous membrane had a hydraulic conductivity of 1.67 109 cm/sec. Nishimura and Koseki (2008) measured air diffusion through the micro porous membrane and reported the diffused air spectrum. The diffused air through membrane accumulated under the micro porous membrane when the applied matric suction was over 60 kPa. 3.4 Test procedure for SWCC test
matric suction, when change of the amount of drainage water became negligible. Vertical deformation of soil specimen was measured using a displacement sensor. As drying process, the matric suction was increased up to 180 kPa using pressure plate technique. After end of test, water content of soil specimen was directly determined by oven dry method. 3.5 Test procedure for unsaturated triaxial test under undrained-unexhausted condition
The ceramic disk and the micro porous membrane were saturated before setting the soil specimen in the steal mould. After placing the soil material in the steal mould, a compression force was loaded to its top surface. The soil specimen had a void ratio of 0.89 and a degree of saturation of 29.7%, respectively. The soil specimens were submerged before applying the air pressure. Therefore, the initial matric suction became nearly zero at once. Subsequently, air pressure was supplied into the air chamber for applying matric suction, which created the water flow in the soil specimen. The air pressure was maintained till approaching the equilibrium condition of the soil moisture distribution in the soil structure for the desired magnitude of
A soil specimen was placed on the modified pedestal, and confining pressure of 100 kPa was loaded isotropically. Application of hydraulic seepage flow was performed for achieving apparent saturation. Thus, the matric suction was lowered to be nearly zero. Subsequently, isotropic consolidation pressure of 100 kPa was loaded. An initial matric suction of 20 kPa was supplied, and soil water was drainaged through the micro porous membrane while maintaining the net normal stress at 100 kPa (i.e. cell pressure was 120 kPa and pore-air pressure was 20 kPa). Deviator stress was loaded with two different rates of shear strain (0.013%/min and 0.005%/ min). In addition to volume change of the specimen, pore-air pressure and pore-water pressure were measured during shear process in order to evaluate the change of matric suction. 4 4.1 TEST RESULTS Stress-strain curve
Figure 5 shows relationship between stress and strain. Deviator stress increased rapidly at commencement of shear. The soil specimens yielded till axial strain reached to 5%. Subsequently, residual conditions were observed at axial strain exceeding 15%. Figure 6 shows the change of excess pore-air pressure. The pore-air pressure increased with axial strain. The increment of excess pore-air pressure became almost
337
250 Shear speed is 0.013 %/min Deviator stress kPa 200 150 100 50 0 0 5 10 Axial strain % 15 20 S Shear speed is 0.005 %/min
Excess pore-water pressure kPa
40 35 30 25 20 15 10 5 0 0 5 10 Axial strain % 15 20 Shear speed is 0.013 %/min Shear speed is 0.005 %/min
Figure 5. Relationship between deviator stress and axial strain.

30 Excess pore-air pressure kPa 25 20 Shear speed is 0.013 %/min 15 10 5 0 Shear speed is 0.005 %/min
Figure 7. Relationship between excess pore-water pressure and axial strain.
30 25 Matric suction kPa 20 15 10 5 0 Shear speed is 0.013 %/min
Shear speed is 0.005 %/min
10 Axial strain %
15
20
10 Axial strain %
15
20
Figure 6. Relationship between excess pore-air pressure and axial strain.
Figure 8. Relationship between matric suction and axial strain.
zero at axial strain of 12%. At end of shearing, excess pore-air pressure achieved to 12 kPa or larger. Change of excess pore-water pressure is shown in Fig. 7. Excess pore-water pressure was generated more rapidly than the pore-air pressure. The increment of excess pore-water pressure gradually decreased, and the amount of residual excess pore-water pressure reached 27 kPa or larger. Such generation of excess pore-water pressure that was larger than that of excess pore-air pressure agreed with the behavior mentioned conceptuality by Fredlund & Rahardjo (1993b). Matric suctions were calculated as difference between pore-air pressure and pore-water pressure. Change in matric suction during the shear process under undrained-unexhausted condition is shown in Fig. 8. The soil specimen was subjected to the decrement of matric suction. Significant decrement was observed till axial strain of 5%. Subsequently, the matric suction became nearly constant. At end of shear, the residual matric suction was approximately 4 kPa.
Change of degree of saturation is shown in Fig. 9. At end of shear, the increment of degree of saturation was approximately 6%. However, the degree of saturation at end of shear was less than 100%, suggesting that the soil specimens did not achieve the state of saturated condition. 4.2 Evaluation of effective stress for an unsaturated soil Unsaturated soil stress variables such as net normal stress and matric suction governs stress-deformation behavior. Bishop (1959), Bishop and Donald (1961) investigated the effect of matric suction on stress-strain behavior obtained from unsaturated triaxial test. Effective stress equation for unsaturated soil was formulated using the empirical parameter. Several researches predicted unsaturated shear strength associated with the soil-water characteristic curve that made it possible to explain the change in shear strength with the matric suction.
338
100 90 Degree of saturation %

Degree of saturation %
100 90 80 70 60 50 40 30 20 10 0
80 70 60 50 40 30 20 10 0 0 5 10 15 Axial strain % 20 Shear speed is 0.005 %/min Shear speed is 0.013 %/min
10
100 Suction kPa
1000
Figure 9. Relationship between degree of saturation and axial strain.
Figure 10. Soil-water characteristic curve.
120 Lateral effective stress kPa
In this test program, the effective degree of saturation was employed instead of Bishops . It was defined as equation (2). Sre (Sr 0 Sr ) (Sr 0 Srr ) (2)
100 80 60 40 20 0 Shear speed is 0.013 %/min Shear speed is 0.005 %/min
Deviator stress kPa
where Sr0 is effective degree of saturation, Sr0 is degree of saturation at zero matric suction, Srr is degree of saturation at residual suction and Sr is degree of saturation at any concerned matric suction. Figure 10 shows the soil-water characteristic curve. Residual degree of saturation for the tested specimen was determined as 15% from the soilwater characteristic curve. Relationship between computed effective stress and degree of saturation during the triaxial tests is shown in Fig. 11. At beginning, the effective stress was 116 kPa. It decreased with axial strain associated with the decrease in the matric suction. At axial strain of 15%, the effective stress became approximately 90 kPa. 4.3 Effective stress path
10 15 Axial strain %
20
Figure 11. Relationship between lateral effective stress and axial strain.
500 400 300 200 100 0 0 100 200 300 400 500 Mean effective principal stress kPa Slope is in the ratio 1:3
Effective stress paths under undrained-unexhausted condition, defined as the relationship between mean effective principal stress and deviator stress, are plotted in Fig. 11. At beginning of shear, the path followed the line at a slope of 1:3. After yielding, the path changed in a nonlinear manner and approached to failure envelope corresponding the internal friction angle of 31.1 degrees. It should be noted that the failure envelop shown in Fig. 12 was obtained based on another series of unsaturated triaxial compression test results by Nishimura et al (2008). Nishimura et al (2008) investigated critical state line of the same
Failure envelope has the internal friction angle (31.1 degrees) & apparent cohesion (5.3 kPa).
Figure 12. Effective stress path under undrainedunexhausted condition.
soil material as tested in this test program, while considering both net normal stress and matric suction measured in three triaxial compression tests.
339
CONCLUSIONS
In this study, a series of triaxial compression tests were conducted on a compacted unsaturated silty soil under undrained-unexhausted condition. An initial matric suction of 20 kPa was applied using a modified triaxial pedestal by means of a pressure membrane technique. As result of measurement of both pore-air pressure and pore-water pressure, change in the matric suction was observed during shear. Both excess pore-air pressure and pore-water pressure were accumulated with axial strain. At the beginning of shear, increment of the excess pore-water pressure larger than that of the excess pore-air pressure. As a result, decrement of matric suction was observed, and the matric suction became nearly constant at the end of shear. Evaluation of effective degree of saturation based on both the soil-water characteristic curve and changes in matric suction significantly contributed to interpretation of the effective stress change. In this study, relationship between deviator stress and mean effective principal stress and the failure envelope were also obtained. They were useful in understanding the unsaturated soil shear properties including the effect of pore pressures. ACKNOWLEDGEMENTS This study was supported by Ashikaga Institute of Technology, Japan. Department of Civil Engineering of Ashikaga Institute of Technology, Japan also supported this experimental study. REFERENCES
Bishop, A.W. 1959. The principle of effective stress. Lecture delivered in Olso, Norway. 1955. Technisk Ukeblad. 106(39). 859863. Bishop, A.W. & Donald, I.B. 1961. The experimental study of partly saturated soil in the triaxial apparatus. In proceedings of the 5th International Conference on Soil Mechanics and Foundation Engineering. Paris. Vol. 1. 1321. Escario, V. 1980. Suction controlled penetration and shear tests. Proceedings of the Fourth International Conference on Expansive soils. Vol. 2. 781797. Escario, V. & Saez, J. 1986. The shear strength of partly saturated soils. Geotechnique. Vol. 36. No. 3. 453456. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of Unsaturated Soils. Canadian Geotechnical Journal. Vol. 15. No. 3. 313321 Fredlund, D.G. & Rahardjo, H. 1987. Soil mechanics principle for highway engineering in arid regions. Transportation Research Record 1137. Washington, D.C. 111.
Fredlund, D.G., Rahardjo, H. & Gan, J. 1987. Nonlinearity of strength envelope for unsaturated soils. Proceedings of 6th International Conference Expansive Soils. New Delhi. Vol. 1. 4954. Fredlund, D.G. & Rahardjo, H. 1993a. Chapter 9. Shear strength theory, Soil Mechanics for Unsaturated Soils. A Wiley-Interscience Publication. JOHN WILEY & SONS. INC. 236247. Fredlund, D.G. & Rahardjo, H. 1993b. Chapter 8. Pore pressure parameters. Soil Mechanics for Unsaturated Soils. A Wiley-Interscience Publication. JOHN WILEY & SONS. INC. 184201. Gan, J.K-M. 1986. Direct shear strength testing of unsaturated soils. M.Sc. Thesis. University of Saskatchewan. Saskatoon. S.K, Canada. Gulhati, S.K. & Satija, B.S. 1981. Shear strength of partially saturated soils. Proceedings of 10th International Conference on Soil Mechanics and Foundation Engineering. Stockholm. Sweden. Vol. 1. 609612. Hilf, J.W. 1956. An investigation of pore-water pressure in compacted cohesive soils. Ph.D. dissertation. Tech. Memo. No. 654. U.S. Dep. Of the Interior. Bureau of Reclamation. Design and Construction Div. Denver. CO. 654. Ho, D.Y.F. & Fredlund, D.G. 1982. A multistage triaxial test for unsaturated soils. Geotechnical Testing Journal. GTJODJ. Vol. 5. No. 1/2. 1825 Jennings, J.E. & Burland, J.B. 1962. Limitation of the use of effective stress in partly unsaturated soils. Geotechnique. 12(2). 125144. Matyas, E.L. & Radhhakrishna, H.S. 1968. Volume change characteristic of partial saturated soils. Geotechnique. 18(4). 432448. Nishimura, T. & Vanapalli, S.K. 2004. Shear strength behavior of an unsaturated silty soil under constant volume conditions. 57th Canadian Geotechnical Conference. GeoQuebec 2004. 1724. Nishimura, T., Toyota, H., Vanapalli, S. & Won,O. 2008. Determination of the shear strength behavior of an unsaturated soil in the high suction range using the vapor pressure technique. Proceedings of the First European Conference on Unsaturated Soils. E-UNSAT 2008. 441447. Nishimura, T. & Koseki, J. 2008. Air diffusion of micro porous membrane used for pressure membrane technique. Seminar for Unsaturated Soil. Hokkaido University. 3334. (in Japanese) Nishimura, T., Toyota, H. & Koseki, J. 2009. Evaluation of apparent cohesion of an unsaturated soil. Proceedings of the 4th Asia Pacific Conference on Unsaturated Soils. Newcastle. Australia. 109114. Satija, B.S. 1978. Shear behaviour of partly saturated soil. Ph.D. thesis. Indian Institute of Technology. Delhi. India. Terzaghi, K. 1936. The shear resistance of saturated soils. Proceedings of 1st International Conference on Soil mechanics and Foundation Engineering 1. 5456. Terzaghi, K. 1943. Theoretical Soil Mechanics. John Wiely and Sons. New York. NY. USA. Vanapalli, S.K. 2009. Shear strength of unsaturated soils and its application in geotechnical engineering practice. Proceedings of the 4th Asia Pacific Conference on Unsaturated Soils. Newcastle. Australia. 579598.
340
The relationship between the elastic and shear modulus of unsaturated soils
W.T. Oh & S.K. Vanapalli
Civil Engineering Department, University of Ottawa, ON, Canada
ABSTRACT: The variation of the elastic (E) and the shear (G) modulus of soils with respect to suction can be predicted using several semi-empirical models. However, limited studies are reported in the literature to investigate the well known interrelationships between the E and the G for soils that are in a state of unsaturated condition. This paper revisits the experimental results of Mendoza et al. (2005) determined using the bender element technique to examine the well known relationship between the E and the G with respect to different degrees of saturation (i.e. suction). The analyses of the results show that the Poissons ratio, is a function of degree of saturation and is a sensitive parameter in the reliable estimation of the E and the G for unsaturated soils. 1 INTRODUCTION are commonly used in the estimation of the shear modulus at small strain (i.e. less than 0.001%), which is referred to as maximum shear modulus (i.e. Gmax). Picornell & Nazarian (1998) and Lee et al. (2007) studied the effect of matric suction on Gmax using the bender element technique by modifying the pressure plate extractor. Qian et al. (1993), Mancuso et al. (2002) and Kim et al. (2003) investigated the variation of the Gmax with respect to matric suction using the resonant column technique. These studies also demonstrate that the Gmax is significantly influenced by matric suction. The determination of the Emax and the Gmax for unsaturated soils is time-consuming and requires elaborate testing equipment. Therefore, it would be valuable to propose empirical or semi-empirical models to estimate or predict the elastic and shear modulus of unsaturated soils using simple techniques. Such studies will be of interest to the practicing engineers. In the present study, typical behaviour of the Emax and the Gmax of soils with respect to suction is discussed. There are limited studies reported in the literature that discuss the interrelationships between the Emax and the Gmax of unsaturated soils. Hence, the measured Emax and Gmax values with respect to degree of saturation for a fine-grained compacted soil (Ip 38%) using bender element technique (Mendoza et al. 2005) were revisited. The analyses of the results show that the Poissons ratio, is a function of degree of saturation (i.e. suction) and is a sensitive parameter in the reliable estimation of the Emax and Gmax values of unsaturated soils. This study is promising and can lead to the justification of deriving Gmax values with the measured or predicted Emax values or vice versa using the existing models.
The elastic and shear modulus of soils are two key parameters that are required in the evaluation of the immediate settlement and dynamic response behavior respectively of geotechnical structures in engineering practice. The elastic and shear modulus of soils are typically assumed to be constant both below and above the ground water table in homogeneous soil deposits. The elastic modulus of a soil below a shallow foundation can be reliably estimated using the linear range of stress versus settlement relationship from in-situ plate load tests extending the theory of elasticity (i.e. Emax) (Timoshenko & Goodier 1951). The shallow foundations are typically placed above ground water table in many arid and semiarid regions and the stresses associated with the foundation loading is not extended into the saturation zone; hence determination of Emax ignoring the influence of matric suction above the ground water table can be conservative. Recently, Oh et al. (2009) and Vanapalli & Oh (2010) investigated the effect of matric suction on the elastic modulus by analyzing the model footing and the in-situ plate load tests results of different unsaturated soils. These studies highlight the importance of taking account of the influence of matric suction in the estimation of the elastic modulus of unsaturated soils. The shear modulus is conventionally determined using in-situ tests (i.e. surface wave test, seismic crosshole test and dynamically loaded drilled shaft) or laboratory tests (i.e. bender element technique, resonant column technique, cyclic triaxial test and cyclic direct simple shear test). These test methods
341
PREDICTION OF THE ELASTIC AND SHEAR MODULUS OF UNSATURATED COARSE-GRAINED SOILS Elastic modulus
2.1
Vanapalli et al. (2008) and Oh et al. (2009) analyzed model footing tests results for different sands to investigate the variation of elastic modulus with respect to matric suction. The elastic modulus was calculated using Eq. (1) for the elastic range of the stress versus settlement relationship (i.e. Emax) (Figure 1). E (1 2 )
q p
where Emax(sat), Emax(unnsat) elastic modulus under saturated and unsaturated condition, respectively, Pa atmospheric pressure (i.e. 101.3 kPa), (ua uw) matric suction, S degree of saturation corresponding to (ua uw); and , fitting parameters. The fitting parameter values of 2.5 and 1 were recommended to obtain good agreement between the measured and the predicted Emax values for coarse-grained soils. 2.2 Shear modulus Wu et al. (1984) and Qian et al. (1993) performed resonant column tests to study the effect of degree of saturation (i.e. suction) on the Gmax for coarsegrained soils. The results show similar trends as elastic modulus for unsaturated coarse-grained soils (Oh et al. 2009) (i.e. curve (i) in Figure 2(b)). Based on the experimental results, Wu et al. (1984) proposed an empirical equation to estimate
BpIw
(1)
where (/qp) slope of settlement versus plate pressure, Poissons ratio, Bp width or diameter of plate, and Iw influence factor. The analyses showed that elastic modulus behavior of unsaturated coarse-grained soils can be represented as curve (i) in Figure 2(b) for the three zones of SWCC (Vanapalli et al. 1999) as below. Boundary effect zone: the elastic modulus linearly increases up to the air-entry value, (ua uw)b. Transition zone: the elastic modulus nonlinearly increases up to a certain matric suction value then gradually decreases. Residual zone: the elastic modulus decreases and tends towards a constant value. A simple model was proposed by these investigators to predict the variation of elastic modulus with respect to matric suction (Eq. (2)). (ua uw ) (S ) Emax( unsat ) Emax( sat ) 1 ( / . ) P 101 3 a (2)
Air Soil Water film

100
Degree of saturation, S (%)
(a)
80 60 40 20 0
Boundary effect zone
Transition effect zone
Residual suction value
Residual zone of unsaturation Air-entry value (ua - uw)b
(b) (ii) E max , G max (i)
Load
Coarse-grained Soil
Curves (i) and (ii): Possible scenarios
Applied stress
Plate (or model footing)
Bp
(c) (i) E max , G max (ii) (iii)
ran ge
Ela s
tic
qp
E=
2 (1- ) Bp Iw
qp )
Fine-grained Soil
Curves (i), (ii), and (iii): Possible scenarios
Suction, (ua - u w)
Settlement
Figure 1. Schematic of the procedure for estimating the elastic modulus from plate load or model footing test.
Figure 2. (a) Typical Soil-Water Characteristic Curve with three different stages of desaturation; and variation of the elastic and shear modulus for (b) coarse- and (c) fine-grained soils with respect to suction.
342
the Gmax with respect to degree of saturation (Eqs. (3) and (4)). Gmax(S) [1 H(S)] Gmax(dry) S (a 1)sin H (S ) 2Sopt (a 1)H1(S )H2 (S ) (3) S Sopt S Sopt (4)
due to the effects that are similar to cementation (Rinaldi et al. 1998) (i.e. curve (ii) in Figure 2). In this case, the highest value of Vs (i.e. Vs(max)) can be obtained at S 0 using Eq. (7). Vs(max) ( m / sec ) 2, 200 D10 (m ) for (5 m D10 300 m ) 3 (7)
where Gmax(S) Gmax at any degree of saturation, S, Gmax(dry) Gmax under dry condition, a maximum value of Gmax(S)/Gmax(dry); and Sopt optimum degree of saturation (i.e. degree of saturation at a). The optimum degree of saturation, Sopt is a function of effective grain size (D10) (Eq. (5)). More details on the functions, H1(S) and H2(S) are available in Wu et al. (1984). Sopt(%) 6.5 log[D10(mm)] 1.5 (5)
PREDICTION OF ELASTIC AND SHEAR MODULUS OF UNSATURATED FINE-GRAINED SOILS Elastic modulus
3.1
Cho & Santamarina (2001) conducted bender element tests on four different coarse-grained soils to study the influence of degree of saturation on the shear wave velocity, Vs ( Gmax/). One of the soils showed drop of Vs after Sopt, whereas the Vs values of the remainder three soils continuously increased up to dry condition (i.e. curve (ii) in Figure 2(b)). Based on the experimental results, Cho & Santamarina (2001) proposed an empirical equation to compute the shear wave velocity, Vs at a given degree of saturation, S (Eq. (6)). 2 eq Vs ( unsat ) Vs ( sat ) 1 (1 K 0 ) v
Mendoza et al. (2005) carried out bender element tests to study the influence of degree of saturation (i.e. suction) on the elastic and shear modulus of unsaturated fine-grained soils (Ip 38%) for both the drying and wetting paths. They established the relationship between the degree of saturation, S and suction, (ua uw) using compacted specimens. The volume of the specimens was allowed to change during both the drying and wetting paths. Mendoza et al. (2005) provided the explanation for the characteristic behavior of the Gmax with respect to suction (i.e. curve (ii) in Figure 2(c)) as below. Boundary effect zone: sharp increase of the Gmax due to the increase in the density of the soil. Transition zone: gradual increase of the Gmax due to some shrinkage. Residual zone: sharp increase of the Gmax due to the particle displacement and rotation due to the influence of soil structure effects without major volume change. Based on the experimental results, they proposed an empirical equation to predict the variation of the Emax (same units as suction) with respect to suction taking account of volume change as below. Emax ( unsat ) 30000 (2.3 e )2 ln(ua uw )1.35 1 e (8)
e Gs eS Gs
(6)
where Vs(sat), Vs(unsat) shear wave velocity under saturated and unsaturated condition, respectively, eq equivalent effect stress due to suction, empirical parameter, e void ratio, Gs specific gravity; and K0 coefficient of lateral earth pressure at rest. Cho & Santamarina (2001) provided explanations of possible reasons for the decrease of net contribution of matric suction towards Emax and Gmax beyond Sopt for coarse-grained soils (i.e. curve (i) in Figure 2(b)) as follows. The menisci between particles separated by a short distance fail earlier than those in contact, which leads to the gradual decrease of the Emax and Gmax beyond optimum degree of saturation. This behavior is generally observed for the uniformly graded coarse-grained soils (i.e. high coefficient of uniformity) which have relatively larger voids within the soil particles and their aggregations (Mendoza & Colmenares 2006). However, if coarse-grained soils desaturate gradually by air drying, the bonding between particles can increase continuously up to dry condition
Vanapalli & Oh (2010) extended the modulus of elasticity model (i.e. Eq. (2)) which was originally proposed for sandy soils (with Ip 0) towards the estimation of elastic modulus for fine-grained soils (8 Ip (%) 16) using model footing tests and insitu plate load test results (Costa et al. 2004, Rojas et al. 2007, Vanapalli et al. 2007). They suggested fitting parameter 1 and 2 are required for coarse- and fine-grained soils, respectively. Relationships were provided for inverse of (i.e. 1/) as a function of Ip. The upper and lower boundary relationships can be used for high and low matric suction values respectively for soils with different values of Ip (Eq. (9); Figure 3). Typical behavior of
343
30 25 20
Lower boundary Upper boundary
(1 / ) 0.5 0.312( I p ) 0.109( I p )2 (1 / ) 0.5 0.063( I p ) 0.036( I p ) 2
15 10 5
Clayey sand Coarse-grained soils Lean clay
Glacial till
at air-entry value, (ua uw)b, parameter that controls the rate of increase of soil stiffness with respect to suction, (p ua)C mean net stress; and r ratio between Gmax(s*) and the threshold value of Gmax. Mendoza et al. (2005) proposed another equation for estimating the Gmax (same units as suction) with respect to suction as given below (i.e. curve (ii) in Figure 2(c)).
20
1/
10
15
Plasticity index, IP
Gmax 133000
(2.3 e )2 ln(ua uw )1.35 1 e
(12)
Figure 3. Relationship between (1/) and plasticity index, Ip (Vanapalli & Oh 2010).
Oh & Vanapalli (2009a, b) proposed a model which is similar to Eq. (2), to estimate the shear modulus of unsaturated sandy soils (i.e. Ip 0%) (Eq. (13)). u u Gmax( unsat ) Gmax( sat ) 1 a w (S ) P a (13)
the elastic modulus for different soils is similar to the curve (i) in Figure 2(c). (1 / ) 0.5 0.312( I p ) 0.109( I p )2 for (0 I p (%) 12 ) (1 / ) 0.5 0.063( I p ) 0.036( I p )2 for (0 I p (%) 16 ) 3.2 Shear modulus
(9)
Mancuso et al. (2002) investigated the influence of suction on the Gmax using resonant column tests on an unsaturated fine-grained soil (Ip 13.7%). They suggested that Gmax tends towards a threshold value at relatively high suctions with an S-shaped trend (i.e. curve (iii) in Figure 2(c)). The effect of bulkwater on Gmax is dominant in the boundary effect zone; therefore, the gradient of Gmax versus (ua uw) relationship is the same as that of Gmax versus p (mean effective stress) relationship. The effect of menisci-water becomes predominant in the residual zone as progressive shift takes place in the transition zone from bulk-water to menisciwater. Extending the above conceptual approach, they proposed a model to predict the Gmax values in boundary (i.e. (ua uw) (ua uw)b; Eq. (10)) and transition and residual (i.e. (ua uw) (ua uw)b; Eq. (11)) zones as given below. ( p ua )C (ua uw ) m Gmax ( unsat ) A OCR pa
n
where , fitting parameters. After analyzing shear modulus data of different sandy soils obtained using bender element (Picornell & Nazarian 1998, Takkabutr 2006, Lee et al. 2007) and resonant column technique (Kim et al. 2003), they suggested the values for the two fitting parameters, and as shown in Table 1. Oh & Vanapalli (2009b) investigated the relationship between the SWCC and the two fitting parameters, and . The fitting parameter, is related to the range of matric suction over which the SWCC extends. For example, the value is estimated as a relatively high value (i.e. 35) for a sandy soil as the SWCC spreads only in a narrow range of matric suction (i.e. PK-sand) and decreases up to 2.5 as the range of the SWCC range of matric suction increases. The fitting parameter, is related to the slope of the linear portion of the SWCC when plotted as {S(%)/100 log (ua uw)}. For example, the value decreases with an increase in the slope. The values were estimated as either 0.5 or 1 for sandy soils. The representative behavior of the Gmax for the soils studied by them is shown as the curve (i) in Figure 2(c).
Table 1. Fitting parameters, and values for the soils analyzed (Oh & Vanapalli 2009a, b). Ip Picornell & Nazarian (1998) (PK-sand) Takkabutr (2006) (T-sand) Kim et al. (2005) (K-sand) Lee et al. (2007) (L-sand) NP NP NP NP 35 3.5 2.5 2.5 0.5 0.5 1 1
(10)
(u u )(u u ) Gmax ( unsat ) Gmax ( s*) 1 r e a w a w b r
(11) where Gmax(sat), Gmax(unsat) maximum shear modulus under saturated and unsaturated condition, respectively, A stiffness index, m, n stiffness coefficients, OCR overconsolidation ratio, Gmax(s*) Gmax
344
Elastic modulus, Emax (MPa)
RELATIONSHIP BETWEEN THE ELASTIC AND SHEAR MODULUS OF UNSATURATED SOILS
1400
S1
1200 1000 800 600 400 200 0 0

(a)
Measured Predicted ( = 0.1) Predicted ( = 0.3) Predicted ( = 0.5)
The relationship between the elastic and shear modulus for homogeneous, isotropic, and linear elastic materials is given below (Eq. (14)). E 2G(1 ) (14)
Eq. (14) indicates that either elastic or shear modulus can be calculated if one of the moduli is known along with the Poissons ratio, . In the present study, the concept in Eq. (14) is extended to Emax and Gmax values measured for unsaturated soils (Eq. (15)) Emax(unsat) 2Gmax(unsat)(1 ) (15)
20
40
60
80
100

900 800 700 600 500 400 300 200 100 0
(b) S3 Measured Predicted ( = 0.1) Predicted ( = 0.3) Predicted ( = 0.5)
The validity of Eq. (15) was critically analyzed using the Emax and the Gmax values with respect to degree of saturation measured using bender element technique for compacted fine-grained soils specimens (Mendoza et al. 2005). Figure 4(a) to (d) show the comparison between the measured Emax values and those estimated using Eq. (15) for different values (i.e. 0.1, 0.3 and 0.5) (the results for S2 are not shown due to the errors related to the testing). There is a good agreement for the value of 0.1 for entire range of degree of saturation; however, the agreement is not good as value increases. In addition, the influence of on Emax values for the relatively high degree of saturation (i.e. low suction value) was less compared to low degree of saturation (i.e. high suction value). This implies that the Poissons ratio, is not a constant value but instead vary with the degree of saturation of soils. To further examine the relationship between and degree of saturation, the values were backcalculated and plotted against degree of saturation. This relationship is compared with the Emax values in different zones (Figure 5). The negative values and greater than 0.5 can be neglected as they are not realistic (solid circles in Figure 5). A trend line is drawn assuming 0.1 which can be representative value of the relatively dry condition (low S and high suction) based on the comparisons between the measured and the estimated Emax values discussed using Figure 4. The values are almost constant with the values close to 0.5 in boundary effect zone, which indicates that the soils specimens were under undrained condition. The values then decreases with decreasing degree of saturation in the transition effect zone and eventually converges to a constant value in the residual zone. This characteristic behavior of with respect to degree of saturation implies that the variation
10
20
30
40
50
60

1200
S4
1000 800 600 400 200 0

(c)
Measured Predicted ( = 0.1) Predicted ( = 0.3) Predicted ( = 0.5)
20
40
60
80
100

800
S5 Measured Predicted ( = 0.1) Predicted ( = 0.3) Predicted ( = 0.5)
700 600 500 400 300 200 100 0 0

(d)
10
20
30
40
50
Figure 4. Comparision between the measured and estimated elastic modulus, Emax (Mendoza et al. 2005).
of with respect to degree of saturation is closely related to the SWCC. In other words, the value can be expressed as a function of degree of saturation (i.e. f(S)) and the SWCC can be used as a tool to estimate or predict the variation of with respect to degree of saturation. Hence, Eq. (15) can be rewritten as below.
*
Emax( unsat ) 2Gmax( unsat )
1 f (S )
(16)
345
Boundary effect zone
Transition zone
Residual zone
1.0 0.8
0.6 0.4 0.2 0.0 -0.2
100
80
60
40
20
Figure 5. (a) Degree of saturation versus Emax and Gmax relationship and (b) back-calculated Poissons ratio, with respect to degree of saturation (using the results from Mendoza et al. 2005).
SUMMARY
Typical behaviour of the elastic (i.e. Emax) and shear (i.e. Gmax) modulus variation with respect to suction are discussed using the different models available in the literature. In addition, Emax and Gmax values measured using bender element technique (Mendoza et al. 2005) were re-analyzed to investigate the relationship between the two moduli for both saturated and unsaturated conditions. The analyses showed that the Poissons ratio, is a sensitive parameter and is also a function of degree of saturation. Also, the unique relationship between the Gmax and Emax values can be justified for both saturated and unsaturated soils when these values are obtained using similar testing techniques or procedures. The discussions presented in this paper would be of interest and value in the numerical analysis of unsaturated fine-grained soils for obtaining reliable comparisons between the measured and the model behavior. The present study is based on limited data from the literature. More studies are therefore necessary to understand the influence of various other parameters on the elastic (E) and the shear (G) modulus of different unsaturated fine-grained soils. REFERENCES
Cho, G.C. & Santamarina, J.C. 2001. Unsaturated particulate materialsparticle-level studies. Journal of Geotechnical and Geoenvironmental Engineering 127(1): 8496. Costa, Y.D., Cintra, J.C. & Zornberg J.C. 2003. Influence of matric suction on the results of plate load tests performed on a lateritic soil deposit. Geotechnical Testing journal 26(2): 219226. Kim, D.S., Seo, W.S. & Kim, M.J. 2003. Deformation characteristic of soils with variations of capillary
pressure and water content. Soils and Foundation 43(4): 7179. Lee, S.H., Seo, W.S. & Kim, D.S. 2007. Evaluation of modulus of unsaturated compacted soils in various matric suctions using modified volumetric pressure plate extractor. Proc. 60th Canadian Geotechnical Conference, Ottawa, Canada, 2124 October 2007: 10371044. Mancuso, C., Vassallo, R. & dOnofrio, A. 2002. Small strain behavior of a silty sand in controlled-suction resonant columntorsional shear tests. Canadian Geotechnical Journal 39(1): 2231. Mendoza, C.E. & Colmenares, J.E. 2006. Influence of the suction on the stiffness at very small strains. Proc. 4th Int. Conf. on Unsaturated Soils, Carefree, AZ, 26 April 2006: 529540. Mendoza, C.E., Colmenares, J.E. & Merchn. 2005. Stiffness of an unsaturated compacted clayey soil at very small strains. In Tarantino, Romero & Cui (eds), Advanced Experimental Unsaturated Soil Mechanics: 199204. Rotterdam: Balkema. Oh, W.T., Vanapalli, S.K. & Puppala, J. 2009. A semiempirical model for the prediction of modulus of elasticity for unsaturated soils. Canadian Geotechnical Journal 46(8): 903914. Oh, W.T. & Vanapalli, S.K. 2009a. A simple method for predicting the shear modulus of unsaturated sandy soils. Proc. 4th Asia-Pacific Conference on Unsaturated soils, Newcastle, Australia, 2325 November 2009: 739744. Oh, W.T. & Vanapalli, S.K. 2009b. A model for predicting the shear modulus of unsaturated sandy soils. Canadian Geotechnical Journal. (Submitted for review ) Picornell, M. & Nazarian, S. 1998. Effect of soil suction on the low-strain shear modulus of soils. Proc. 2nd International Conference on Unsaturated Soil, 2730 August 1998, Beijing, China: 102107. Qian, X., Gray, D.H. & Woods, R.D. 1993. Void and granulometry: effects and shear modulus of unsaturated soils. Journal of Geotechnical Engineering 119(2): 295314. Rojas, J.C., Salinas, L.M. & Seja, C. 2007. Plate-load tests on an unsaturated lean clay. Experimental Unsaturated Soil. Timoshenko, S. & Goodier, J.N. 1951. Theory of Elasticity. McGraw-Hill Inc., New York. Takkabutr, P. 2006. Experimental investigations on smallstrain stiffness properties of partially saturated soils via resonant column and bender element testing. PhD thesis, University of Texas, Arlington. Vanapalli, S.K., Fredlund, D.G. & Pufahl, D.E. 1999. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Gotechnique 49(2): 143159. Vanapalli, S.K. Oh, W.T. & Puppala, A.J. 2007. Determination of the bearing capacity of unsaturated soils under undrained loading conditions. Proc. 60th Canadian Geotechnical Conference, 2124 October 2007, Ottawa: 10021009. Vanapalli, S.K., Oh, W.T. & Puppala, A.J. 2008. A simple model for predicting modulus of elasticity of unsaturated sandy soils. In Toll et al. (ed.), Unsaturated soils: Advances in Geo-engineering: 503509. Rotterdam: Balkema. Vanapalli, S.K. & Oh, W.T. 2010. A model for predicting the modulus of elasticty of unsaturated soils using the soil-water charateristic curve. International Journal of Geotechnical Engineering. (Accepted for publication) Wu, S., Gray, D.H. & Richart, F.E. Jr. 1984. Capillary effects on dynamic modulus of sands and silts. Journal of Geotechnical Engineering, ASCE 110(9): 11881203.
Poisson's ratio,
E max ,Gmax
346
Shear strength of a pyroclastic unsaturated soil from suction-controlled direct shear tests
R. Papa & M.V. Nicotera
Dipartimento di Ingegneria Idraulica, Geotecnica ed Ambientale, Universit Federico II, Napoli, Italy
ABSTRACT: The paper analyses some experimental results on natural unsaturated pyroclastic specimens carried out by means of direct shear tests both on undisturbed unsaturated and lab saturated samples. The results of suction-controlled tests were interpreted and shear strength envelope recovered. Although an extended Mohr-Coulomb criterion can be expressed in terms of matric suction and net stress, a more satisfactory description of the shear strength of this unsaturated pyroclastic soil was achieved by taking into account both matric suction and saturation degree.
INTRODUCTION
The paper analyses some experimental results of a project on mudflows in pyroclastic soils in the southern Italian region of Campania. Based on geological and geomorphological considerations a test site was selected in order to collect experimental data from laboratory tests on undisturbed samples, as well as monitor climatic conditions (affecting infiltration), matric suction and water content in the subsoil (Papa, 2007; Evangelista et al., 2008). Elsewhere (Papa et al., 2008; Papa and Nicotera, 2008) the results of suction-controlled triaxial tests on undisturbed unsaturated specimens from the test site have been interpreted and the critical state conditions recovered. It emerged that matric suction was not suitable, as an independent stress variable, to describe the shear strength of this unsaturated pyroclastic soil. By contrast, a satisfactory interpretation was achieved by taking into account both matric suction and saturation degree, representing the data in terms of Bishop stress. However txt tests did not allow shear strength to be investigated at low stress levels or at high strains. Therefore a number of direct shear tests both on undisturbed unsaturated and lab saturated samples were performed. 2 MATERIALS AND METHODS
consist of: 1) topsoil; 2) weathered and humified ashy soil; 3) pumices from the Avellino eruption (3.7 ky b.p.); 4) palaeosoil consisting of weathered volcanic ashes; 5) pumices from the Ottaviano eruption (8.0 ky b.p.); 6) palaeosoil consisting of weathered volcanic ashes; 7) volcanic sand; 8) highly weathered fine-grained ashy soil. Mechanical and hydraulic behaviour of these soil were investigated in both unsaturated and saturated conditions (Papa, 2007; Papa and Nicotera, 2008; Papa et al., 2008). In the present paper the investigation focuses on the shear strength of soil 4 alone. The grain-size distribution of soil 4 exhibits limited scatter; the soil is well-graded, ranging from sand to silt with a small clay fraction. As regards water retention properties, soil 4 behaves like coarse-grained material; it has an air entry value of about 68 kPa and starting from saturated conditions becomes almost dry when the applied matric suction reaches about 100 kPa. In all, 10 conventional and 11 suction-controlled direct shear tests were employed to investigate the shear strength of soil 4. (see Table 1 and 2). 2.1 Traditional direct shear tests Traditional direct shear tests were performed both on natural water content and on lab saturated samples (Table 1). The tests on natural water content consisted of a load-controlled vertical compression followed by a displacement-controlled shearing phase (displacement rate of 0.1 mm/hours). Two different values of the vertical stress were considered: 150 kPa and 250 kPa. The rather more complex tests on lab saturated samples consisted of:
The soil samples utilized for the tests were recovered at the test site by means of a thin wall sampler in deep trenches. At the test site the stratigraphic succession can be described as a series of soil layers essentially parallel to the ground surface that
347
Table 1. Direct shear tests on specimens at natural water content (NW) and on saturated specimens (ST). Test type NW # 01 02 03 04 Initial state e 2.725 2.585 2.625 2.671 Sr 0.69 0.82 0.65 0.53 End of consolidation e 2.408 2.509 2.077 1.946 Sr 0.76 0.82 0.80 0.72 v ua (kPa) 150 150 250 250 250 250 250 250 250 75 150 150 250 250 250 250 250 250 250 0.004 0.005 0.006 0.007 0.003 0.007 Saturation e End of shearing stage # e 2.154 2.339 1.873 1.636 1.647 1.690 1.707 1.690 1.669 2.256 2.050 2.153 1.829 1.954 1.396 1.380 1.412 1.450 1.475 Sr 0.78 0.71 0.84 0.74 0.70 0.68 0.67 0.68 0.69 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 1.00 (kPa) 106 110 180 177 196 193 199 197 192 52 93 92 167 172 158 180 184 186 183
1 2 3 4 5 6
ST
01 02 03 04 05 06
2.803 2.788 2.512 2.788 2.521 2.678
0.78 0.72 0.81 0.74 0.66 0.53
2.432 2.275 2.332 2.026 2.202 1.723
1.00 1.00 1.00 1.00 1.00 1.00
1 2 3 4 5
Table 2. Suction-controlled direct shear tests (SC). Test Type SC # 01 02 03 04 05 06 07 08 09 10 11 Initial state e 2.644 2.637 2.471 2.610 2.360 2.596 2.390 2.731 2.516 2.630 2.402 Sr 0.71 0.85 0.87 0.61 0.75 0.82 0.71 0.85 0.71 0.65 0.90 S (kPa) 17 5 6 13 13 8 17 6 6 15 7 End of consolidation e 2.644 2.637 2.471 2.610 2.360 2.596 2.390 2.731 2.516 2.630 2.402 Sr 1.00 0.90 0.83 0.61 0.71 0.77 0.74 0.85 0.72 0.68 0.90 S (kPa) 0 3 12 12 20 20 20 20 6 6 12 6 End of equalization e 2.452 2.517 2.379 2.497 2.261 2.501 1.981 2.001 2.550 2.117 2.245 2.046 Sr 1.00 0.89 0.85 0.69 0.74 0.79 0.88 0.88 0.94 0.86 0.81 1.00 v ua (kPa) 37 37 37 37 37 37 250 37 76 151 151 251 End of shearing e 2.350 2.337 2.313 2.411 2.251 2.354 2.076 2.355 2.002 2.093 1.905 Sr 1.00 0.91 0.85 0.74 0.74 0.83 0.86 1.00 0.84 0.83 0.95 (kPa) 27 32 32 33 42 42 43 65 121 121 161
a load-controlled vertical compression phase; a saturation phase performed by submerging with distilled water the specimen inside the shear box and waiting for the completion of the collapse process; a displacement-controlled shearing phase (displacement rate of 0.1 mm/hours). In this case three different values of the vertical stress were considered: 75 kPa, 150 kPa and 250 kPa. In the experimentation some additional attention was paid to the behaviour of the soil at very large shearing strains. Therefore in the cases of two tests
(NW_04 and ST_06) performed at the same stress level (250 kPa), the first on a natural water content specimen and the second on a lab saturated one, the shearing phases were repeated more than once: at the end of the first shearing stage the shearing box was taken back to its initial position, maintaining the vertical load constantly applied to the specimen; the specimen was then re-sheared. This procedure was repeated five times (6 shearing stages in all) in the case of test NW_04 (see table 1) and four times (5 shearing stages in all) in the case of test ST_06.
348
2.2
Suction-controlled direct shear apparatus
The apparatus (see Figure 1) was designed to test parallelepiped specimens of 60 mm of base side and 20 mm in height. Suction control is based on the axis translation technique. The shear box is located inside a steel cell in which air pressure is applied. The axial and shear loads are applied by means of two independent hydraulic pistons. A submergible LVDT is used to measure the vertical displacements while the relative shear displacements between the upper and lower parts of the shear box are measured by means of an external LVDT activated by a stiff beam connected to the shear piston. The vertical and shear load are measured by means of inner load cells. The water content variations are determined according to Aversa and Nicotera (2002) by measuring with a high accuracy differential pressure transducer the difference in pressure in two double-walled burettes, one of which is connected to the drainage circuit, the other operating as a reference. The system is connected to an air supply at a stable pressure of 1100 kPa and can regulate four pressures independently (i.e. cell pressure, pc, axial pressure, pr, shear pressure ps, and pore water pressure, uw) by means of electro-pneumatic regulators which can be adjusted over a range from 0 to 1000 kPa with an accuracy of 1 kPa. The cell pressure, pc (equal to pore air pressure, ua), is directly applied to the cell and to the air filling the specimen through the upper porous stone (with a low air entry value). The pore water pressure is applied to the specimen
through the lower porous stone (with a high air entry value); a system for measuring water content changes is located in the circuit, between the regulator and the porous stone. 2.3 Suction-controlled direct shear tests
Direct shear tests on unsaturated undisturbed specimens were performed at values of normal net stress ranging from 37 kPa to 251 kPa. Matric suction values for suction-controlled direct shear tests were selected in order to study mechanical behaviour during the transition between fully saturated and partially saturated conditions. Each test consisted of the following phases: measurement of the initial matric suction by means of the axis translation technique; equalisation to an assigned matric suction value; vertical compression at constant suction; shearing at constant suction and constant normal net stress (Table 2). In one case (test 07) the specimen was overconsolidated by means of a loading-unloading cycle before shearing. Suction measurements lasted about 24 hours while equalisation phases took from 48 to 72 h. The vertical compressions were completed at a loading rate of 5 kPa/hour. The shearing phases were performed in displacement-controlled conditions at a rate of 0.1 mm/ hour. The shearing rate was selected by scaling down the optimal displacement rate determined for saturated tests on the basis of a trial and error procedure (Papa 2007).
Figure 1. Scheme of pressure regulation circuits and transducers used in the suction-controlled direct shear apparatus.
349
Results of direct shear tests on lab saturated samples (ST) are summarized in Figure 2a, c and e. In particular in Figure 2a and c the effective stress ratio / and the dilation rate dv/du are plotted against shear displacement u; in Figure 2e effective stress ratio is measured against dilation rate. In each diagram the thin continuous lines refer to conventional tests (first stage); the dashed lines refer to the intermediate stage of the multistage test; the thick continuous line refers to the last stage of the multistage test. It is worth noting that in all tests during the first shearing stage the soil had a ductile behaviour; furthermore at the end of the first shearing stage the saturated specimens still showed a highly contractive behaviour. Hence the stationary or critical shear
strength was never completely mobilized at the end of this stage. The final effective stress ratios at the end of the tests ranged between 0.614 and 0.697 with an average of 0.653 (corresponding to a mobilized effective friction angle of 33.16). By contrast, during the subsequent shearing stage the effective stress ratio peaked against the shear displacement curve, followed by strain softening; moreover dilation rate plots revealed an initially slight dilative behaviour, rapidly reducing as the shearing process approached constant volume conditions. These considerations are clearly validated by the diagram in Figure 2e; in the dilation rate versus effective stress ratio plane all tests converge toward a unique point (/ 0.734 corresponding to a friction angle of 36.26) on the ordinate axis representative of stationary or critical shearing conditions. In particular, the second and
Figure 2. Traditional direct shear tests: tests on saturated specimens (ST), a) effective stress ratio, c) dilation rate and e) dilation rate versus effective stress ratio; tests on natural water content specimens (NW), b) total stress ratio, d) dilation rate and f) dilation rate versus effective stress ratio.
350
Figure 3. Suction control direct shear tests: a) average skeleton stress ratio; b) dilation rate versus average skeleton stress ratio; c) dilation rate; d) Interpolation of experimental data with Fredlund Morgenstern & Widger (1978) shear strength criterion.
subsequent shearing stages reached these conditions while the first stage and the single stage tests did not. The results of direct shear tests on natural water content specimens (NW) reported in Figure 2b, d and f are almost similar to the previously described ones. However, in this case the net stress ratio / was employed instead of the effective one. Some differences can be observed between ST and NW tests. Although they behaved in a rather similar way to lab saturated ones, natural water content specimens were less contractive and ductile. The net stress ratios at the end of the first shearing stage range between 0.708 and 0.736 with an average of 0.719 (corresponding to a mobilized effective friction angle of 35.70).The critical value of the net stress ratio approached in NW tests (/ 0.781 corresponding to a friction angle of 37.99) is higher than those ones ST tests; this difference is probably due to the effect of matric suction, but this point will be better discussed below. According to Papa and Nicotera (2008) the shear strength of the particular unsaturated pyroclastic soils of concern was investigated by taking into account both matric suction and degree of saturation combined in the Bishop stress as originally proposed by Jennings (1960) and subsequently adopted by others (e.g. Jommi, 2000; Gallipoli et al. 2003). Hence the results of suction-controlled
direct shear tests (SC) were interpreted in terms of average skeleton stresses:
Sr s
(1)
In Figure 3 SC test results are summarised. The investigated soil showed a ductile and contractive behaviour in all SC tests with the exception of test 7 performed on an over-consolidated specimen. The diagram in Figure 3b clearly demonstrates that as for ST and NW tests the paths described in the dilation rate versus average skeleton stress ratio plane converge toward a unique point (/ 0.751 corresponding to a friction angle of 36.91) on the ordinate axis. It is worth noting that due to the different operating range of the suction-controlled direct shear apparatus the final values of shear displacement in the SC tests exceeded those in ST and NW tests; it follows that at the end of SC tests the dilation rates were smaller than those in ST and NW tests and hence the critical shear strength was almost completely mobilized unlike what happened in these last cases. 4 SHEAR STRENGTH CRITERION
Two different approaches can be adopted to analyze the experimental data in order to derive the
351
shear strength criterion of the investigated soil in both saturated and unsaturated conditions. The first approach according to Fredlund et al. (1978) consists in distinguishing the contribution to the shear strength due to matric suction from that due to the net stress. Hence the classical expression of the extended Mohr-Coulomb criterion can be rearranged as follows (in the hypothesis of a zero cohesion intercept): s tan tan b (2)
is of concern, while for modeling peak strength conditions an approach with two independent stress variables is likely to be more suitable. REFERENCES
Aversa, S. & Nicotera, M.V. 2002. A triaxial and oedometer apparatus for testing unsaturated soils. Geotechnical Testing Journal, GTJODJ, 25(1):315. Evangelista, A., Nicotera, M.V. & Aversa, S. 2004. Unapparecchiatura di taglio a suzione controllata per lo studio delle condizioni di innesco dei fenomeni franosi nelle coltri piroclastiche. Proceedings of the XXII Convegno Nazionale di Geotecnica, Palermo: 8188. Patron, Bologna. Evangelista, A., Nicotera, M.V., Papa, R. & Urciuoli G. 2008. Field investigation on triggering mechanisms of fast landslides in unsaturated pyroclastic soils. In:Unsaturated Soils: Advances in Geo-Engineering. 1st European Conference on Unsaturated soil. Durham, UK. 24 July 2008. (pp. 909915). LONDON: Taylor & Francis Group plc (UK). Fredlund, D.G. & Morgenstern, NR. 1977. Stress state variables for unsaturated soils. Journal of the Geotechnical Engineering Division (ASCE), 103(GT5): 447466. Fredlund, D.G., Morgenstern, N.R & Widger 1978. The shear strength of unsaturated soils. Can. Geotech, J. 15: 313321. Gallipoli, D., Gens, A., Sharma, R., & Vaunat, J. 2003. An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Gotechnique 53(1): 123136. Jennings, L.E. 1960. A revised effective stress law for use in the prediction of the behaviour of unsaturated soils. Pore Pressure and Suction in Soils. London: Butterworths. Jommi, C. 2000. Remarks on the constitutive modelling of unsaturated soils. In A. Tarantino & C. Mancuso (eds.) Experimental evidence and theoretical approaches in unsaturated soils: 139153. Rotterdam: Balkema. Papa, R. 2007. Indagine sperimentale sulla coltre piroclastica di un versante della Campania. PhD thesis, Universit di Napoli Federico II. Papa, R., Evangelista, A., Nicotera, M.V. & Urciuoli G. 2008. Mechanical properties of unsaturated pyroclastic soils affected by fast landslide phenomena. In:Unsaturated Soils: Advances in Geo-Engineering. 1st European Conference on Unsaturated soil. Durham, UK. 24 July 2008. (pp. 917923). LONDON: Taylor & Francis Group plc (UK). Papa, R. & Nicotera, M.V., 2008. Critical state of an unsaturated pyroclastic soils. Proc. of 3rd International Workshop on Between theory and Practice in Unsaturated Soil Mechanics. Trento, IT. 46 February 2008. Millpress Science Publisher, Rotterdam (in press). Tarantino, A. & Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique 55(4): 307317. Tarantino, A. 2007. A possible critical state framework for unsaturated compacted soils. Gotechnique 57(4): 385389.
In Figure 3d all the data (end of shearing point) from SC tests are reported in s/ versus / plane. The regression line of the experimental data is also reported ( 36.22; b 35.20). It results that the intercept of the regression line represents an upper boundary for the / ratios from ST tests. However, the fit of the experimental data is rather crude (R2 0.868). The second possible approach consists in assuming that the shear strength criterion is expressed in terms of average skeleton stress:
tan
(3)
Hence the value of the friction angle should be that estimated by representing the data either from SC tests ( 36.91) or from ST tests ( 36.26) in the dilation rate versus average skeleton stress ratio. In both cases the estimated friction angle practically coincides with that inferred by Papa and Nicotera (2008) from suction-controlled triaxial tests. This second approach is more satisfactory in terms of data fitting and in some sense is more physically based as observed by some authors, albeit referring to quite different soils and stress levels (Tarantino & Tombolato 2005, Tarantino 2007). 5 CONCLUSIONS
By performing direct shear tests on undisturbed specimens we investigated the shear strength in both saturated and unsaturated conditions of one of the soils at the test site. The soil behaved in a highly ductile and contractive fashion in all types of performed tests (i.e. ST, NW and SC tests). Therefore shear strength has to be evaluated taking into account the dilation rate at the end of each test. As regards the shear strength criterion, a model based on two independent stress variables was compared to one expressed in terms of average skeleton stress. This second approach seems to be preferable. However, it must be pointed out that these conclusions are valid only if the critical shear strength
352
Experimental study of the collapsible behavior of a tropical unsaturated soil

R.A. Rodrigues & O.M. Vilar
University of So Paulo, So Carlos-SP, Brazil
ABSTRACT: This paper presents wetting induced (or collapse) deformation of an undisturbed sandy soil from Brazil. In this investigation, the main testing program has been performed using oedometric tests especially designed to apply and control the suction through axis-translation technique. Two types of test have been carried out: tests in which a combination of loading paths at constant suction were applied; collapse tests under constant loading in order to determine the collapse potential during wetting. Also, the effect of wetting-drying cycle on the collapse behavior was studied.
INTRODUCTION
2 2.1
MATERIAL AND METHODS Soil properties
Some soils under constant stress show volume decrease related to an increase of water content. This wetting-induced deformation or soil collapse is a typical behavior of the so called collapsible soils which are characterized by an unstable structure. The analyses of collapsible behavior of soils are commonly undertaken using conventional oedometric test results in which a soil, at certain moisture content, is firstly loaded and then wetted. However, some evidence from laboratory tests and field observation indicate that soil collapse can arise from moisture variation or, more appropriately, from suction variation conditions that are far from saturation or far from zero suction. Laboratory evidence indicates that deformations due to collapse may occur due to the gradual increase in moisture content or gradual reduction in soil suction and wetting-drying cycles, as shown by Escario and Saez (1973), Vilar and Davies (2002), among others, when using controlled soil suction tests. In this paper selected results from an experimental program specifically undertaken to identify the collapse mechanism on a tropical sandy soil are presented. The study rests on conventional and suction controlled tests that are analyzed considering some fundamental concepts of the mechanics of unsaturated soils. The influence of the stress state and suction on volume change was studied. Also, the influence of the initial soil suction and wetting-drying cycle on the collapse potential were investigated.
The soil used in this study is a typical collapsible soil from the Northwest of Sao Paulo State, Brazil. Large areas of this region are covered by sandy soils that are of colluvial nature and reach depths of about 10 meters. These soils present low units weights and low degree of saturation. The main characteristics of the studied soil are: specific gravity, s 2.63 g/cm3; dry density, d 1.44 g/cm3; water content, w 7.2%; void ratio, e 0.845; porosity, n 46%; liquid limit, wL 18%; plastic limit, wP 11%; clay fraction 15%; silt fraction 6% and sand fraction 79%. The soil classify as SC in the Unified Soil Classification System (ASTM 2003a). 2.2 Conventional oedometric tests A total of eight oedometric tests were performed, of which two tests were with constant water content (double oedometer tests) and six tests with collapse induction. After applying the vertical stress of 40, 80, 160, 320, 640 and 1280 kPa and allowing the equilibrium of deformation, the specimens were wetted and deformations developed during this wetting stage were registered. The wetting-induced or collapse deformations (CP) were calculated by Equation 1 from Jennings and Knight (1975): CP e 100% 1 ei (1)
353
Patm
vertical load
1,0 0,9 0,8
liquid oil
in situ moisture content
burette specimen
e/eo
air pressure
0,7 0,6 0,5
wetted
coarse porous stone
water pressure
0,4 0,3 1 10 100 1000 10000
high air entry pressure porous plate
(kPa)
Figure 1. Oedometer with suction control by axis translating technique.
Figure 2. tests.
1,0 0,9 0,8
Compression curves in double oedometric
Where, e: void ratio variation of the specimen after collapse; ei: void ratio of the specimen before the wetting.
e/eo
2.3
Suction controlled oedometric tests
0,7 0,6 0,5 0,4 0,3 1 10 100 1000 10000
Suction controlled oedometric tests were performed using the axis translating technique (Hilf, 1956) in a cell similar to that developed by Escario and Saez (1973), according to Figure 1. These tests were undertaken using specially designed oedometers where a 500 kPa air-entry value ceramic disk was incrusted in cell base. Soil specimen dimensions were 70 mm in diameter and 20 mm in height. The specimens were first wetted to release suction and then subjected to an initial suction of 60 kPa, that is associated to the average in situ moisture content, following a drainage path. After equilibrium, they were loaded under net vertical stress of 50, 100, 200 and 400 kPa allowing for the equilibrium of deformation under load. Then the suction was gradually reduced to 60, 50, 25, 10 and finally to 0 kPa and the corresponding collapse strains and volume of water that entered the soil were registered. To study the influence of suction on soil compression (volume change) others specimens were monotonically loaded under constant soil suction, that varied between 0 and 400 kPa. 3 3.1 ANALYSIS AND DISCUSSION OF THE RESULTS Conventional oedometric tests
(kPa)
40 kPa 80 kPa 160 kPa 320 kPa 640 kPa 1280 kPa
Figure 3. tests.
Soil collapse in conventional oedometric
Figure 2 shows the results of double oedometric tests normalized through the initial void ratio (eo), while Figure 3 shows the results of collapse deformation in individual tests. As can be seen, in both kind of tests, collapse deformation increases upon to a maximum and then decreases with load. Test results of collapse deformation of the two types of tests are close, as can be seen in Table 1. Maximum collapse was of the order of 5% and was related to vertical stress of 80 kPa. 3.2 Suction controlled oedometer tests
The compression curves from the conventional oedometric tests are shown in the Figures 2 and 3.
Figure 4 shows the soil compression curves from oedometric tests with suction control. As it is usual, suction tends to increase preconsolidation stress. In these tests, the compression index was not significantly affected, however it tended to increase with suction as shown in Figure 5.
354
Table 1. Collapse potential values obtained in the conventional oedometric tests. Collapse potential Vertical stress 40 kPa 80 kPa 160 kPa 320 kPa 640 kPa 1280 kPa Double 3.4% 5.2% 4.7% 3.8% 2.8% 1.2% Individual 3.4% 5.1% 4.1% 2.5% 1.8% 0.3%
Cc
0,330 0,320 0,310 0,300 0,290 0,280 0,270 0,260 0,250 0 100 200 300 400
1,1 1,0 0,9 0,8
s (kPa)
Figure 6. Compression index against suction for vertical stress lower than 200 kPa.
e/eo
0,7 0,6 0,5 0,4 0,3 1 10 100 1000 10000
1,0 0,9 0,8
e/eo
- ua (kPa)
s = 0 kPa s = 100 kPa s = 50 kPa s = 400 kPa s = 75 kPa
0,7 0,6 0,5 0,4 0,3
Figure 4. tests.
0,280
Soil collapse in suction controlled oedometric
10
100
1000
10000
- ua (kPa)
50 kPa
160
100 kPa
200 kPa
400 kPa
0,270 120 0,260
Preconsolidation stress (kPa)
Compression index
Figure 7. tests.
Soil collapse in suction controlled oedometric
0,250 0,240 0,230 0,220 0 100 200 300
80
Preconsolidation stress
40
0 400
s (kPa)
Figure 5. Compression index and preconsolidation stress against suction for the complete range of vertical stress up to 1600 kPa.
This behavior is not in total accordance with BBM model (Alonso et al. 1987; 1990), which considers that Cc(s) reduces as suction increases. However if the low values of stress are considered, that is, until 200 kPa, it can be seen in Figure 6
that Cc (s) tended to decrease with suction until an asymptotic value. The set of test results shown in Figure 7 were performed on samples with initial suction of 60 kPa. Each specimen had the suction reduced at known loads to measure the soil collapse deformation. Figure 7 and Table 2 shows the magnitude of measured collapse deformation in these tests, where it can be observed that the maximum collapse took place at 100 kPa of vertical stress and that the collapse tends to reduce as the stress is increased. A comparison between data from Tables 1 and 2 (plotted in Figure 8) show that the collapse potentials of specimens obtained in the suction controlled oedometric tests are slightly different from that obtained from a conventional oedometric tests, although both curves show the same pattern of behavior (maximum collapse).
Cc
355
Table 2. Collapse potential values obtained in the suction controlled oedometric tests. Net vertical stress 50 kPa 100 kPa 200 kPa 400 kPa Collapse potential 2.8% 5.8% 4.8% 3.6%
CP (%)
6 5 4 3 2 1 0 0 20 40 60 80 100 120
50 kPa
wetting-drying cycle
7,0 6,0 5,0
s (kPa)
Suction controlled oedometric tests
Figure 9. 50 kPa.
6 5 4
Specimen wetted under net vertical stress of
CP (%)
4,0 3,0
100 kPa
CP (%)
2,0 1,0 0,0 0 100 200 300 400 500 600 700
3 2 1 0 0 20 40 60 80 100 120
Conventional oedometric tests
ua (kPa)
Figure 8. Collapse potential measured on wetting under different vertical stress in the oedometric tests. Figure 10. 100 kPa.
s (kPa)
The collapse potential obtained in the tests carried out using conventional oedometric tests is a consequence of the fact that suction varies during the stress increments application. It is worth to note that the collapse deformations basically depend on the initial suction and load applied, explaining this difference. In Figure 9 to 12, the collapse potentials are plotted as function of the suction. In these figures, the progressive reduction of the soil suction, the wetting and drying cycle and corresponding collapse development can be seen. The sample tested under net vertical stress of 50 kPa (Figure 9) presented an almost negligible collapse potential when suction was decreased from 60 to 10 kPa. The largest part of collapse developed when the suction was reduced from 10 to 0 kPa. On the other extreme, a rather different behavior is observed. For the sample tested under net vertical stress of 400 kPa (Figure 12), appreciable collapse potential begins to develop for suction lower than 25 kPa. Thus, the soil is capable of supporting the suction reduction until a limit value, from which the collapse deformation begin to develop more intensively changing the soil structure.
6 5 4
200 kPa
CP (%)
3 2 1 0 0
20
40
60
80
100
120
s (kPa)
Figure 11. 200 kPa.
It can be observed in Figures 9 to 12 that the specimens previously submitted to a wettingdrying cycle did not present changes on collapse deformation development. In the wetting-drying cycle, the deformations produced by the new wetting (suction varying between 30 and 10 kPa) are insignificant.
356
6 5 4
400 kPa
REFERENCES
Alonso, E.E., Gens, A., and Hight, D.W. 1987. Special problem soils. In Groundwater Effects in Geotechnical Engineering: Proceedings of 9th European Conference on Soil Mechanics and Foundation, Dublin, 31 August3 September 1987. A.A. Balkema, Rotterdam, The Netherlands. Vol. 3, pp 10871146. Alonso, E.E., Gens, A., and Josa, A. 1990. A constitutive model for partially saturated soil. Gotehnique, 40(3): 405430. ASTM. 2003a. Standard Practice for Classification of Soils for Engineering Purposes (Unified Soil Classification System)D248700. In Annual Book of ASTM Standards, Vol. 04.08. American Society for Testing Materials, West Conshohocken, Pa. Escario, V. and Sez, J. 1973. Gradual collapse of soils originated by a suction decrease. In Proceedings of 8th International Conference on Soils Mechanics and Foundation Engineering, 611 August 1973, Moscow, v. 4.2, pp. 123124. Hilf, J.W. 1956. A investigation of pore-water pressure in compacted cohesive soils. Ph.D. thesis, Faculty of Graduate School of the University of Colorado. Jennings, J.E. and Knight, K. 1975. A guide to construction on or with materials exhibiting additional settlement due to collapse of grain structure. In Proceedings of 6th Regional Conference for Africa on Soil Mechanics end Foundation Engineering, September 1975, Durban, South Africa, Vol. 1, pp. 99105. Vilar, O.M. and Davies, G.I. 2002. Collapse behavior analysis of a clayey sand using different testing procedures. In: Proceedings of 3th International Conference on Unsaturated Soils, 1012 March 2002, Recife, Brazil, Vol. 2, pp. 571576.
CP (%)
3 2 1 0 0 20 40 60 80 100 120
s (kPa)
Figure 12. 400 kPa.
CONCLUSION
An experimental program using oedometric tests has been carried out to determine the effect of suction and collapse development in a tropical collapsible sandy soil. The results show that collapse deformations depend upon net vertical stress and suction. The soil has shown maximum collapse strains at about 100 kPa of stress, after which they have decreased and approached negligible values. The larger the stress, the more distributed is the collapse strain with suction. For low stress, significative collapse deformation starts to develop only at low suction. However, the largest deformation takes place at zero suction, whatever the net vertical stress. The cycling of suction did not introduce additional strain in the specimens.
357
Swelling pressure evolution in compacted bentonite: Experiments and modelling

M. Snchez
Texas A&M University, College Station, USA
M.V. Villar & R. Gmez-Espina

CIEMAT, Madrid, Spain
A. Lloret & A. Gens

UPC, Barcelona, Spain
ABSTRACT: This work presents experimental results about the saturation of bentonite under constant volume and the simultaneous development of swelling pressure. The tests have been performed at room temperature under conditions of free availability of liquid water. Water intake and swelling pressure development were measured. The tests have been performed with a bentonite compacted to dry densities between 1.4 and 1.7 g/cm3. The results obtained show that swelling pressure does not develop in a uniform way, but present more that one swelling stage. After a sharp initial increase, there is a period of time in which pressure increases more slowly orfor the lower densitieseven decreases. This behaviour has also been detected in other clays and can be explained as a consequence of the complex fabric of this material, with the typical double porosity observed in expansive clays. A double structure framework specifically developed to analyse the behaviour of swelling clays has been used in this work to model the swelling pressure evolution of the FEBEX bentonite.
INTRODUCTION
The behaviour of expansive soils is potentially very complex owing to the interaction between the volume change of aggregates made up of highly expansive clay minerals (microstructure) and the rearrangement of the granular-like skeleton formed by the aggregates (macrostructure). The possible use of expansive clays as buffer material for nuclear waste disposal has emphasised the interest on the study of these materials. In this context, the understanding of the main phenomena involved in the evolution of swelling pressure upon saturation of compacted clays under confined conditions is a matter of great interest. This work presents experimental results about the saturation of FEBEX bentonite under constant volume and the simultaneous development of swelling pressure. The tests have been analysed with a double structure model for expansive clays (Snchez et al. 2005). A fundamental characteristic of the double structure framework is the explicit distinction of the two structural levels present within the material: the macrostructural level, which accounts for the larger scale structure of the material and the microstructural level, associated with the active
clay responsible for the swelling behaviour. In addition, the model considers the interaction between the two structural levels. This is a key mechanism to describe the behaviour of swelling clays. The constitutive laws are able to model key aspects of the complex behaviour of highly expansive material, such as large swellings under wetting, yielding, stress path dependency, clay-fabric changes. Even though the model is general, it has been mainly applied to explain and reproduce the behaviour of expansive clays used as engineered barrier to isolate high-level radioactive waste (HLW), which is the context of the research reported herein. 2 2.1 MATERIAL AND METHODS Material
The tests have been performed with the FEBEX bentonite, which was selected by ENRESA (the Spanish Agency for Radioactive Waste Management) as suitable material for the backfilling and sealing of HLW repositories (ENRESA 2006). The FEBEX bentonite was mined from the Cortijo de Archidona deposit (Almera, Spain).
359
The processing of the material at the factory consisted in disaggregation and gently grinding, drying at 60C and sieving to discard particle sizes larger than 5 mm. The FEBEX bentonite has a content of montmorillonite higher than 90 percent. In addition, it contains variable quantities of quartz, plagioclase, K-feldspar, calcite and opal-CT (cristobalitetrydimite). The cation exchange capacity (CEC) varies from 96 to 102 meq/100 g, and the major exchangeable cations are: Ca (3542 meq/100 g), Mg (3132 meq/100 g), Na (2427 meq/100 g) and K (23 meq/100 g). The liquid limit of the bentonite is 1024 percent, the plastic limit is 533 percent, the specific gravity 2.700.04, and 673 percent of particles are smaller than 2 m. The hygroscopic water content in equilibrium with the laboratory atmosphere is 13.71.3 percent. The value obtained for the external specific surface area is 323 m2/g and the total specific surface is about 725 m2/g. The hydraulic conductivity of the saturated FEBEX clay can be related to dry density through an exponential law. The values of permeability to deionised water for densities around 1.6 g/cm3 are in the order of 1014 m/s. The swelling pressure (Ps, MPa) of samples manufactured by compacting the clay with its hygroscopic water content at different dry densities (d, g/cm3) and saturated with deionised water, determined in standard oedometers, can be related to dry density through: ln Ps 6.77 d 9.07 (1)
Figure 1.
High-pressure oedometer.
The maximum difference between the experimental values and the fitting of this equation is 25 percent. 2.2 Methodology
The evolution of swelling pressure upon saturation in compacted and confined cylindrical samples of bentonite has been studied in highpressure oedometers (Figure 1). The granulated clay was compacted uniaxially and statically in the oedometer ring, which had an inner diameter of 5.0 cm, the length of the resulting specimen being 1.2 cm. Nominal dry densities of 1.50, 1.60 and 1.70 g/cm3 were reached by applying vertical stresses of 10.01.3, 16.01.5 and 29.11.0 MPa, respectively. The specimens thus obtained were confined between porous filters. The samples were hydrated at constant volume through the bottom surface with deionised water injected at a pressure of 0.6 MPafor most of the samples of dry density 1.60 g/cm3or 0.01 MPa for the other samples, while the upper
outlet remained open to atmosphere. At the same time, the load cell measured the swelling pressure exerted by the clay. The small vertical deformation of the specimen, due mainly to the load cell and frame deformability, was measured by LVDTs. An automatic volume change apparatus measured the water intake of the specimen. The values of load, strain and water intake were automatically recorded. Once the sample was completely saturated (which is assumed by the stabilisation of swelling pressure development), the injection of water was stopped, and the pressure recordes is considered the swelling pressure value for that particular dry density. The actual density differed slightly from the nominal one due to the small displacement allowed by the equipment (about 0.1 mm when the vertical stress is 1 MPa and 0.4 mm when the vertical stress is of 7 MPa, i.e. between 1 and 3 percent of the sample height). The tests were performed at temperatures between 25 and 30C. On completion of the tests, the equipments were disassembled, the samples were weighed and measured and their final water content determined by drying at 110C for 24 hours. 3 EXPERIMENTAL RESULTS
One test was performed for the nominal dry density of 1.40 g/cm3, six tests for the nominal dry density of 1.50 g/cm3, five tests for the nominal dry density of 1.60 g/cm3, and four tests for the nominal dry density of 1.70 g/cm3. The average values of dry density at the end of the process
360
of saturation were in fact 1.49, 1.57 and 1.65 g/ cm3, due to the small displacement allowed by the equipment during saturation and to initial differences with respect to the nominal value caused during manufacturing of the specimens (Villar & Gmez-Espina, 2009). The curves describing the evolution of swelling pressure over time in some of the tests are plotted in Figure 2. Swelling pressure started to develop as soon as the water came in contact with the clay, but it did not develop in a uniform way. After a sharp initial increase, there was a period of time in which pressure increased more slowly orin the case of nominal densities 1.5 and 1.6 g/cm3even decreased. This behaviour was more apparent in samples of lower dry density whose swelling pressure was lower. The final swelling pressure values are plotted in Figure 3 as a function of dry density. The
9.0 8.0
Swelling pressure (MPa)
line resulting from Equation 1, corresponding to the values obtained at laboratory temperature in standard oedometers, has also being plotted. The dispersion of data can be mostly attributed to the variations in dry density, as the swelling pressure value is very sensitive to small density changes. 4 DOUBLE STRUCTURE MODEL
Expansive clays generally present a marked double structure, made up from clay aggregates and large pores (e.g. Pusch, 1982). The larger voids would correspond to the inter-aggregate pores. The pore space inside the aggregates is constituted by voids of a much smaller size. The two dominant pores size could be associated with two basic structural levels The macrostructure, related to the global arrangements of clay aggregates (the skeleton of the material), with macropores between them. The microstructure, which corresponds to the active clay minerals and their vicinity (inside the clay grains). Only these two basic structural levels identified above are considered herein. The approach is open enough and it could be extended to include more structural levels in the analysis, if it deemed relevant. The soil fabric plays a crucial role to understand and to reproduce the behaviour of expansive clays. In this model, the inclusion of the clay fabric in the analysis is considered in the definition of laws for: 1) the macrostructural level, 2) the microstructural level, and 3) for the interaction between both structural levels. 4.1 Macrostructural model
7.0 6.0 5.0 4.0 3.0 2.0 1.0 0.0 0.01
Dry density (g/cm ) 1.7 1.6 1.5 1.4
0.1
1 10 Time (h)
100
1000
Figure 2. Evolution of swelling pressure in infiltration tests performed with bentonite compacted at different nominal dry density.
9.0 Swelling pressure (MPa) 8.0 7.0 6.0 5.0 4.0 3.0 2.0 1.0 1.30 1.40 1.50 1.60 1.70
The inclusion of this structural level in the analysis allows the consideration of phenomena that affect the skeleton of the material, for instance deformations due to loading and collapse. The BBM (Barcelona Basic Model) has been adopted to describe the macrostructural behaviour (Alonso et al., 1990). The BBM considers two independent stress variables to model the unsaturated behaviour: the net stress () computed as the excess of the total stresses over the gas pressure, and the matric suction (sM), computed as the difference between gas pressure and liquid pressure (pg pl) in the macropores. Figure 4a shows the BBM yield surface (FLC), defined as: FLC g 2 3J M p ps p0 p 0 g 30
2 2
Dry density (g/cm3)
Figure 3. Swelling pressure values obtained in the highpressure oedometers. The line corresponds to Equation 1.
(2)
361
a)
which corresponds to the initial (first stage) swelling observed on Figure 2. In order to describe this expansive behaviour a dependence of the elastic compression index on suction has been considered, as follows: s 0.1 kMi ( s ) kMio 1 i log M 0.1 where i is a constant. 4.2 Microstructural model
Current state
(4)
b)
0 p
The microstructure is the seat of the basic physicochemical phenomena occurring at clay particle level. The strains arising from microstructural phenomena are considered elastic and volumetric (Gens & Alonso, 1992). The microstructural effective stress is defined as: p sm p (5)
al ur ct n ru io st ss ro re ic p M com
0 p
icr sw ostr ell uc in tu g ra
where sm is the microstructural suction. It is assumed that there is no hydraulic equilibrium between the water potentials of both structural levels. The mass transfer of water (w) between the two pore levels is controlled by the following law: w ( sM sm ) (6)
Figure 4. a) BBM yield surface, b) Microstructural load directions on the p-s plane.
where M is the slope of the critical state line, po is the apparent unsaturated isotropic pre-consolidation pressure, g( ) is a function of Lodes angle and ps considers the dependence of shear strength on suction. A basic point of the model is that the size of the yield surface increases with matric suction. The trace of the yield function on the isotropic p-s plane is called LC (Loading-Collapse) yield curve, because it represents the locus of activation of irreversible deformations due to loading increments or wetting collapse. The position of the LC curve is given by the pre-consolidation yield stress of the saturated * state, p0 (hardening variable), according to:
* * p 0 p0
were is the leakage parameter. The increment of microstructural strains is expressed as: m s p p m Km Km Km (7)
( 0 )
1 e
p v p
(3)
v is the volumetric plaswhere e is the void ratio, tic strain, is the elastic compression index for changes in p and (0) is the stiffness parameter for changes in p for virgin states of the soil in saturated condition. In this paper it has been assumed that the more accessible clay particles that compound the macrostructure can swell when wetted,
where the subscript m refers to the microstructural level, the subscript v refers to the volumetric component of the strains and Km is the microstructural bulk modulus. The microstructural stiffness is calculated following a similar approach to the one adopted for the macrostructural level (i.e. using the law (4) for the elastic compression index), but now in terms of the microstructural suction. The Neutral Line (NL) corresponds to constant p and no microstructural deformation occurs when the stress path moves on the NL (Figure 4b). The NL divides the p-s plane into two parts, defining two main generalized stress paths, which are identified as: MC (microstructural contraction) and MS (microstructural swelling). 4.3 Interaction between macro and microstructure In expansive soils there are other mechanisms in addition to the ones included in the BBM which
362
induce plastic strains. This irreversible behaviour is ascribed to the interaction between the macro and micro structures (Gens & Alonso, 1992). It is assumed that the microstructural mechanical behaviour is not affected by the macrostructure but the opposite is not true. An assumption of the model is that the irreversible deformations of the macrostructure are proportional to the microstructural strains according to interaction functions f. The plastic macrostructural strains are evaluated by the following expression:
p p v vLC vm f
(8)
in Section 3. The aim of this paper is not to reproduce all the tests, but to show the capabilities of the proposed model to describe the main features of soil/behaviour observed during the hydration of the FEBEX bentonite. The simulation of the tests has been made solving a boundary value problem using a 1-D model. Table 1 contains the main parameters adopted in the modelling. According to Gens & Alonso (1992), in this framework, the dependence of swelling on initial density can be taken into account through * the parameter p0 . It is assumed that the water
Table 1. Model parameters. s (0) r (MPa1) i 1.371
p is the plastic strains induced by the where vLC yielding of the macrostructure (BBM). In fact the coupling is given by p0*, the hardening variable of the macrostructure (Figure 4a), which depends on the total plastic volumetric strain (Equation 3). In this way it is considered that the microstructural processes can affect the global arrangements of aggregates. More details can be found in Snchez et al. (2005).
Parameters defining the BBM (macrostructure) 1.251 1.32 51 8.21 80. Parameters defining microstructural behaviour (emicro 0.10) 8.0e2 i 1.751 1 Interaction function fS 0.20 1.0 1 pr /p0 p0* (MPa) d 1.6 (Mg/m3) e macro 0.631 p0* (MPa) d 1.5 (Mg/m3) e macro 0.749 p0* (MPa) d 1.4 (Mg/m3) e macro 0.885 k0 1.0 1021 m2
3
4.4
Permeability model
Advective fluxes are computed using the generalized Darcys law, expressed as: q l Kl Pl l g (9)
p0* 10.0 (MPa) p0* 7.5 (MPa) p0* 5.5 (MPa) b 50
where P is the phase pressure, K is the permeability tensor of phase and g is the gravity vector. The permeability tensor is evaluated according to Kl k krl l (10)
-6
-5
Swelling Pressure (MPa)
where k is the intrinsic permeability tensor, is the dynamic viscosity of the phase and kr is the phase relative permeability. The dependence of intrinsic permeability on porosity has been based on: k k0 exp[ b M M 0 ]I (11)
d = 1.60 g/cm3
-4
-3
d = 1.50 g/cm3
where is the macro porosity, k0 is the intrinsic permeability at a reference porosity (0) and b is a model parameter. 5 MODELLING
-2
d = 1.40 g/cm3
-1
0 0.01 0.1 1
The constitutive models presented in Section 4 have been implemented in the finite element program CODE_BRIGHT (Olivella et al., 1994) and have been used to model the three tests presented
Time (hours)
10
100
1000
Figure 5. Evolution of swelling pressures tests at three different densities. Symbols represent experimental data and lines model results.
363
potentials associated with each pore level are initially in equilibrium. An initial suction of 200 MPa has been adopted for both media. The swelling pressure tests presented in Figure 2 show more than one swelling step. Three stages can be detected: i) an initial stage, associated with the hydration and swelling of the more accessible active clay; ii) a transition stage between the first and the second swelling; and iii) a final stage, corresponding to the eventual hydration and swelling of the clays aggregates (i.e. grains) with active clay particles. The model can reproduce this behaviour by distinguishing each swelling phase according to the structural level involved. The first swelling is ascribed to the macrostructure; the second swelling stage is due to the delayed hydration of the expansive clay grains, considered in the model through the mass transfer term between the two media. Figure 5 shows the experimental results of three tests alongside with the model outputs. It can be seen that the model is able to reproduce in a satisfactory way the three observed stages, both in terms of kinetics of swelling and maximum swelling pressure values. 6 CONCLUSIONS
content, that is translated into a diminution of the initial swelling pressure. When the degree of saturation is higher, swelling affects all the material and predominates over the collapses, for what reason swelling pressure increases again and reaches a stable value (Imbert & Villar 2006). However, the analysis reported herein has shown that the transfer of water between the micro and the macrostructure of the clay and the different permeability of the two structural levels play also a key role in the hydration kinetics of the clay. The aim of this work was not to check the predictive capabilities of the formulation, but to explore its suitability to analyse the evolution of swelling clays submitted to wetting under confined conditions. A satisfactory response of the model in the cases analysed has been noted. It seems that the relevant physical phenomena, and couplings, are well captured by the double structure model for expansive clays. REFERENCES
Alonso, E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40(3): 405430. ENRESA (2006): FEBEX Full-scale Engineered Barriers Experiment, Updated Final Report 19942004. Publicacin Tcnica ENRESA 05-0/2006. Madrid. 590 pp. Gens, A. & Alonso, E.E. 1992. A framework for the behavior of unsaturated expansive clays. Can. Geotc. Jnl. 29: 10131032. Imbert, Ch. & Villar, M.V. 2006. Hydro-mechanical response of a bentonite pellets/powder mixture upon infiltration. Applied Clay Science 32: 197209. Olivella S, Gens A, Carrera J. Alonso E.E. 1994. Numerical formulation for a simulator (CODE-BRIGHT) for the coupled analysis of saline media. Engineering Computations; 13(7): 87112. Pusch, R. 1982. Mineral water-interaction and their influence on the physical behaviour of highly compacted Na bentonite. Can. Geotech. Jnl., 19: 381387. Snchez, M., Gens, A., Guimares, L. & Olivella, S. 2005. A double structure generalized plasticity model for expansive materials. Int. Jnl. Num. Anal. Meth. in Geom. 29: 751787. Villar, M.V. & Gmez-Espina, R. 2009. Report on thermo-hydro-mechanical laboratory tests performed by CIEMAT on FEBEX bentonite 20042008. Informes Tcnicos CIEMAT 1178. Madrid, 67 pp.
In this paper, the results of swelling pressure tests performed with FEBEX bentonite compacted at different dry densities have been presented. More than one stage of swelling development has been observed in practically all the samples. As expected, the final swelling pressure depends on dry density. The transition stage between the first and second swelling is more marked in samples of low dry density. A double structure model has been used to describe the behaviour observed in the tests. The problem exhibits distinctive features connected to the strong coupling between the unsaturated flow and the swelling and stiffness of the two media, characterized by an important swelling of the clay aggregates. The two-maxima path observed in the tests was also detected in infiltration tests performed in other clays and was explained as a consequence of the collapse of the macropores caused by the decrease in suction due to the increase in water
364
Constant volume triaxial compression tests on unsaturated soil prepared from slurry
M. Shimizu J. Terakata
Graduate School of Engineering, Tottori University, Tottori, Japan OYO Corporation, Sendai, Japan
ABSTRACT: Hvorslevs postulate is examined whether it can be a principle for the shear strength of unsaturated soil or not. The applicability of Bishops effective stress B and Fredlund et al.s idea to the shear strength is also focused on. A series of triaxial compression tests were conducted under the condition of constant volume on a silt sample. Specimens were prepared, from slurry, by the consolidation and unsaturation technique to raise the reproducibility of their initial states. Void ratio was regulated by changing the history of isotropic consolidation. As a conclusion, a unique failure envelope was obtained based on B in spite of the differences in void ratio and stress history, and therefore Hvorslevs parameters were found to be independent of them for the soil used. 1 INTRODUCTION constant volume in order to prevent the void ratio from changing during shear. 2 2.1 SAMPLE AND METHODS Sample
Hvorslev (1960) postulated that, if shear strength parameters in Mohr-Coulomb failure criterion are determined for an identical value of water content or void ratio for saturated cohesive soils, the parameters would not be dependent on the consolidation history but unique. Parameters thus determined were defined as the effective angle of friction and effective cohesion. His idea is approved if it is considered that even cohesive soils are particulate as sands for which the angle of shear resistance strongly depends on void ratio. One of objectives of this study is to examine whether Hovrslevs postulate can be applied for the shear strength of unsaturated soil or not. It seems still disputable that the effective stress concept applies to the shear strength of unsaturated soils, and Bishops effective stress (Bishop, 1959) can be the effective stress. Alternatively, instead of seeking the effective stress, it is intended that the shear strength of unsaturated soils is described as a function of stress parameters such as the net normal stress and suction, as Fredlund et al (1978) proposed. The present study focuses on the applicability of Bishop's effective stress and Fredlund et al.'s idea to the shear strength. A series of triaxial compression tests were conducted on a silt sample that contains a clay fraction. Specimens were prepared, from slurry, by the consolidation and unsaturation technique to raise the reproducibility of their initial states. Void ratio at the beginning of shear was regulated by changing the history of isotropic consolidation. The shearing was made under the condition of
Clayey silt available on the market was used (see Fig. 1 and Table 1). The sample contains clay fraction but it does not appear to be cohesive; in fact, the consistency limits, especially the liquid limit, could be determined with difficulty. In determining it by Casagrande method, the groove could not be cut successfully in the bowl with the amount of sample regulated in the standard method; therefore, the amount was reduced. 2.2 Apparatus Figure 2 shows the apparatus used for preparing specimens as well as for conducting triaxial tests. It was used also in a previous study (Shimizu et al., 2006). A ceramic plate (denoted by g in the figure) is installed on the pedestal (h); a tube is connected to the loading cap (e) to apply the pore-air pressure through it. A part, designated as inner cell (c), is fixed inside the outer cell (d) to facilitate the measurement of volume change. The apparatus has a problem. The inside volume of the outer cell can change when the cell pressure 3 changes; the change will make the water level in the inner cell vary; and therefore the measurement of specimens volume will be inaccurate. To fix the problem, 3 was held constant throughout every triaxial test; the net confining stress 3net ( 3 ua) was controlled by changing ua.
365
100 Percent finer (%) 80 60 40 20 0 0.001
0.01
0.1
Diameter (mm)
Figure 1. Table 1.
Particle size distribution curve. Physical properties of the sample used. s wL wP Silt Clay Mg/m3 (%) (%) (%) (%) 2.689 32.8 17.0 92.4 7.6
Soil particle density Consistency limits Particle size
through a preliminary test showing that the slurry can be homogeneous with the water content of nearly 60%. It was, next, consolidated in a cylindrical mold that was temporarily installed to mount the pedestal in the triaxial cell. Vertical pressure was applied by increments up to 49 kPa. The mold has the inner diameter of 3.5 cm, and is so high that the slurry is consolidated to be as tall as 8 cm. After the consolidation, the vertical pressure was reduced to 6 kPa and, at the same time, the suction of 50 kPa was applied. After the completion of the drainage due to the suction, the mold was removed leaving the sample on the pedestal. The application of the suction made specimens unsaturated and also made them keep the shape without the mold. These procedures yielded good reproducibility of specimens except for one specimen, as shown in Figure 3. 2.4 Method for triaxial tests
2.4.1 Initial state Specimens prepared by the method described above are lightly unsaturated. In spite of its state of saturation during the triaxial test, every specimen was saturated again in the triaxial cell, with latex membrane, before conducting triaxial test so that physical conditions would be common to all the specimens. For the saturation, water pressure of 10 kPa was applied from the specimens bottom, without applying the cell pressure so that pore air was flushed out through the tube connected to the loading cap. Once no air bubble was found in the tube, it was considered that full saturation was achieved. After the saturation, confining pressure of 3 kPa was applied to drain extra water possibly existing between the specimen and the membrane. The drainage was allowed only from the cap.
a: load cell, b: displacement transducer, c: inner cell, d: outer cell, e: loading cap, f: specimen, g: ceramic plate, h: pedestal, i: pressure transducer, j: circle-seal valve, k: pressure difference transducer, l: double tube burette, ua: pore-air pressure, uc: cell pressure (=3), ub: back pressure.
Figure 2.
Apparatus for triaxial tests.
w (%)
The drainage of pore air was always allowed. The drainage of pore water was controlled by the valve (j). The water pressure to be measured with the pressure transducer (i) is regarded as the porewater pressure, which is denoted by uw. 2.3 Method for preparing specimens Specimens for triaxial tests were prepared by one-dimensionally consolidating the slurry of the sample in the triaxial cell. The procedures are as follows. The sample was, at first, thoroughly remolded with the water content of 60%, which was determined
2.4.2 Unsaturation and consolidation The procedures for unsaturation and consolidation were followed as shown in Figure 4. Every specimen was, at first, isotropically consolidated by the effective confining stress of '3(1) in the state of full saturation. For specimens to be sheared in the unsaturated state, suction was applied with the net confining pressure, 3net(1), identical to '3(1); the
35 33 31 29 0.98
Sr =90% 88 86 84 82 80
x: average C-1 test specimen 1.00 1.02 1.04 1.06 1.08
Figure 3.
Initial conditions of specimens.
366
Suction 70kPa
Saturated Unsaturated
'3, 3net '3(1) =3net(1) 3net(3) 3net(2)
Figure 4. Procedures for consolidation and unsaturation prior to axial compression. Table 2. No. Conditions given to specimens. 3net(2) 3net(3) OCR ec Sr
2.4.3 Axial compression During the axial compression, the drainage of pore water through the specimens bottom was not allowed but the drainage of pore air was allowed through its top surface; the rate of compression was 600 minutes per 1 cm, which corresponds to 0.021%/ minute for the specimen of 80 mm in height. The volume was held constant by manually regulating the pore-air pressure ua. When the specimen tended to contract, ua was increased. Contrarily, when it tended to expand, ua was decreased. The increase in ua causes the decrease in the net confining pressure 3net; and vice versa. It will be evidenced that the constant volume condition was satisfactorily implemented. 3 RESULTS AND DISCUSSION Behavior in unsaturation and consolidation
3net(1)
(a) Tests on unsaturated specimens. A1 25 25 25 B1 25 75 75 C1 25 100 100 C2 25 150 75 C6 25 150 25 D1 150 150 150 D6 150 150 25 (b) Tests on saturated specimens. No. C1S C6S D1S 3(2) 85 90 250 3(3) 85 15 250 1 1 1 2 6 1 6 0.954 0.944 0.925 0.927 0.928 0.796 0.794 64 64 63 63 61 70 67
3.1
OCR 1 6 1
ec 0.926 0.929 0.798
suction applied is 70 kPa without exception. Then, the net confining stress was increased to a specified value, denoted by 3net(2), and, for some specimens, it was decreased to the value of 3net(3). In these processes of consolidation and unsaturation, and also in those that follows, cell pressure 3 was held constant at 340 kPa by the reason mentioned in section 2.2. Net confining stress 3net ( 3 ua) and effective confining stress '3 ( 3 uw) were changed by regulating pore-air pressure ua and pore-water pressure uw, respectively. Table 2 summarizes the conditions given to specimens, in which OCR is defined as the ratio of 3net(2) to 3net(3) for unsaturated specimens and '3(2) to '3(3) for saturated specimens. Void ratio was changed by varying 3net(2), and specimens with similar values of void ratio but with different values of 3net were prepared by varying the value of 3net(3). In this study, specimens are classified into 4 groups from A to D, according to the level of the void ratio at the end of the processes for consolidation and unsaturation before shear. The number following the group name, A to D, stands for the value of OCR, and S for saturated specimens.
The behavior of the drainage of pore water and the volume change in the processes for unsaturation and consolidation is shown in Figures 5 and 6, respectively. The results of C6 test are exemplified. For this test, suction of 20, 50 and 70 (kPa) was applied incrementally when 3net = 25 kPa in the unsaturation process, and 3net was increased from 25 to 150 (kPa), and decreased to 25 kPa in the process of consolidation. Figure 5 shows that the drained volume, Vw, is more than the volume change, V, at any level of suction, which indicates that the unsaturation developed. Figure 6 shows that the amounts of Vw and V are comparable each other for each stage of consolidation. This indicates that the change in the degree of saturation is trivial for the applied levels of 3net. In these figures, we observe that the volume change V fluctuates with time although the fluctuation is exaggerated because the logarithmic scale is used for time. The fluctuation reflects the variation of room temperature. In fact, the temperature varied in the range of 20 1 (C) and V correspondingly fluctuated (see Fig. 7).
0.1 -2 0 Vw (cm )
3
t (min) 10 100
1000 10000 -0.2 0.0 0.2 0.4 V (cm )

3
2 4 6 8 Vw V s=20 kPa 50 70
0.6 0.8 1.0 1.2
10 12
Figure 5. Behavior of Vw and V when suction was increased.
367
0.1 -0.2 0.0 0.2 Vw (cm3) 0.4 0.6 0.8 1.0 1.2
t (min) 10 100
1.05
1000 10000 -0.2
Unsaturated at 3net=25 kPa C6 C1 C1S C2 C6S
0.2 0.4 Vw V 3net=50 kPa 100 150 0.6 0.8 1.0 1.2 V (cm3)
Void ratio
0.0
1.00
0.95
Final states for consolidation.
0.90
10
100 3net, ' (kPa)
1000
Figure 6. Behavior of Vw and V when 3net was increased.

1000 0.0 0.2 V (cm ) 0.4 0.6 0.8 1.0 V
3
Figure 8. Relationships between void ratio and 3net or '3.

150 ua-ua0 (kPa)
1500
t (min) 2000
2500
3000 22 Temperature (C) 21 20 19 18 17
100 50 0 -50 0 5 10 a (%) 15
C6 C2 C1
Temperature
-100
Figure 7. Effects of the variation of room temperature to the volume change measurement.
Figure 9. The variation of the regulated pore-air pressure, where ua0 is the value of ua at the start of axial compression.
1.0 0.5
3.2
Void ratio vs. net confining stress

V (%)
C6
Void ratio was regulated in this study so that some specimens possess an identical value of it. In Figure 8, void ratio is plotted against the net confining stress 3net for tests in the group C. In these cases void ratio is regulated well to be in the range from 0.925 to 0.928. 3.3 Implementation of constant volume condition
0.0
-0.5 -1.0 0 5 10 a (%) 15
Figure 10. The variation of the volumetric strain during the constant-volume shearing. (Results from test C6).
Pore-air pressure ua was regulated so that the volume of specimen is held constant. Figure 9 exemplifies how the pore-air pressure ua was regulated, in which results from C1, C2 and C6 tests are shown. In Figure 10, the variation of the measured volumetric strain in test C6 is shown as an example showing that the condition of constant volume was achieved. In fact, the maximum change of the volumetric strain was less than 0.1% in all the tests. The amplitude of the variation, 0.1%, is partly due to the variation of the room temperature as was noted in Figure 7, and partly due to inaccurate regulation of ua. The pore-water pressure at the bottom of specimen, uw, changed due to the change in the pore-air pressure, ua. Changes in ua and uw resulted in the change in suction (Fig. 11). The suction continued to decrease although ua was, at first, raised and, subsequently, lowered as already shown in Figure 9.
3.4
Behavior of shear stress and strain
The relationships between shear stress q and axial strain a are shown in Figure 12, where q is defined as q
1 3 2
(1)
For comparison, the result from test C1S, conducted on a saturated specimen, is also shown. It should be noted that the axial strain is regarded as a measure for the shear strain because the volumetric strain was kept not to develop. It seems difficult to find, in this figure, certain effects for the unsaturation or over-consolidation to cause in the shear stress and strain behavior. In Figures 13 and 14, stress ratios q/pnet and q/pB are plotted, instead of q, against the axial strain, a,
368
10 s-s 0 (kPa) 0 C6 C2 C1
with
-10 -20 -30 0 5 10 a (%) 15
net ua and B net Sr s
(3)
20
Figure 11. The change in the suction due to the change in the pore-air pressure.
250 200 q (kPa) 150 100 50 0 0 5 10 a (%) 15 C1 C2 C6 C1S C6S 20
In Figure 13, the effects of suction and OCR are clearly seen. On the other hand, in Figure 14, such effects are scarcely found; this implies that the parameter pB, or B can be used to unify the shear stress and strain behavior in spite of the differences of the stress history and the degree of saturation. 3.5 Paths of stress parameters
Paths that stress parameters changed along are shown in Figures 15 and 16. In Figure 15, paths of pnet clearly depend on the degree of saturation and also on the degree of over-consolidation. On the other hand, in Figure 16, almost no difference in the paths of pB is found as the failure is approached. These observations are similar to those made when referred to Figures 13 and 14. 3.6 Strength characteristics Characteristics of shear strength are discussed by referring to the results from all the tests cited in Tables 1 and 2. In the following, shear strength is defined as the maximum value of the shear stress parameter q. Shear strengths are plotted against pnet in Figure 17, and against pB in Figure 18, respectively. In Figure 17, results of tests on unsaturated specimens are clearly separated from results of saturated ones, and they locate along a line parallel, but close to the straight line representing the state for saturated specimens. The reason for this seems to be that the suction at failure for unsaturated specimens is nearly the same, because the change in suction during shear was not quite different between specimens as was exemplified in Figure 11. The last finding can be explained by the equation, proposed by Fredlund et al., that qf pnetf sin s cos tan b (4)
Figure 12. Relationships between the shear stress q and the axial strain a.
1 0.8 q/p net 0.6 0.4 0.2 0 0 2 4 a (%) 6 8 C6 C2 C1 C1S 10
Figure 13.
1 0.8 q/p B 0.6 0.4 0.2 0
Relationships between q/pnet and a.
C6 C2 C1 C1S 0 2 4 a (%) 6 8 10
Figure 14.
Relationships between q/pB and a.
As for the relationships between qf and pBf that Figure 18 shows, almost all the plots locate along the straight line representing shear strengths of saturated specimens. This means that shear strengths are uniquely explained as a function of pB as qf pBf sin c cos (c 0 ) (5)
respectively, where the parameters pnet and pB are defined as follows: pnet
1net 3net 2
and
pB
1B 3B 2
(2)
Here, we note that Figures 17 and 18 include results from specimens with different values of void ratio and different consolidation histories. Considering this, we can say that the differences in
369
250 200 q (kPa) 150 100 50 0 0
C1 C6 C6S
C2 C1S
CONCLUSIONS
100
200 pnet (kPa)
300
400
Figure 15.
250 200 q (kPa) 150 100 50 0
Paths of pnet during shear.

C1 C6 C6S C2 C1S
100
200 p B (kPa)
300
400
Figure 16.
400 300 q f (kPa) 200 100 0
Paths of pB during shear.

A-1 C2 D1 D1S
C1S Line passing through C1S.
B1 C1 C6S
C6 D6 C1S
Specimens were prepared by the consolidation and unsaturation of slurry. The procedure yielded good reproducibility of their states before triaxial tests. The axial compression in triaxial tests was performed under the condition of constant volume. Void ratio was made identical between specimens by regulating the history of consolidation and unsaturation before the axial compression. The parameter pB, a function of Bishops effective stress, had the possibility to unify the behavior in the portion approaching the failure, in spite of the differences of the stress history and the degree of saturation. Strength characteristics shown in the qf vs. pnetf diagram can be explained by the equation proposed by Fredlund et al. In the qf vs. pBf diagram, almost all the plots located along a straight line representing shear strengths of saturated specimens. This means that shear strengths are uniquely explained as a function of pB. Differences in void ratio and in the consolidation history did not result in different values of the angle of shear strength and the cohesion intercept in both diagrams. Therefore, Hvorslevs postulate is not a necessary principle to explain the shear strength of unsaturated specimens. However, it should be added that different conclusions might be obtained for more cohesive samples than that used in this study. ACKNOWLEDGEMENTS
200 p netf (kPa)
400
600
Figure 17.
400 300 q f (kPa) 200 100 0
Relationships between q and pnet at failure.
Experiments for the paper were done by the second author in Tottori University. Authors sincerely thank Mr. K. Iwanari for his help in doing experiments. REFERENCES
Bishop, A.W. 1959. The principle of the effective stress. Teknisk Ukeblad. 106: 859863. Fredlund, D.G., Morgenstern, N.R., & Widger, R.A. 1978. The shear strength of unsaturated soils. Can. Geotech. J. 15: 313321. Hvorslev, M.J. 1960. Physical components of the shear strength of saturated clays. Proc. Research Conf. on Shear Strength of Cohesive Soils, ASCE, 169273. Shimizu, M., Sakamoto, S. &. Nishioka, T. 2006. Effects of drainage conditions on the shear strength of unsaturated soil, Proc. 4th Int. Conf. on Unsaturated Soils, ASCE, 12231234.
A1 C2 D1 D1S
B1 C1 C6S
C6 D6 C1S
C1S Line passing through CIS.
200 p Bf (kPa)
400
600
Figure 18.
Relationships between q and pB at failure.
void ratio and history did not result in different values of the angle of shear strength and the cohesion intercept. And therefore, Hvorslevs postulate is not required to explain the shear strengths of unsaturated specimens.
370
Assessment of the cv coefficient for unsaturated Warsaw Clay

Z. Skutnik, M. Lendo-Siwicka & K. Garbulewski
Department of Geotechnical Engineering, Warsaw University of Life Sciences, Poland
ABSTRACT: In order to determine the effect of saturation degree on Warsaw clay stiffness a series of laboratory consolidation tests were conducted. The tests were carried out on undisturbed clay samples for different net stresses and degree of saturation values using the suction controlled consolidometer (called UPC- consolidometer). The UPC apparatus allows to performing consolidation tests on unsaturated soil samples controlling suction through the application of the axis-translation technique. The Casagrandes method was applied for determination of the coefficient of consolidation cv. depending on the clay saturation degree. It was concluded that the value of the cv estimated for different values of suctiondegree of saturation, decreases while loading increases. However, for the same level of loading the cv values decrease if applied suction increases (degree of saturation decreases). 1 INTRODUCTION where cv ks w gmv (2)
Consolidation is defined as the reduction of the volume of a soil due to the expulsion of water (Terzaghi, 1943). This will be accompanied by the dissipation of pore water pressure. To determine the compressibility characteristics of the soil a laboratory consolidation test on an undisturbed soil sample is usually performed. A conventional consolidation test is conducted over a number of load increments. The number of load increments should cover the stresses range from the initial stress state (in situ) of the soil to the final stress state due to the proposed construction. Based on the test results the compressibility characteristics of the soil are: parameters to estimate the amount of consolidation settlement and parameters to estimate the rate of consolidation settlement in the field. While the first ones do not depend directly on the saturation of the soil, the second group of consolidation parameters was depend on soil water characteristics. The rate of consolidation settlement is estimated using the coefficient of consolidation cv. This parameter is determined for each load increment in the test. The value of cv for particular pressure increment in the oedometer test can be determined by comparing the characteristics of the experimental and theoretical condition curves, the procedure being referred to as curve fitting. For saturated soil consolidation process can be described by Terzaghi, (1943) formula (1): 2uw uw cv 2 t y (1)
cv,coefficient of consolidation, kscoefficient of permeability with respect to water for fully saturated soil, wdensity of water, ggravitational acceleration, mvcoefficient of volume change for saturated soils. For an unsaturated soils Fredlund & Rahardjo 1993, proposed the following equation (3):
2 uw w uw cv t y 2
(3)
where
w cv
kw w w gm2
(4)
w cv coefficient of consolidation with respect to the water phase, kwcoefficient of permeability with respect to water as a function of matric suction, mw 2 coefficient of water volume change with respect to a change in matric suction. The paper presents the experimental results of consolidation tests carried out on unsaturated Warsaw clay samples. The pore-air and pore water pressures on the boundaries of the soil specimens during consolidation were controlled. The total volume change of the soil specimen as well as the water volume change was measured during the consolidation test. The tests were performed on several Warsaw clay specimens using a
371
consolidometer designed and constructed at the Geotechnical Laboratory of the U.P.C. (Technical University of Catalonia, Barcelona, Spain). The UPC consolidometer allows to perform compressibility tests under suction controlled conditions (Romero, 1999). 2 PROPERTIES OF WARSAW CLAY
The Warsaw area is located within a tectonic unit composed of Upper Cretaceous deposits, the Mazovian Basin filled with Tertiary marls and marly high-plasticity clays. The thickness of Pliocene deposits is variable and depends on the intensity of later glacitectonical processes. Termogravimetric analyses using LabsysTM TG-DTA12 showed that the clay fraction primarily consists of beidelite, as well as illite and kaolinite (Table 1), minerals known for being moderately to high expansive. The general properties of Warsaw clays are listed in Table 1. The scanning microscope photograph of clay is presented in Fig. 1. The soil samples for the consolidation tests have been taken from the experimental plot in Stegny (district of Warsaw). The index properties of tested soil samples of Warsaw clays are listed in Table 2. 3 CONSOLIDATION TESTS
inside a cell where controlled air pressure can be applied. The ceramic disk allows to control the negative water pressure in the soil specimen. Vertical load is applied with compressed air. The loading chamber from the pore air pressure is separated with 1 mm thick rubber membrane. Water volume change is measured by means of a graduate glass burette enclosed in cylinder. Axial displacements are measured by means of mechanical micrometer with resolution 2 m. The research program on unsaturated Warsaw clay included the determination of the consolidation
Consolidation tests were carried out using UPC consolidometer (Fig. 2), which allows controlling both matrix and total suctions during whole test (Alonso et al. 1990, Gens & Alonso 1992). Matrix suction is controlled through the application of the axis translation technique. The specimen rests on a high air entry value ceramic disc (1.5 MPa)
Table 1. General properties of Warsaw Clay. Unit % g/m3 g/m3 % % % % % % % % % Value 35.090.0 2.672.75 1.952.10 33.048.0 43.085.0 22.038.0 23.038.0 0.751.25 44.162.8 04.6 9.411.2 0.75.2 0.00.3
Figure 1. Scanning microphotograph of clay.
Table 2. Sample no. 1 2
Properties of tested soil samples. s d e (g/m3) (g/m3) (g/m3) 2.69 2.71 1.95 2.05 1.48 1.61 0.82 0.69 wn (%) 31.7 27.5 wP (%) 35.0 22.5 wL (%) 83.0 69.5
Parameter Content of clay Density of solids [s] Bulk density [] Porosity[n] Liquid limit[wL] Plastic limit[wP] Plasticity index[IP] Activity[A] Mineral composition Beidelite Illite Kaolinite Geotyte Syderyte
Figure 2.
The testing equipment for unsaturated soils.
372
curves for different load increments and for different soil suction. The consolidation tests were performed for two levels suctions: s ua uw 400 kPa for soil sample No. 1, and s 50 kPa for soil sample No. 2. The values of net vertical stresses (v ua) for soil sample No. 1 were equal: 250 kPa, 300 kPa, 350 kPa and 400 kPa. The values of pore-air pressure (ua) for each load increment were the same and were equal to 410 kPa. The pore water pressure (uw) was equal to 10 kPa. For the soil sample No. 2 for which the suction, s 50 kPa, the net vertical stresses (v ua) were equal 350 kPa, 400 kPa, 450 kPa, 500 kPa and 550 kPa, and ua 60 kPa, and uw 10 kPa. The loading paths are shown in Figure 3. The consolidation test results for different net stresses (250, 300, 350, 400 kPa) for suction equal to 400 kPa are presented in Fig. 4. The Casagrandes method was applied for cv determination (Head, 1982). In the log time method the experimental curve is obtained by plotting the dial gauge reading in the oedometer test against the logarithm of time. The coefficient of consolidation has been calculated from the formula (5): cv 0.196 h2 t50 (5)
4,80
4,78
4,76
4,74
Gauge reading [mm]
4,72
4,70
4,68
4,66 250 4,64 300 350 400 4,62

1,E+01 1,E+02 1,E+03 1,E+04 1,E+05 1,E+06
where hheight of sample, t50time at U 50% determined from the construction which example is given in Fig. 5. In Fig. 6 the consolidation test results for net stress (350, 400, 450, 500, 550 kPa) and suction 50 kPa are presented. In Table 3 the values of cv obtained from the tests with different values of suction and net stress are presented.
Time [s]
Figure 4. Consolidation curves for different net stress (v ua) values and suction s 400 kPa, soil sample No. 1.
4,70 Gauge reading [mm] 4,70 4,69 4,69 4,68
t50 =20000
t1
c b a ab=bc
4t
v - ua = 500 kPa 0%
500
50 %
400 Suction, ua - uw [kPa]
300
100
Sample No. 1 s = 400 kPa Sample No. 2 s = 50 kPa
4,68 10 100 1000 10000 Time [s] 100000 1000000
200
100
0 0 200 400 600 Vertical net stress, v - ua [kPa]
Figure 5. Casagrande procedure of t50 time determination for loading v ua 500 kPa and suction s 50 kPa.
CONCLUSIONS
Figure 3.
Loading paths during consolidation tests.
For the purposes of the research it was assumed that both processes the dissipation of excess
373
u=100 %
4,75
Table 3. Calculated values of coefficient of consolidation for different loading and suction. Sample no. 1 Sr* ua uw v v ua (kPa) (kPa)(kPa) t50 (s) cv (cm2/s) 660 710 760 810 410 460 510 560 610 250 300 350 400 350 400 450 500 550 2000 5800 12000 70000 3000 7000 7000 20000 30000 0.000394 0.000136 0.000066 0.000011 0.000263 0.000113 0.000113 0.000039 0.000026
4,74
4,73
0.799 0.886 400
4,72
2
Gauge reading [mm] 4,71
0.798 0.881 50
4,70
Sr*value of initial water saturation of specimen.
4,69
4,68
4,67
4,66
4,65 1,E+01 1,E+02 1,E+03 1,E+04 1,E+05 1,E+06 Time [s]
Figure 6. Consolidation curves for different net stress (v ua) values and suction s 50 kPa, soil sample No. 2.
5,0E-04 Coefficient of consolidation cv [cm 2/s] 4,0E-04 3,0E-04 2,0E-04 1,0E-04 0,0E+00 0 100 200 300 400 500 600 Sample No. 1 s = 400 kPa Sample No. 2 s = 50 kPa
pore-air and pore-water pressures are going simultaneously. Therefore, interpretation of unsaturated soil behavior and determination of cv values depending on the clay saturation degree were conducted with assumptions (Casagrandes procedure) for saturated soils. Based on the test results obtained it can be concluded that the values of coefficient of consolidation cv decrease if the loading (net stress) increases. For the same level of loading the cv values are higher if the suction level is lower (degree of saturation is higher). It should be underlined that lowest applied net stress in all consolidation tests was higher than swelling pressure which is 150 200 kPa. REFERENCES
Alonso, E.E., Gens, A., Josa, A. 1990. A constitutive model for partially saturated soils. Geotechnicue 40, No. 3, 405429. Fredlund, D.G., Rahardjo, H., 1993. Soil mechanics for unsaturated soils, John Wiley & Sons, Inc. Gens, A., Alonso, E.E., 1992. A framework for the behavior of unsaturated expansive clays. Canadian Geotechnical Journal 29(6):10131032. Head, K.H. 1982. Manual of Soil Laboratory Testing Vol. 3. Romero, E. 1999. Characterization and thermohydro-mechanical behaviour of unsaturated Boom clay: an experimental study. Doctoral Thesis, Universitat Politecnica de Catalunya, Barcelona. Terzaghi. K. 1943 Theoretical Soil Mechanics. New York: Wiley.
Vertical net stress ( v ua), [kPa]

Figure 7. Changes of the coefficient of consolidation cv as a function of net stress for different values of suction.
374
Laboratory investigations on static liquefaction potential of pyroclastic soils involved in rainfall-induced landslides of the flow-type
G. Sorbino, G. Migliaro & V. Foresta
University of Salerno, Department of Civil Engineering, Fisciano, Salerno, Italy
ABSTRACT: Rainfall-induced landslides of the flow-type often cause huge social and economical disasters, posing threat to life and livelihood all over the world. In Campania Region (southern Italy) these landslides occur in loose unsaturated pyroclastic covers originated from eruptive activities of different volcanic apparata. For these landslides, a clear understanding of the mechanisms responsible for the transition from a frictional soil-type to a viscous fluid-type is one of the basic requirements to perform an adequate landslide hazard analysis. This transition is likely from soil liquefaction due to landslide occurrence. The present work investigates the potential liquefaction of the pyroclastic soils through a new unsaturated direct simple shear apparatus. To this aim, stress-controlled and strain-controlled tests were carried out on initially unsaturated pyroclastic soil specimens characterized by a metastable structure. The results evidence the influence of the volumetric collapse on the liquefaction potential due to both wetting process and soil destructuration. 1 INTRODUCTION (e.g. Sladen et al. 1985, Lade 1993, Chu et al. 2003, Olivares & Damiano 2007). However, it should be noted that in situ stress and strain paths experienced by a slope could not be precisely simulated by triaxial tests, either when the minor principal stress 3 remains constant while major principal stress 1 increases, or when both 1 and 3 remain constant while the pore-water pressure is increased from an initial negative value until failure occurs. Alternatively, stress and strain paths more adherent to the slope in-situ conditions can be better reproduced by simple shear tests. In the present paper, after a brief description of the Unsaturated Direct Simple Shear Apparatus designed at the University of Salerno, the results of some tests on initially unsaturated pyroclastic soil specimens are presented. The obtained results put in evidence the role of the wetting and of soil structure changes on the volumetric behaviour of the analyzed soil specimens. They also provide indications on the possible occurrence of static liquefaction during the triggering stage of these landslides. 2 THE SIMPLE SHEAR APPARATUS
Rainfall-induced landslides of the flow-type affect a large portion of the Campania Region (southern Italy) and they can certainly be considered one of the most insidious landslide phenomena because of their potential for causing casualties and huge economic damages. They involve unsaturated pyroclastic soil coversoriginating from the explosive phases of different volcanic apparatawhich mantle the limestone and tuffaceous slopes. These landslides can be considered complex slope instability phenomena since they exhibit distinct kinematic characteristics during the failure, post-failure and propagation stages (Fell et al. 2000, Leroueil 2004, Cascini et al. 2009). Failure stage is characterised by the formation of a continuous shear surface through the entire soil mass (Leroueil 2001). Post-failure stage is represented by the rapid generation of large plastic strains and the consequent sudden acceleration of the failed soil mass (Hungr 2004). In the literature, the acceleration of the failed mass during the post-failure stage is mainly attributed to static liquefaction, a process in which a soil mass, sheared under undrained condition, loses a large part of its shear strength as a result of induced high pore-water pressure, and consequently flows like a fluid. Most previous investigations on static liquefaction were based largely on triaxial tests on saturated and unsaturated soils
The equipment used in the present work is the Unsaturated Direct Simple Shear Apparatus (UDSSA) designed at the University of Salerno (Sorbino & Foresta, in prep.). This apparatus is
375
able to apply a simple shear mode of strain (i.e. zero horizontal strain in the direction of shear together with plain strain in the orthogonal direction) to cylindrical soil specimens. The layout of the apparatus is shown in Figure 1. The soil suction is either measured or applied by means of the axis translation technique. The pore-water pressure can be controlled both at the head and at the toe of the specimen. The pore-air pressure in the specimen is equal to the cell-air pressure in the pressurized chamber containing the simple shear box. The shear box (Fig. 2) has an inner circular cross section and can accommodate a soil specimen having 51 mm in diameter. Specimens have a height varying between 12.5 mm and 22 mm. The zero lateral strain conditions in the specimen during shear are attained by means of a stack of twelve rings. The lowest ring has the same height (10 mm) of the base pedestal. The pedestal is connected with the horizontal ram and can slide on a pair of submergible linear bearings. The upper ring is also 10 mm thick and it is assembled with an arm (goose-neck) that constrains the upper part of the box to the horizontal load gauge. The other rings have height of 1.9 mm and can slide over each other obeying to a synchronized movement. This movement is allowed by two linear fluting (grooves)machined on the upper face of each diskand two external holed metal straps fixed on the lower and upper ring. Shear load can be applied both in stresscontrolled or strain-controlled modes. As for the stress-controlled mode, an electro-pneumatic regulator provides the air pressure control through an air-water interface connected to the horizontal ram. In strain-controlled mode, the horizontal ram displacements are regulated by means of a Bishop pump, activated by a stepper motor controlled by an A/D card. The apparatus is normally used in conjunction with a computer to run different types of shear tests (Sorbino & Foresta, in prep.). These tests involve the measurements of the applied vertical
9 4 1 2 3 1 2 3 4 5 6 7 8 9 Pressure chamber Internal LVDT Vertical Ram Vertical load cell Horizontal ram Horizontal load cell Shear box Linear bearing Pneumatic press
Top cap with Double drainage system (air/water) Stainless steel rings
Goose-neck connection HAEV porous stone Join to Horizontal load ram Submergible linear bearing 0 5 10 cm
Figure 2.
Shear box details.
load, the resulting volume change and shear stresses as the sample is shared, as well as the control of the applied pore pressures (air and water) and the resulting change in the water content. 3 3.1 MATERIALS AND METHODS Tested materials
Laboratory tests were performed on a non-plastic pyroclastic soil samples whose grain size distributions and main physical properties are summarized in Table 1. These samples, named SL and Q, were obtained from two slope sites of the Pizzo dAlvano massif (Campania Region-southern Italy). As already pointed out by Bilotta et al. (2005) these samples are representative of the pyroclastic covers susceptible to rainfall-induced landslides of the flow-type. They are characterized by a very loose state (high void index values) and by a metastable structure. SL soil has an average void index e0 2.44 and specific gravity Gs 2.55, while the Q soil has e0 2.87 and Gs 2.65. 3.2 Simple shear tests
6
LOAD CELL
LOAD CELL
5 9 0 5 10 cm
0510
Figure 1. Cross-section of the UDSSA apparatus with indication of the main mechanical components.
To asses the liquefaction potential of the studied pyroclastic soils, some simple shear tests were performed. All the tested specimens were firstly equalized to an imposed value of suction. The suction applied was 0 kPa for tests A1, C1 and 50 kPa for tests A2, B and C2 (Table 2). A maximum time of 48 hours was sufficient to obtain pore water pressure equalization within the specimen. After equalization the net vertical stress was applied and about 24 hours were needed to complete consolidation phase. Finally, the shear phase was performed
376
Table 1. Grain size characteristics and some physical properties of tested soils. Grain size distribution Clay (%) Silt (%) 53.6 40.9 49.0 Sand (%) 40.6 51.3 44.6 Grav. (%) 1.1 6.4 1.8 2.87 2.65 e0 Gs
SL upper 4.7 bound limit SL lower 1.4 bound limit Q 4.6 Table 2.
2.44 2.55
Summary of the performed simple shear tests.
Test ua uw ua uw type initial final Specimen kPa kPa kPa v ua v0 S0609 S0709 S0809 S0909 S1009 S1309a S1309b S1309c S1409 A2 A2 B A1 A1 A1 C1 C1 C2 50 50 50 0 0 0 0 0 50 0 0 0 0 0 0 0 0 0 100 50 30 50 100 50 100 150 50 3.89 3.78 3.49 3.65 3.72 3.36 3.36 3.36 3.28
vc 3.68 3.68 3.43 3.45 3.54 3.24 3.02 2.94 3.20
soil Q Q SL SL Q SL SL SL SL
at constant suction (type A1, C1) or at variable suction (type A2, B, C2). Particularly: 1. Tests A1 and A2 are constant vertical load strain controlled shear type. A shear strain rate of 0.14 mm/h was imposed allowing the volume strains of the specimen to develop during shearing. For A2 tests, suction was reduced to 0 kPa by flooding the specimens, in correspondence of a fixed value of shear strain; 2. Tests B are constant vertical loadstress controlled shear type. In the first phase of these tests, a shear stress rate of 3.6 kPa/h was applied to the specimens until a fixed value of shear stress was attained. The adopted shear stress rate was chosen in order to ensure any variation of suction in the specimen. In the second phase, shear stress was maintained constant and the suction was reduced to 0 kPa by flooding the specimens; 3. Tests C1 and C2 are constant volume strain controlled shear type. A strain shear rate of 5 mm/h was applied to the specimen. Constant volume is attained by automatically regulating net vertical stress. For C2 tests a saturation step, obtained by flooding the specimen, was provided when the mobilized shear stress of the material attained a value lower than the saturated shear strength.
Figure 3, shows the results obtained by straincontrolled tests (A1 and A2 tests) carried out at constant vertical stress of 50 kPa and 100 kPa. From the Figure 3 it can be observed that the two specimens tested in saturated condition (A1 tests) show the typical stress-strain behaviour of contractive material. In the same figure are also reported the results of A2 tests obtained through a rapid decrease in suction. As it can be noted, the suction decrease produces a volumetric collapse and a drop of the shear stress. In both the tested specimens this drop results in shear stress values that are lower than those attained, at the same shear strain, by the saturated specimens. After this drop, both the specimens show an increase of the shear stress with the shear strain and approach the stress-strain curves of the saturated specimens. The attainment of shear stress values lower than those showed by the saturated specimens could be explained by an increase of the pore water pressure due to volumetric collapse during wetting. The increase in pore water pressure seems to be confirmed also by a stress-controlled shear test at constant vertical stress (B test). In this test, the adopted net vertical stress was equal to 30 kPa and the initial applied suction was 50 kPa. As for the shear stress, a value of 18 kPa was applied and maintained constant. This value is lower than the shear strength of the saturated soil at an effective vertical stress of 30 kPa. Figure 4 shows the stress-strain followed by the specimen together with both the vertical displacements and the shear strain time trends during the shear phase. As it could be expected, a volumetric collapse develops during the wetting. A sudden increase of the shear strain rate was also observed during this stage of the test. This behaviour highlights that the applied shear stress at collapse is
shear stress (kPa)
80 60 40 20 0.0 0 0 0.2 v -u a = 50 kPa wetting v -u a = 100 kPa
shear strain
0.4 0.6 008
vert. displ. (mm)
-0.5 -1.0 -1.5 -2.0 collapse
Figure 3. Stress-strain relationship showed in strain controlled shear tests: dashed line (A1 test) and continuous line (A2 test).
377
shear stress (kPa)
v coll = Hwet/Hs
25 20 15 10 5 0 0.4 0.3 0.2 0.1 0.0 0 0
(a)
wetting
10 0.00 0.02 0.04 0.06 0.08 0.10 0.12 2.1 2.0 1.9 1.8 1.7 1.6 10
100
v -ua (kPa) 1000

v 0 =3.05
(a)
v 0 =3.31
v -ua = 30 kPa
shear strain
0.1 0.2 0.3 0.4
v 0 mean =3.55
vert. displ. (mm) vv shear strain
(b)
specimens heigth H (cm)
(b)
Hwet due to collapse
time (min)
100 200 300 400
-0.4 -0.8 -1.2 collapse
ESA02_06 s = 50 kPa ESA01_05 s = 0 kPa
100
v -ua (kPa) 1000
Figure 4. Shear stress-strain behaviour (a) and time evolution of vertical displacements and shear strain (b) in stress controlled simple shear test.
Figure 5. Results of oedometer tests: a) specific volume variation due to wetting collapse vs. vertical net stress for specimens with initial specific volume between 3.64 to 3.45; b) specimen height variation with net vertical stress in saturated condition and during repeated wettingdrying paths (after Migliaro 2008).
higher than the available shear strength of the soil. This means that an increase in pore water pressure have to be developed in the specimen. The obtained results confirm that the volumetric collapse is responsible of pore water pressure increase during wetting paths. The observed behaviour is, however, a necessary but not sufficient condition for the development of liquefaction phenomena. The same condition becomes also a sufficient one if the amount of the volumetric collapse is so much to induce in the specimen a pore water pressure equal to the mean total stress. For the tested soil, Migliaro (2008) showed that the potential amount of the volumetric collapse can be related to two different conditions: one is represented by the wetting paths on unsaturated specimens; the other one by the modification of the soil structure. Figure 5a shows the variation of the specific volume due to volumetric collapse obtained in oedemeter tests (Migliaro, 2008) upon wetting for unsaturated soil specimens at different net vertical stresses and having an initial suction of 50 kPa. As it can be noted, the results highlight a clear dependence of the volumetric collapse from the net vertical stress and from initial void ratio. Similar values of volumetric collapse were also obtained from wetting tests performed in triaxial conditions (Migliaro 2008). It is interesting to note that the observed volumetric-deformation upon wetting does not imply for the tested soil a change in its structure. This is clearly evidenced by the results of oedometer tests shown in Figure 5b. Here, a series of
wetting and drying paths have been carried out on the same specimen (Migliaro 2008). As it can be noted, whatever is the vertical net stress, the specimen approaches, upon wetting, the compressibility curve pertaining to the saturated condition. This, in turn, excludes the presence of a structural change due to wetting paths. The influence of the structural change of the tested soil on the volumetric collapse has been investigated by performing a series of oedometer tests in saturated conditions on undisturbed and remoulded samples. Remoulded samples were obtained by adopting different techniques of mechanical mixing (Migliaro 2008). Figure 6 shows the effect of the mixing time on the reduction of the initial specific volume (v0) of the tested specimens. As already showed by Migliaro (2008) this specific volume reduction is due to the macro-aggregate destruction (destructuration) without any variation in the soils grain size distribution. Figure 7 shows the results obtained from the oedometer tests. By comparing Figure 5a and Figure 7, it can be noted that the amount of the volumetric collapse due to the destructuration (i.e. difference among the specific volumes belonging to the undisturbed and remoulded specimens at a given effective vertical stress) is larger than the one obtained from wetting paths. Macro-aggregates destructuration can be also reproduced by performing cyclic drained simple shear tests on undisturbed specimens. Typical results of this type of tests are illustrated in Figure 8. As can be noted, most of the volume
378
3.19 2.99
'c
v0 2.79
2.59 2.39 0 500 1000 1500 2000 2500
mixing time (min)
0.40 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 0 0.2
Saturated Failure envelope
'v = 100 kPa
'v = 150 kPa 0.4 0.6 0.8 1
'v 'c
1.2
Figure 6. Variation of the initial specific volume of Q and SL soils specimens for different mixing time (after Migliaro 2008).
3.75
I = undisturbed CL1 = remoulded S1309 S1409
Figure 9. Normalized stress paths during a constant volume shear for reconsolidated S1309 saturated specimen.
1 0.8 0.6
Inversion suction zeroed
specific volume
3.55 3.35 3.15 2.95 2.75 2.55 1
Suction from 50 to 0 kPa
CL1
/(v - ua)c
0.4 0.2 0 -0.2 -0.4 0 0.2 0.4 0.6 0.8 1 1.2
after 1st cicle 10 100 1000
'v (kPa)
10000
Figure 7. Comparison among the oedometer tests performed on undisturbed and remoulded specimens and the values of specific volumeafter reconsolidationof two specimens cyclic sheared in simple shear tests.
40 20 0 0 -20 -40 0 0.2 0.4 0.6 0.2 0.4 0.6 0.8 decreasing shear strain Shear strain 0.8 1 1 increasing shear strain
(v -ua)/(v ua)c
Saturated Failure envelope
Figure 10. Normalized stress-path for S1409 specimen during the constant volume shear test.
Shear stress (kPa)
-0.01 0 -0.02 -0.03 -0.04
Figure 8. Stress-strain behaviour for S1309 saturated specimen during cyclic drained simple shear test.
reduction of the specimen is attained when reversing the shear strain. The destructuration of the specimen during this cycle is confirmed by the oedometric reconsolidation of the specimen that (Fig. 7) travels from the undisturbed compressibility domain to the remoulded one. A confirmation of the occurred destructuration of the specimen is given by the results obtained from constant volume simple shear tests (C1 test) carried out at vertical stress of 100 kPa and 150 kPa. Figure 9 shows the results of these tests in terms of dimensionless stress paths (i.e. the values of shear stress and
effective vertical stress referred to effective vertical stress at the end of the consolidation phase c). As it can be noted, the specimen shows a marked dilative behaviour, typical of remoulded dense material. In order to evaluate the influence on the potential for liquefaction of the volumetric collapse due to both wetting paths and destructuration, a specific simple shear test was carried out (C2 test). During this test, the specimen was initially equalized at an imposed suction of 50 kPa and then consolidated under 50 kPa net vertical stress value. At the end of the consolidation phase the specimen was sheared to a constant rate of shear strain by imposing a constant volume condition. Figure 10 shows the dimensionless stress path (i.e. the shear stress and net vertical stresses referred to the value of the net vertical stress at the end of the consolidation phase (v ua)c) followed by the specimen during the shear phase. As it can be noted, during this stage of the test the specimen shows an initial tendency to dilate. This tendency can be revealed by the progressive increase of the applied net vertical stress (Fig. 10). In the subsequent stage of the test, suction was rapidly reduced from the initial value of 50 kPa to 0 kPa by flooding the specimen. During this stage the specimen also shows an initial decrease of the shear stress. As the shear strain increases,
vert. strain a
379
a continuous decrease of the net vertical stress at constant shear stress was experienced by the specimen, probably due to dissipation of excess pore water pressure generated by wetting. The decrease of effective vertical stress leads the specimen to approach the failure envelope in saturated condition under an effective vertical stress greater than zero. This means that the avoided contractive behaviour of the soil and the avoided collapse during wetting path, is not able to conduct the specimen to the liquefaction. In the final stage of the test an inversion in the shear strain was applied to the specimen in order to simulate the soil destructuration. As it can be noted (Fig. 10), the specimen immediately reduces its shear and vertical stresses that approach the zero stress conditions (soil liquefaction). This behaviour is due to a more consistent volume variation tendency that the specimen develops through destructuration. After the occurrence of the liquefaction phenomena, the specimen exhibits again a dilative behaviour. This last circumstance can be ascribed to the denser state achieved to the soil following the occurred destructuration. 5 CONCLUDING REMARKS
now obtained, it seems that the occurrence of liquefaction could be possible also for the same soil in medium/dense state. To support these hypotheses, further simple shear tests have to be carried out, taking into account different initial structures of the soil. ACKNOWLEDGEMENT The Authors would like to thank Professors E. Bilotta and L. Cascini for supporting and encouraging them during the research. REFERENCES
Bilotta, E., Cascini, L., Foresta, V. & Sorbino, G. 2005. Geotechnical characterisation of pyroclastic soils involved in huge flowslide. Geotechnical and Geological Engineering 23: 365402. Cascini, L., Cuomo, S., Pastor, M. & Sorbino, G. 2009. Modelling of rainfall-induced shallow landslides of the flow-type. ASCEs Journal of Geotechnical and Geoenvironmental Engineering (in press). Chu, J., Leroueil, S. & Leong, W.K. 2003. Unstable behaviour of sand and its implication for slope stability. Canadian Geotechnical Journal 40: 873885. Fell, R., Hungr, O., Leroueil, S. & Riemer W. 2000. Keynote LectureGeotechnical engineering of the stability of natural slopes, and cuts and fills in soil. GeoEng November 2000, Vol. 1: 21120. Technomic Publishing, Lancaster. Hungr, O. 2004. Flow slides and flows in granular soils. In Proc. of the Int. Workshop Flows 2003Occurrence and Mechanisms of Flows in Natural Slopes and Earthfill, Sorrento, Patron Ed. Lade, P.V. 1993. Initiation of static instability in the submarine Nerlerk berm. Can. Geotech. J. 30: 895904. Leroueil, S. 2001. Natural slopes and cuts: movement and failure mechanisms. Geotechnique 51(3): 197243. Leroueil, S. 2004. Geotechnics of slopes before failure. Landslides: Evaluation and Stabilization, Lacerda, Ehrlich, Fontoura & Sayao (eds), Vol. 1: 863884. Migliaro, G. 2008. Il legame costitutivo nei terreni piroclastici per la modellazione di scavi in ambiente urbanizzato ed influenza della parziale saturazione. Ph.D. Thesis (in Italian) University of Salerno. Olivares, L. & Damiano, E. 2007. Postfailure Mechanics of Landslides: Laboratory Investigation of Flowslides in Pyroclastic Soils. Journal of Geotechnical and Geoenvironmental Engineering ASCE 133(1): 5162. Sladen, J.A., DHollander, R.D. & Krahn, J. 1985. The liquefaction of sands, a collapse surface approach. Canadian Geotechnical Journal 22: 564578. Sorbino, G. & Foresta, V. (in prep.). A new Unsaturated Direct Simple Shear Apparatus.
Simple shear tests up to now performed on undisturbed pyroclastic soils highlight an unstable volumetric behaviour when subjected to wetting paths during shearing. This instability, which can be attributed to the collapsible tendency of the investigated soils, is showed both in constant vertical load and constant volume shear tests. However, the amount of the volumetric collapse, due to transition versus saturated conditions, seems not to be sufficient to induce liquefaction. More consistent volumetric strains, such as those produced by destructuration of the soil structure, are needed. Such additionally volumetric strainsthat can be attained in the soil specimens by reversing shear strain in simple shear testsseem to be responsible for liquefaction occurrence. In situ, a strain reversal can take place when, in addition to rainfall, further boundary conditions (coming for example from buried springs) act at the base of the loose pyroclastic covers. It is also evident from Figure 6 that soil skeleton structure can be progressively lost. A soil specimen not completely destructurated can thus experience consistent volume decrease. From the results up to
380
Desorption and consolidation behaviour of initially saturated clays

S. Tripathy, R. Bag & H.R. Thomas
Geoenvironmental Research Centre, Cardiff University, Cardiff, UK
ABSTRACT: One-dimensional consolidation and desorption behaviour of highly plastic clays (MX 80 bentonite and Speswhite kaolin) were experimentally determined in this study. Initially saturated clay specimens were subjected to an increase in suction employing axis-translation technique and salt-solution technique to cover a suction range of 10 kPa to 300 MPa. Consolidation tests were carried out with a maximum applied vertical pressure of 3.2 MPa. The water content and the volume of the clay specimens at each applied suctions were measured. The study revealed that a vertical pressure increase was more effective in reducing the water content and the void ratio for the bentonite studied, whereas a suction increase was more effective in removing water from the kaolinite. The possible structure and fabric of the clays at saturation, at commencement of desaturation, and at residual shrinkage stage are discussed. The observed differences in the volume change and water retention behaviour were attributed to the differences in the mineralogical properties of the clays. 1 INTRODUCTION vertical pressure, and isotropic loading conditions. Marcial et al. (2002) compared the volume change behaviour of bentonites due to an increase in the vertical pressure and suction. Fredlund (1964) and Flereau et al. (1993) noted that an increase in the vertical pressure and suction have similar influence on the volume change behaviour of expansive clays for a large range of applied suction. Marcial et al. (2002) stated that due to isotropic loading conditions, the void ratio decrease due to a suction increase was greater than that obtained due to a vertical pressure increase. Fredlund and Rahardjo (1993)s results on silty clay (Fig. 1) showed that the volume change behaviour
Highly plastic saturated clays generally exhibit a significant volume decrease due to a change in the water content caused by an increase in suction and stress. The volume change behaviour of saturated clays can be studied either by subjecting clays specimens to an increase in suction or an increase in the vertical pressure. Oedometer tests are generally carried out to establish the vertical pressure-void ratio relationships. The relationship between the water content and suction of the soil at zero external stress is called the soil-water characteristic curve (SWCC). Similarly, to establish the suction-void ratio SWCC, the volume of the soil specimens are measured at several applied suctions. The suction-water content and the suction-void ratio SWCCs are extremely useful while dealing with analysis and prediction of engineering behaviour of unsaturated clays. On the other hand, the pressure-void ratio relationships are used to calculate settlement and heave of saturated soils due to stress changes. Comparisons of consolidation test results with suction loading results for initially saturated soils are limited in the literature. Fredlund (1964) compared the volume change behaviour of Regina clay due to suction and vertical pressure loading. Fredlund and Rahardjo (1993) reported the suction-void ratio and the pressure-void ratio relationships for a silty clay (Fig. 1). Fleureau et al. (1993) compared the volume change behaviour of a Ca-bentonite and a kaolinite due to suction,
Figure 1. Influence of matric suction and net stress on volume change behaviour of an initially saturated silty clay; w water content, e void ratio, Gs specific gravity of soil solids, (ua uw) matric suction, (ua) net stress, ua pore air pressure, uw pore water pressure, am, at, bt, and bm, are the slopes of the corresponding plots; initial conditions with a subscript 0 (from Fredlund and Rahardjo, 1993).
381
of the soil due to an increase in suction and net stress (isotropic consolidation tests were carried out) were similar until the air-entry value of the soil was reached. They stated that once the soil starts to desaturate, a matric suction change will not be as effective as a total stress change in producing a volume change (in Fig. 1, the matric suction-void ratio plot indicated by B, remained above that of the net stress-void ratio plot, indicated by A). On the other hand, a matric suction increase is more effective than a net normal stress increase in removing water from the soil (in Fig. 1, the suction-wGs plot indicated by C remained below that of plot A). Direct mechanical interactions between soil particles govern the volume change behaviour of granular soils. On the other hand, the volume change behaviour of clays is primarily governed by the physico-chemical forces arising due to the clay-fluid-ion interactions. In the latter, the mineralogical and chemical properties are known to influence the fabric and structure of clays that in turn influence the changes in the water content and the void ratio due to suction and stress increase. The intent of the study was to investigate in detail the effects of an increase in suction and vertical pressure on the volume change behaviour of initially saturated clays. Two highly plastic clays with wide variations in their mineralogical and plasticity properties were chosen for the study. 2 BACKGROUND
The influence of physico-chemical forces on the volume change behaviour of clays has been brought out in the past (Sridharan and Jayadeva 1982; Tripathy et al. 2006; Tripathy and Schanz 2007). The charged clay platelets, the exchangeable cations, and the water molecules form an electrical double layer around the clay platelets. In a clay-water ion system, if two clay platelets are brought to a distance that is less than twice the double layer thickness, overlapping of the electric double layers will occur. This in turn will cause repulsion between the clay platelets. For an applied load, clays with smaller specific surfaces would attain smaller void ratios than that for clays with larger specific surfaces, all other factors, such as the temperature, the exchangeable cations, and the bulk fluid characteristics being the same. The structure and fabric of saturated clays composed of kaolinite and montmorillonite have been reported by several researchers (Sridharan et al. 1971; Colins and McGown 1974; van Olphen 1977; Delage and Lefebvre 1984; Mitchell 1993; Marcial et al. 2002). The bonding between the structural units of montmorillonite is via exchangeable ions and water. The structure of montmorillonite at very high water contents is usually described to have
edge-to-edge flocculated (i.e., edge to edge association of aggregates) and aggregated structure (i.e., face to face association of several clay platelets) (Fig. 2a). Group interaction takes place between the aggregates at the edges and as well as between the aggregates separated by the pore fluid which is being present in the larger void spaces. Overlapping of the diffuse double layers occurs within the aggregates. The structural units of kaolinte are usually strongly bonded to one another to form thicker clay particles (clay aggregates) (Fig. 3a). Also, due to a smaller specific surface and cation exchange capacity, the diffuse double layer thickness is relatively thin. Thus, the van der Waals attractive pressure is dominant that leads to an aggregated but flocculated structure (i.e., edge-to-face clay particle arrangement) at very high water contents. The group interactions persist via the weakly formed diffuse double layer on the outer surfaces of the aggregates and the positively charged edges. A large quantity of the pore fluid is expected to be trapped within a relatively strong flocculated clay structure (Fig. 3a). The concentration of the pore fluid in the bigger pores, in the double layers, and in the adsorbed water layer (i.e., the Stern layer) depends upon the distribution of the available exchangeable ions in the clay. In montmorillonites, most of the exchangeable ions generally reside very close to the clay platelet surfaces and in the interlayer spaces. A lesser amount of ions may be expected in the bigger pores. On the other hand, a minority of the exchangeable cations can be expected to be within the interlayer pores for kaolinite.
Figure 2. Particle association for montmorillonite during drying: (a) at liquid limit, (b) commencement of desaturation, and (c) at residual/no shrinkage stage.
Figure 3. Particle association for kaolinite during drying (a) at liquid limit, (b) commencement of desaturation, and (c) at residual/no shrinkage stage.
382
MATERIALS AND METHODS
Laboratory tests were carried out on two highly plastic clays (MX 80 bentonite and Speswhite kaolin) with wide range of plasticity characteristics. The properties of the clays are presented in Table 1. The specific gravity of soil solids, the Atterbergs limits, the specific surface areas, and the cation exchange capacities for the clays were determined following standard laboratory procedures. X-ray diffraction patterns of the clays indicated that the dominant mineral present in MX 80 bentonite was montmorillonite and that in Speswhite kaolin was kaolinite. The cation exchange complexes of the clays clearly indicated that the clays were not homoionised; rather, different proportions of Na, Ca, Mg, and K ions were present in the clays. The fractions of Na, Ca-Mg, and K ions in MX 80 bentonite were found to 57%, 42%, and 3%, respectively, whereas the fractions of these cations in Speswhite kaolin were 33%, 55%, and 12%. The initial water content under the laboratory environment was found to be 11% for MX 80 bentonite and 0.7% for Speswhite kaolin. Clay-water mixtures were prepared by mixing clay powders thoroughly with deionized water to targeted initial water content equals to about 1.15 times the respective liquid limit of the clays. The clay-water mixtures were then stored in sealed plastic bags and kept in air-tight containers to allow for water equilibration to take place for about seven days prior to preparing for the test specimens. During this period, the clay-water mixtures were mixed intermittently. All the laboratory tests were carried out at near constant temperature of 25C. To study the desorption behaviour of the clays, thick-walled cylindrical stainless-steel specimen rings were used to prepare the clay specimens. The inner surfaces of the specimen rings were lubricated using silicon grease. The specimen ring was made to rest on a wet filter paper kept over a flat surface. The filter paper used had a sufficiently greater diameter than that of the specimen ring. Clay-water mixture was carefully worked into the ring, a part at a time, using a spoon and a spatula. Light tapping helped to eliminate air bubbles from the clay specimen.
Table 1. Properties of clays studied. Clays Properties Specific gravity Liquid limit (%) Plastic limit (%) Specific surface area (m2/g) MX 80 bentonite 2.8 437 39 676 Speswhite kaolin 2.61 51.9 28 48
The excess clay was trimmed off using a spatula. The mass of the clay specimen along with the filter paper and the lubricated specimen ring was measured. 3.1 Suction-water content SWCC
The suction-water content SWCCs for the clays were established by employing pressure plate tests and salt solution tests to cover a wide range of suction. The pressure plate tests (Fredlund and Rahardjo 1993) were carried out by applying air pressures of 10, 30, 50, 100, 200, and 400 kPa, using a 5 bar pressure plate extractor. In the pressure plate tests, the applied air pressure is equal to matric suction of the soil specimen at equilibrium. At each suctions, at least three specimens were tested. Sufficient care was exercised while preparing the duplicate clay specimens. The mass of the clay specimens was monitored periodically to infer the suction equilibration. Figure 4 shows the transient time versus water content decrease for the clays studied to better visualize the equilibration time at various suctions considered in the pressure plate tests. In Figure 4, the equilibration time for both clays studied were about one to two weeks for all suctions considered. It is important that during the pressure plate tests, the ceramic disks remain saturated to avoid air-entrapment and delay in the expulsion of water from the soil specimens (Leong et al. 2004). The salt solution tests (i.e., the vapour equilibrium technique), were carried out in closed-lid desiccators for inducing higher suctions of 3.3, 7.5, 21.8, 38, 82.3, 113.3, and 294.8 MPa. Saturated salt solutions of K2SO4, KNO3, KCl, NaCl, Mg(NO3)2, K2CO3, and LiCl, were used (ASTM E 104-85 1998). The clay specimens tested in the desiccators were first subjected to a matric suction of 400 kPa in the pressure plate device before being transferred to the test desiccators containing various saturated salt solutions. The procedure adopted saved significant testing time. The relative humidity within the closed-lid desiccators was periodically measured using a commercial relative humidity probe. It was noted that the targeted relative humidity was well within the measurement error range of the humidity probe for all cases. The relative humidity of saturated salt solutions is related to the total soil suction (Fredlund and Rahardjo 1993). The equilibrium time for salt solution tests varied between about two weeks to a couple months (Fig. 5). In general, a greater equilibrium time was required at lower suctions. 3.2 Suction-void ratio SWCCs and pressure-void ratio relationships
The final volumes of the clay specimens at all suctions considered in the pressure plate tests and salt
383
550
(a) MX 80 bentonite Matric suction (kPa) 10 30
250
(a) MX 80 bentonite Total suction (MPa) 3.3 7.5 21.8 300
500
200
450
Water content (%)
50 200
Water content (%)
100 400
400
150
350
100
300
50
250
200 0
70
(b) Speswhite kaolin Matric suction (kPa) 10 30 50 100 200 400
0
5 10 15 20
10
20
30
40
50
60
70
35 30 25 Water content (%) 20 15 10 5 0 0
(b)
Speswhite kaolin
Total suction (MPa) 7.5 21.8 38 82.3 113.3 300
65 60
Water content (%)
55 50 45 40 35 30 0 5 10 Time ( days) 15
20
10
20
30
40
50
60
70
Time ( days)
Figure 4. Equilibration time plots in pressure plate tests for (a) MX 80 bentonite and (b) Speswhite kaolin.
Figure 5. Equilibration time plots in salt solution tests for (a) MX 80 bentonite and (b) Speswhite kaolin.
solution tests were measured using the wax method (ASTM D 494308 2008). The duplicate clay specimens tested were used for this purpose. The void ratios of the clay specimens were calculated based on the volume-mass relationships and the equilibrium values of the following at each applied suctions: (i) the total volume of the specimen measured by wax method, (ii) the total mass of the specimen, and (iii) the water content. Knowledge of the density of the molten wax at specified test temperature is crucial for precise determination of the volume of soil specimens. The density of the molten wax used was measured to be 0.745 Mg/m3 in this study. Consolidation tests were carried out on initially saturated clay specimens in a conventional fixedring oedometer. The maximum applied vertical pressure for the consolidation tests was 3.2 MPa. The initial water content of the clay specimens in the consolidation tests were kept similar to that of the specimens tested in the pressure plate tests. Though the applied maximum vertical pressure was far smaller than the maximum applied suction of 300 MPa, useful comparison of the volume change behaviour of the clays could be made. No attempt was made to measure the water content at each loading steps for the consolidation tests. The odeometer test results were corrected for the system deformations at each loading steps.
The suction-void ratio SWCCs (i.e., s versus e) and suction versus wGs plots (i.e., s versus wGs) for MX 80 bentonite and Speswhite kaolin are shown in shown Figures 6 and 7. The consolidation test results (i.e., p versus e, equivalent to p versus wGs) for the clays are included in the Figures 6 and 7 for comparison. For MX 80 bentonite (Fig. 6), the s-e, the s-wGs, and the p-e results were found to be very nearly similar up to a suction or vertical pressure of about 200 kPa, indicating that both vertical pressure and suction had similar influence on the volume change behaviour of the bentonite. For suctions or vertical pressures greater than 200 kPa, the p-e results remained distinctly below that of the s-e and the s-wGs plots indicating that the vertical pressure was more effective in changing the volume of the clay than that due an increase in suction. A lower void ratio of the bentonite due to an applied vertical pressure than that due to an applied suction is attributed to an oriented fabric of the bentonite (in relative terms). For suctions greater than 200 kPa and less than about 10 MPa, a similar s-e and s-wGs results was obtained. This indicated that the bentonite remained saturated during the desorption process up to a suction of about 10 MPa.
384
14
MX 80 bentonite
G s = 2.8, w L = 437% Suction vs e Suction vs wGs Vertical pressure vs e
Void ratio, e or water content X specific geravity, wGs
12
10
8 Suction vs e Suction vs wG s Vertical pressure vs e Air-entry value ? 10 MPa 2
0 1 10 100 1000 10000 100000 1000000 Suction /Vertical pressure (kPa)
Figure 6. Influence of suction and vertical pressure on volume change behaviour of MX 80 bentonite.
1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 1 10 100 1000 10000 100000 1000000 Suction/ Vertical Pressure (kPa)
Suction vs e Suction vs. wG s Vertical pressure vs e Air-entry value ? 0.8 MPa
Speswhite kaolin
Gs = 2.61, w L = 51.9%
Void ratio, e or water content X specific geravity, wGs
Suction vs e Suction vs wGs Vertical pressure vs e
Figure 7. Influence of suction and vertical pressure on volume change behaviour of Speswhite kaolin.
Note that the s-e and s-wGs data are based on independent void ratio and water content measurements. At suctions greater than the air-entry value (about 10 MPa), the s-e plot remained above that of the s-wGs plot. Comparing the test results in Figure 6 with the results for the silty clay in Figure 1, the difference in the volume change behaviour of the bentonite and that for the silty clay was only in the position of the p-e plot, whereas the positions of the s-e and s-wGs plots for the bentonite and the silty clay remained similar. For Speswhite kaolin (Figure 7), the s-e and the s-wGs plots were found to be similar up to a suction of about 800 kPa. However, the p-e plot was found to remain distinctly above that of the s-e and the
s-wGs plots indicating that the volume change due to a vertical pressure increase was less than that due to an increase in suction. Various factors may be attributed to such a behaviour; (i) there could have been some pre-stressing on the kaolin during specimen preparation stage and (ii) in kaolinte the diffuse double layer is relatively thin and the clay aggregates are usually in an edge to face arrangement that possibly offered a greater resistance to compression under K0-loading condition than that occurred due to isotropic suction loading. Since the specimen preparation method adopted for the suction tests and the consolidation test were similar and that the s-e plot exhibited a near linear relationship for suctions less than 800 kPa, the second factor seems to be more justifiable. At suctions greater than about 800 kPa, the s-e plot remained clearly above that of the s-wGs plot indicating that the air-entry value of the clay was reached. Additionally, the s-wGs plot remained below that of the p-e plot indicating that an increase in suction was more effective in reducing the water content of the clay than due to an increase in the vertical pressure. Comparing the test results in Figure 7 with the results for the silty clay in Figure 1, the positions of the s-e, the s-wGs, and the p-e plots were found to be very similar for both soils after the air-entry values were reached. The energy required for removing the pore fluid of clays from the bigger pores, the overlapped diffuse double-layers in the inter layers, and from the adsorbed water layers, differs quite significantly and depends upon the interactions of these fluids with the mineral surfaces and aggregates. Destauration of soils may be expected to begin when the adsorbed water layers of two soil particles have overlapped. In bentonites, the exchangeable cations present in the pore-fluid of the bigger voids contribute to the repulsive pressure between the aggregates structure. Similarly, the exchangeable cations present in the inter layers are responsible for the repulsion between the clay platelets within the aggregates. Additionally, the edge to edge interaction between aggregates is via the positively charged edges of the mineral crystals and the available anions in the clay. The void ratio decrease due to an increase in suction and vertical pressure initially is accompanied by the compression of the bigger voids (Marcial et al. 2002). This occurs due to the movement of aggregates against relatively weak resistance offered at the edge to edge contact regions. Once a more uniform pore-size distribution is attained, the compression of the interacting diffuse double layers commences within the inter layers. A possible fabric and structure of bentonites at commencement of desaturation and at residual shrinkage stage are shown in Figs. 2b and c. In Figs. 2a, b, and c, the
385
number of clay platelets has been kept the same and the qualitative volume decrease from saturated state to residual shrinkage state is noticeable. For kaolinite clays, it is hypothesized that an increase in suction initially causes the pore water to expel from the thin double layer formed around the clay aggregates followed by the expulsion of the fluid from the regions of the overlapping adsorbed water layers of the clay aggregates. Compression of the bigger voids is resisted by the edge to face aggregate matrix resulting in a lesser volume change. With a further increase in the applied suction, the air-water interface retreats inwards into the edge to face contact regions of the clay aggregates. Desaturation of kaolinite possibly occurs quite abruptly (Fig. 3b). A steep slope in the s-wGs plot in Figure 7 clearly indicated such a phenomenon. Since most of the pore water expelled immediately after the desaturation point was reached (Fig. 7), at large applied suctions the pore water expulsion took place from the edge to face and face to face contact regions accompanied by an insignificant change in the void ratio (Figs. 3c and 7). In Figs. 3a, b, and c, the number of clay aggregates has been kept the same. The volume decrease for Speswhite kaolin from saturated state to no shrinkage stage was found to be far lesser than that occurred for MX 80 bentonite. The test results clearly indicated that a change in suction did not always result in the same change in effective stress. For MX 80 bentonite, the void ratio changes were found to be similar at smaller suctions or vertical pressures, whereas for Speswhite kaolin the void ratio changes were quite dissimilar for the vertical pressures and suctions considered in this study. This implies that the structure of clays plays a significant role on their volume change behaviour and that there may not be a unique relationship between the degree of saturation and the related effective stress parameter associated with suction for highly plastic clays. 5 CONCLUSIONS
behaviour of clays due to an increase in suction and vertical pressure. The study showed that a vertical pressure increase was more effective in reducing the void ratio and water content as compared to an increasing in suction for the bentonite studied. On the other hand, a suction increase was more effective in reducing the water content of the kaolinite. For suctions beyond the air entry values, the void ratio changes for both clays were small. REFERENCES
ASTM D 494308. 2008. Standard test methods for shrinkage factors of soils by the wax method. ASTM International, West Conshohocken, Pa. ASTM E 10485. 1998. Standard practice for maintaining constant relative humidity by means of aqueous solutions. 1998 Annual book of ASTM standards, Vol. 11.03, Water and Environmental Technology, pp. 781783. Collins, K. & McGown, A. 1974. The form and function of microfabric features in a variety of natural soils. Gotechnique 24: 223254. Delage, P. & Lefebvre, G. 1984. Study of the structure of a sensitive champlain clay and its evolution during consolidation, Canadian Geotechnical Journal 21: 2135. Fleureau, J.M. Kheirbek-Saoud, S. Soemitro, R. & Taibi, S. 1993. Behaviour of clayey soils on drying-wetting paths. Canadian Geotechnical Journal 30: 287296. Fredlund, D.G. 1964. Comparison of soil suction and one dimensional consolidation characteristics of a highly plastic clay. M.Sc. thesis, University of Alberta, Edmonton, Alta. Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils. New York: John Wiley and Sons. Leong, E.C. Tripathy, S. & Rahardjo, H. 2004. A modified pressure plate apparatus. ASTM Geotechnical Testing Journal 27(3): 322331. Marcial, D. Delage, P. & Cui, Y.J. 2002. On the high stress compression of bentonites. Canadian Geotechnical Journal 39: 812820. Mitchell, J.K. (2nd ed.). 1993. Fundamentals of soil behaviour. New York: John Wiley and Sons. Sridharan, A. Altschaeffl, A.G. & Diamond, S. 1971. Pore size distribution studies, Journal of the Soil Mechanics and Foundation division, ASCE 91: 771786. Sridharan, A. & Jayadeva, M.S. 1982. Double layer theory and compressibility of clays. Gotechnique 32(2): 133144. Tripathy, S. Kessler, W. & Schanz, T. 2006. Determination of interparticle repulsive pressure in clays. ASCE Geotechnical Special Publication 147: 21982209. Tripathy, S. & Schanz, T. 2007. compressibility behaviour of clays at large pressures. Canadian Geotechnical Journal 44: 355362. van Olphen, H. (2nd ed.). 1977. An Introduction to Clay Colloid Chemistry: For Clay Technologists, Geologists and Soil Scientists. New York: Wiley-Interscience.
One-dimensional volume change and desorption behaviour of two highly plastic clays were studied. Initially saturated clay specimens were subjected to an increase in suction employing axis-translation technique and salt-solution technique to cover a suction range of 10 kPa to 300 MPa. Consolidation tests were carried out with a maximum applied vertical pressure of 3.2 MPa. The water content and the volume of the clay specimens at each applied suctions were measured. It was shown that mineralogy and the physico-chemical interactions between the clay particles and the pore fluid have a significant influence on the volume change
386
Compressive strength of unsaturated fissured clays

L.E. Vallejo
Department of Civil and Environmental Engineering, University of Pittsburgh, Pittsburgh, USA
ABSTRACT: Unsaturated clays often contain isolated fissures. This study reports laboratory and theoretical investigations on the mechanics of fissure propagation and failure of clay samples subjected to unconfined compression tests. The samples tested had fissures inclined at 15, 30, 45, 60 and 75 degrees with the vertical. They also had water contents (w) of 3, 9, 15, 22 and 27%. The samples with w smaller than 20% were found to behave like brittle materials and their failure was by fissure-induced tension. The samples with w in excess of 20% were found to behave like plastic materials and their failure was in shear. The strength of the samples that failed in tension was found to increase with an increase in w. The strength of the plastic samples was found to decrease with an increase in w. The failure modes of the samples was explained using the pendular, funicular, and saturated water distribution in the samples. 1 INTRODUCTION content (w) was about 30%, cracks were artificially made in the samples by a process of inserting and removing thin glass sheets 1 mm in thickness and 2,5 cm in width in a direction normal to the samples face. The cracks were made at 15, 30, 45, 60 and 75 degrees with the direction of uni-axial compressive load (vertical direction). In order to investigate the effect of water content (w) on the uni-axial compressive strength of the fissured clay samples, groups of clay samples with 5 different water contents were tested in the laboratory. These five different water contents were developed in the samples by allowing them to air dry after the cracks were made in the samples (w 30%). By allowing the samples to air dry (air temperature 20C) for about 2 hours, 4 hours, 2, 3 and 5 days, the final water contents in the samples were measured to be 27, 22, 15, 9, and 3%. After these final water contents were reached in the samples, they were tested under a uni-axial compressive stress conditions. The rate of deformation used in the testing was 0.5 mm/min. 2.1 Mode of failure of the samples
Some clays forming part of slopes and earth dams exist in the fissured state (Duncan and Dunlop, 1969, Sherard, 1973) For the assessment of the short-term stability of these earth structures, the unconfined compressive, qu, of the clays is needed. Very little is known about the role fissures play in the failure process of unsaturated clays when subjected to compressive loads. The purpose of this study is to report laboratory and theoretical investigations designed to understand the role of water content and its distribution on the unconfined compressive strength of fissured clays. 2 LABORATORY AND THEORETICAL ANALYSES
2.1 Laboratory preparation of samples For the experimental investigation, laboratory prepared samples of kaolinite clay with pre-existing cracks were used. The kaolinite used in the experiments had a LL 58% and PL 28%. The dry kaolinite was mixed with distilled water to form a soft soil mass with water content of about 40%. After the mixing was done, the clay-water mixture was placed in plexiglass containers that measured 7.62 cm in length, 7.62 cm in width and 4 cm in thickness. The samples were consolidated in these containers using a weight that applied a normal pressure of 25.7 kPa for a period of 5 days. After removal of the weight, the samples had consolidated and measured 7.62 in length and width and about 2.54 cm in thickness. Immediately after removal of the samples, when their water
Under uni-axial compressive stress, the samples experienced two very different type of failure depending upon the water content (w) in the samples. For the samples with w less than 20%, the failure of the samples was controlled by the fissures in the samples. Under compression, the failure started at the tip of the pre-existing cracks in the form of secondary cracks. These secondary cracks grew in the samples from the tip of the original cracks in a direction that was parallel to the uni-axial compressive loads (Figs. 1 and 2).
387
Figure 1. w 20%.
Mode of failure of fissured clay samples with
Figure 3. w 20%.
Mode of failure of fissured clay samples with
Figure 2. Mode of failure of a clay sample with a crack inclined at 60 degrees with respect to vertical load (w 3%).
Figure 4. Closing of the fissure under uni-axial compression in a clay sample with w 27%.
For the samples with w 20%, the samples first experienced a closing of the fissures. After closing of the fissures, the samples failed in shear, with the shear plane developing in the lateral section of the samples (Figs. 3, 4 and 5). 2.2 Type of stresses causing failure in samples with w 20%
Figure 5. Failure in the lateral section of the fissured clay sample with w 27%.
Theoretical investigations conducted by Brace and Bombolakis (1963), Kobayashi (1970), Ashby and Verral (1977), and Gdoutos (1984) have indicated that when samples of brittle material containing an inclined open crack is subjected to uni-axial compressive stresses, the tips of the crack concentrate stresses. According to Ashby and Verral (1977), the upper-right and lower-left regions of the crack tip are regions of intensified tensile stresses (Fig. 6). The upper-left and lower-right regions of the crack tip are regions of intensified compressive stresses. Since soils are weaker in tension, the samples will fail first in the regions of tensile stresses (Figs. 1, 2 and 6). This distribution of stresses around cracks in brittle materials seem to apply to the fissured samples with w 20%.
2.3
Influence of the crack inclination and water content on the failure of the samples
Figure 7 shows a plot of the uni-axial compressive strength (given by the maximum uni-axial compressive stress, c ), versus the inclination of the pre-existing fissures in the samples, . The uni-axial compressive strength reported in Figure 7 is the one at which the samples failed in tension by the developing of secondary cracks (Figures 1 and 2) or failed in shear. (Figures 3, 4 and 5). The plot in Figure 7 has also been done for samples having five different water contents.
388
type of stresses that caused the failure of the samples and the influence of the water content and its distribution in the unsaturated clay samples. 2.4 The influence of the type of stresses and the amount and distribution of the water in the unsaturated clay samples
Figure 6. Tensile and compressive stresses around an open inclined crack in a prismatic brittle material under uni-axial compression.
Figure 7. Relationships between the uni-axial compressive strength, c and crack inclination, , for clay samples with different water contents, w.
Figure 7 shows that the inclination of the preexisting cracks have a marked influence on the compressive strength of the fissured clay samples. For the case of the samples with water contents (w) less than 20%, Figure 7 shows that the compressive strengths reached minimum values when the preexisting cracks in the samples reached a critical angle, , that varied between 45 and 60. For the case of the samples with w 20%, the samples reached a maximum c when the crack inclination in the samples was about 55. Figure 7 also indicates that for the case of the brittle samples (w 20%), the uni-axial compressive strength of the samples increased with an increase in water content. For the case of the samples with w 20%, Figure 7 indicates that the c of the samples decreased with an increase in w in the samples. This behavior is explained next using the
In order to explain why the brittle (w 20%) fissured clay samples experience an increase in unaxial compressive strength with an increase in water content, the amount of water and its distribution in the samples will be used. According to Schubert et al. (1965) and Kim and Hwang (2000) the state of water saturation can be either pendular, funicular and saturated (Fig. 8). The amount of water in the unsaturated samples increased in value as well as in its distribution in the samples from the pendular, funicular and finally in the saturated state. These states create negative pore water pressures in the samples. The effect of the negative pore water pressures is to pull together the clay particles in the samples. This pulling together of the clay particles by the negative pore water pressures will make the samples more resistant against failure in tension (Figs. 1, 2 and 6 show tension zones in samples with w 20%). Also the higher the water content in an unsaturated sample, the larger the number of clay particles affected by the negative pore water pressures and thus the larger the tensile strength of the samples (Fig. 7). As the sample become close to saturation (w 30%), the samples tend to develop positive pore water pressures when subjected to compression. These positive pore water pressures will cause a decrease in compressive strength in the samples (see Fig. 7 for samples with w 22% and 27%). Recent reports by Vesga and Vallejo (2005, 2006) on the tensile strength of unsaturated kaolinite samples with varying degrees of water content seem to substantiate the previous contention. Vesga and Vallejo (2005, 2006) conducted direct and indirect tensile strength tests on samples of kaolinite with different water contents. The indirect tensile tests were of the Brazilian type. Figure 9 shows the type of tensile strength tests conducted by Vesga and Vallejo (2005, 2006), Figure 10 shows the results of the strength tests in function of the water content in the samples.
Figure 8. Degree of saturation in unsaturated soils: (a) pendular, (b) funicular, and (c) saturated (after Schubert et al., 1975).
389
Figure 9. Direct and indirect tensile strength tests on kaolinite clay samples (Vesga & Vallejo, 2005, 2006).
in a direction parallel to that of the axial load. When w 20% in the samples, they failed in shear. (3) The unsaturated samples that failed in tension (w 20%) had a pendular and funicular state of water saturation. These states induced negative pore water pressures that pull the particles together. This in turn causes an increase in the unconfined compressive strength of the samples. The samples with w 20% were close to saturation (w 30%). These samples tend to develop positive pore water pressures when subjected to compression, thus, their unconfined compressive strengths decreases with an increase in water content. (4) For the samples with w 20%, their unconfined compressive strengths were at minimum values for samples with cracks inclined between 45 and 60 with respect to the uni-axial load. The samples with w 20% had maximum values of compressive strength when the cracks in the sample were inclined at 55 with respect to the uniaxial load. REFERENCES
Ashby, M.F. & Verral, R.A., 1977. Micro-mechanisms of flow and fracture and their relevance to the rheology of the upper mantle. Philosophical Transactions of the Royal Society of London, 288A, 5995. Brace, A.W. & Bombolakis, 1963. A note on brittle crack growth in compression. J of Geophysical Research, 68 (12), 37093713. Duncan, J.M. & Dunlop, P., 1969. Slopes in stiff fissured clays and shales. Journal of the Soil Mechanics and Foundation Division, ASCE, 95 (2), 467491. Gdoutos, E.E., 1984. Problems of Mixed Mode Crack Propagation. Martinus Nijhoff Publishers, The Hague, Holland. Kim, T.H. & Hwang, C., 2003. Modeling of tensile strength on moist granular earth material at low water content. Engineering Geology, 69, 233244. Kobayashi, S. 1970. Fracture criteria for anisotropic rocks. Memoirs of the Faculty of Engineering. Kyoto University, Kyoto, Japan, 32 (3), 307333. Schubert, H., Herrmann, W. & Rumpf, H. 1975. Deformation behavior of agglomerates under tensile stresses. Powder Technology, 11, 121131. Sherard, J.L., 1973. Embankment dam cracking. In: Embankment Dam Engineering, Casagrande Volume. Hirschfield R.C. & Poulos, S.J. (Eds). Wiley Intersciences, New York, 271353. Vesga, L. & Vallejo, L.E., 2005 Strength of an unsaturated kaolinite clay under suction pressures. Proceedings of the 16th International Conference on Soil Mechanics and Geotechnical Engineering, Osaka, Japan, 2, 611614. Vesga, L.F., & Vallejo, L.E., 2006. Direct and indirect tensile tests for measuring the equivalent effective stress in a kaolinite clay. In: Unsaturated Soils, ASCEs Geotechnical Special Publication No. 147, Miller, G.A., Zapata, C.E., Houston, S.,& Fredlund, D.G. (Eds), 12901301.
Figure 10. Tensile strength of unsaturated kaolinite clay samples With different states of unsaturation (Vesga & Vallejo, 2006).
An analysis of Figure 10 shows that as the degree of saturation increases in the unsaturated samples (from pendular to funicular), the tensile strength of the samples increases. An increase in tensile strength in the samples will also represent an increase in uniaxial compressive strength in the prismatic samples with water content w 20% (Fig. 7). Figure 10 also shows that as the water content in the samples reaches the saturation value (w 30%), the tensile strength of the samples decreased in value with an increase in water content. Also, an increase in water content will cause the samples to become plastic and fail in shear (Figs. 3 to 5). 3 CONCLUSIONS
Laboratory and theoretical investigations designed to understand the effect of fissures in unsaturated samples of clay subjected to unconfined compression strength tests indicated the following: (1) The samples have two different modes of failure depending on the water content (w) in the samples. When w 20%, the fissures controlled the mode of failure of the samples. The samples failed in tension. They developed secondary tensile cracks that extended from the tip of the pre-existing cracks
390
Effect of suction and plasticity on the shear strength of sand/silt mixtures

M.V. Villar, P.L. Martn & R. Gmez-Espina
CIEMAT, Madrid, Spain
A. Lloret
UPC, Barcelona, Spain
ABSTRACT: This paper presents results of an experimental study on the effect of suction on the shear behaviour of sand/silt mixtures as determined with direct shear tests performed under suction control. The device used to perform these tests is a direct shear apparatus in which the shear box has been replaced by an air pressure chamber where matric suction can be applied by increasing the air pressure in the cell chamber while the water pressure at the bottom of the sample remains atmospheric (axis translation technique). The samples have been initially consolidated under constant normal stresses between 100 and 900 kPa for different matric suctions ranging from 0 to 1200 kPa. Afterwards, while the normal stress and the matric suction are kept constant, the samples have been sheared. The results obtained in the shear tests show the expected increase in strength with suction, particularly in cohesion, although the increase is less significant towards high suctions and for the high vertical stresses. In the range of stresses tested the relation between suction and cohesion is linear. The friction angle with respect to suction is lower than the internal friction angle, and decreases with the level of suction until reaching a constant, low value for the suctions corresponding to the residual zone of the water retention curve. 1 INTRODUCTION to its preference over direct shear testing for constitutive modelling studies, the latter is still useful for solving geotechnical problems related to prediction of stability of slopes and embankments, bearing capacity of foundations and pressure against earth-retaining structures. In addition, the main advantage of shear testing of unsaturated soils is that, due to the short drainage path in the specimen, the time to failure is greatly shortened. This paper presents results of an experimental study on the effects of suction on the shear behaviour of sand/silt mixtures as determined with direct shear tests performed under suction control. 2 2.1 MATERIAL AND METHODS Material
The mechanical behaviour of an unsaturated soil is affected differently by changes in net normal stress than by changes in matric suction. The suction tension within an unsaturated soil causes additional grain-to-grain forces and consequently shear strength increase. Thus, the shear strength, f, of an unsaturated soil can be considered to consist of an effective cohesion, c, and independent contributions from the net normal stress, ua, and the matric suction, ua uw. The effective angle of internal friction, , can be associated with the shear strength contribution from the net normal stress state variable; and the angle b, is related to the shear strength contribution from the matric suction stress state variable (Fredlund et al. 1978): f c ( ua) tan (ua uw) tan b (1)
Current knowledge assumes that both friction angles are similar for matric suctions lower than the air entry value and that b is smaller than for higher suction values. In turn, it has been proven by several authors that increases with suction (e.g. Wheeler & Sivakumar 1995). Although the improvement of unsaturated triaxial testing equipment in the last decades has led
Sand/silt mixtures of 70/30 and 50/50 dry weight proportion have been used in the investigation (Villar et al. 2006). The sand used is a pure quartz sand with most of the grain sizes comprised between 0.1 and 1 mm, for which a specific gravity of 2.65 has been assumed. The silt is composed by moscovite (39%), quartz (30%), caolinite (12%), smectite (9%) and feldspars. It has a liquid limit of 51 percent, a plasticity index of 24 percent, and
391
58 percent of the particles are silt-sized, the remaining being of clay size. The specific gravity is 2.81. For the 70/30 mixture, the optimum water content for normal proctor energy is 12 percent and the dry density reached is 1.96 g/cm3, which have been the initial conditions of the samples tested. The permeability of the mixture compacted at these conditions is 6 107 m/s and it has no swelling pressure. In order to allow a better comparison, the 50/50 mixture has been also compacted at the same water content and dry density. Under these conditions the permeability of the 50/50 mixture is 7 108 m/s and its swelling pressure 194 kPa. 2.2 Methodology The device used to perform these tests is a direct shear apparatus in which the shear box has been replaced by an air pressure chamber (Figure 1) where matric suction can be applied by increasing the air pressure in the cell chamber while the water pressure at the bottom of the sample remains atmospheric (axis translation technique). Between the sample and the lower porous stone a semipermeable cellulose membrane is interposed. The design is based in that presented by Escario & Sez (1986). Square surface specimens of 25 mm side and 15 mm height have been manufactured by uniaxial compaction up to dry density 1.96 g/cm3 of the mixtures with an initial water content of 12 percent. Before compaction the sand/silt/water mixtures were allowed to stabilise for at least 24 h. The compaction pressure applied has been 43 MPa for the 70/30 mixtures and 12 MPa for the 50/50 mixtures. Air pressure has been applied by injecting nitrogen gas at a pressure controlled by an external manometer. The vertical load is applied by weights in a lever yoke, and the horizontal stresswhich is applied at a constant displacement rate by an electric motoris measured by a 500-kp load cell. Vertical and horizontal displacements are measured with 10-mm displacement transducers. All
the transducers are connected to the same data acquisition system. The samples have been initially consolidated under constant normal stresses between 100 and 900 kPa for different matric suctions ranging from 0 to 1200 kPa. Suction and normal stress were applied simultaneously to the samples. Afterwards, while the normal stress and the matric suction were kept constant, the samples have been sheared at a velocity of 0.001 mm/min, in order to assure drained conditions. At the end of the shear tests, the gravimetric water content of the samples has been determined by oven drying, what has allowed to determine the water retention curves at failure. Also, the water retention curves of the mixtures have been determined with the contact filter paper method. The specimens needed for this determination were obtained by uniaxially compacting the mixtures at dry density 1.95 g/cm3 with different water contents. Three specimens were prepared for each water content, and they were allowed to stabilise in contact with the filter paper for two weeks at a temperature of 20C. Afterwards, the water contents of the filter paper and the sample were determined by oven drying and the filter paper water content was correlated to matric suction through a calibration curve. 3 RESULTS
3.1 Shear tests Consolidation of the samples took two days for the 70/30 mixture and between 3 and 5 days for the 50/50 mixture, the difference in permeability accounting for this disparity. During the shear tests the samples experienced a small initial vertical contraction followed by significant dilation, which was higher the lower the vertical stress and the higher the suction. Time to failure was between 20 and 38 h for the 70/30 mixture, corresponding to horizontal displacements between 0.85 and 1.76 mm which increased with vertical stress. For the 50/50 mixture the time to failure and corresponding horizontal displacements were slightly higher. As an example, Figure 2 shows the evolution of shear stress and vertical displacement for the shear tests performed with the 70/30 mixture under a normal stress of 900 kPa. It can be observed that the value of the final stress reached after failure is rather independent of the suction level. Also, as for the rest of normal stresses tested, the higher the suction the larger the dilatancy. Figure 3 shows the void ratios at the end of the tests performed with the 70/30 mixture; the final volume is larger the higher the suction under which the sample was tested. The results of the shear tests experimental programme have allowed the determination of the
Figure 1.
Controlled-suction shear box.
392
Shear strength, f (kPa)
1800 1600 1400 1200 1000 800 600 400 200 0 0.0 1.0
1800 1600 1400 1200 1000 800 600 400 200 0

s=0 kPa s=400 kPa s=800 kPa s=1200 kPa
Shear stress (kPa)
s=0 s=400 kPa s=800 kPa s=1200 kPa
s=0 kPa s=400 kPa s=800 kPa s=1200 kPa 2.0 3.0 4.0 5.0
0.5 0.4 0.3 0.2 0.1 0.0 -0.1 0.0 1.0 2.0 3.0 4.0 5.0
200 400 600 800 Net normal stress, - u a (kPa)
1000
Figure 4. Shear strength for different matric suctions. Open symbols: 70/30 mixture, filled symbols: 50/50 mixture.
600 Cohesion, c ' (kPa) 500 400 300 200 100 0 0 400 800 1200 Matric suction (MPa) cohesion friction 58 54 52 50 48 46 44 42 1600 Internal friction angle, ' 56
Figure 2. Evolution of shear stress and vertical displacements during the shear tests performed with the 70/30 mixture under normal stress 900 kPa and different matric suctions.
0.48 0.47 Final void ratio 0.46 0.45 0.44 0.43 0.42 0 200 400 600 800 1000 Net normal stress, - u a (kPa) s=0 s=400 kPa s=800 kPa s=1200 kPa
Figure 5. Strength parameters deduced from the shear tests. Open symbols: 70/30 mixture, filled symbols: 50/50 mixture.
Figure 3. Void ratio at the end of the shear tests performed under different suctions with the 70/30 mixture.
shear strength envelopes in the f, ( ua) and f, (ua uw) planes. So far most of the tests have been performed with the 70/30 mixture. Figure 4 shows the shear strength at failure as a function of the net normal stress for the different suctions tested. As expected, shear strength increases with suction for both materials. As shown in Figure 5, this increase is mainly due to the contribution of cohesion, which increases lineally with suction. The friction angle for samples tested under suction could be considered constant and higher than that for the saturated sample, with an average value of 54 for the 70/30 mixture and 47 for the
50/50 mixture. Logically, the more clayish mixture shows higher cohesion and lower friction angles. When shear strength is plotted as a function of suction (Figure 6), it is observed that the increase in shear strength with suction is not linear, and that the slope of the envelope decreases with suction. Two ranges of suction could be considered, with a threshold between 400 and 800 kPa. For the low suctions, b takes values between 17 and 39, i.e. smaller than , whereas for higher suctions, a value between 5 and 7 has been found. These observations apply just to the 70/30 mixture, since not enough data are yet available for the 50/50 mixture. 3.2 Water retention curves At the end of the shear tests, the gravimetric water content of the specimens tested was determined by oven drying. The results obtained for the 70/30 mixture are plotted in Figure 7 as a function of the suction during the test. These can be considered the water retention curves at failure. A clear trend
393
1800 1600 Shear strength, f (kPa) 1400 1200 1000 800 600 400 200 0 0 400 800 1200 1600 Matric suction, u a - u w (kPa) ns=100 kPa ns=300 kPa ns=500 kPa ns=900 kPa
0.35 Volumentric water content (%) 0.30 0.25 0.20 0.15 0.10 0.05 1 70/30 0 kPa 100 kPa 300 kPa 500 kPa 900 kPa
50/50
10 100 1000 10000 Matric suction (kPa)
Figure 6. Shear strength versus matric suction for different normal stresses (ns) indicated in the legend. Open symbols: 70/30 mixture, filled symbols: 50/50 mixture.
7.8 7.6 Water content (%) 7.4 7.2 7.0 6.8 6.6 6.4 6.2 0 100 kPa 300 kPa 500 kPa 900 kPa 300 600 900 1200 1500 Matric suction (kPa)
Figure 8. Retention curves obtained at failure and in unstressed specimens (points labelled 0 kPa and continuous lines, corresponding to the van Genuchten fittings). Open symbols: 70/30 mixture, filled symbols: 50/50 mixture. The initial conditions of the shear tests are pointed by arrows.
Figure 7. Retention curves of the 70/30 mixture obtained at failure under different normal stresses (indicated in the legend).
to find higher water contents as the normal stress during the test is lower can be observed, especially towards high suctions. Also, it seems that the effect of normal stress on the retention curve is less significant for low suctions. The volumetric water content at the end of the shear tests has been computed considering the dry density at the end of the consolidation under suction stage for each test. The values obtained for the two mixtures are plotted in Figure 8 as a function of the matric suction during the test. The values obtained with the contact filter paper method for unstressed samples compacted to similar initial dry density (1.95 g/cm3) are also included in the figure, as well as the van Genuchten fittings for these values. For the 70/30 mixture, the water content values obtained at failure are in the order, or slightly higher, than those obtained for unstressed samples
at the same matric suctions. However, for the 50/50 mixture, the water content values obtained at failure are much lower than those obtained under unstressed conditions, since, being a softer material, the volume change during consolidation and shear has been higher. According to the curves obtained for the unstressed specimens, the suction of the samples at the beginning of the shear tests (dry density 1.96 g/cm3, water content 1213%), would be of 6 kPa in the case of the 70/30 mixture and of 1500 kPa in the case of the 50/50 mixture (both values are indicated in the figure). Consequently, and taking into account that the range of suctions applied in the suction-controlled shear tests was between 400 and 1200 kPa, at the end of the shear test there has been a significant decrease in water content with respect to the initial one in all the tests performed with the 70/30 mixture. For the 50/50 mixture, the decrease in water content at the end of shear with respect to the initial one is smaller, since the change in suction has been less important. 4 DISCUSSION
A clear increase in shear strength with suction has been observed for the two mixtures tested. This seems to be due mainly to a linear increase in cohesion with suction and to an overall increase in internal friction angle of the unsaturated material with respect to the saturated one. This increase in friction angle shows no clear dependence on the suction level. The friction angle with respect to matric suction, b, is lower than , as is usually the case,
394
and decreases with suction. Figure 9 shows the van Genuchten fitting for the water retention curve of the 70/30 mixture with indication of the stages suggested by Vanapalli et al. (1996). It can be observed that the suctions applied in the shear tests are well above the air entry value (AEV) and very close to the residual zone. According to these authors, the following stages, related to the strength behaviour, can be identified: Up to the air-entry value there is a linear increase in shear strength with suction. This stage has not been explored in our tests (except for the saturated tests). Friction angles b higher than those measured herein, i.e. closer to , would be expected. In the transition zone, the increase in shear strength with suction is not linear. Depending on the normal stress, the b values found in this work were between 17 and 39. The samples of the 50/50 mixture have been tested under suctions in this transition zone, but there are not enough data to determine b. In the residual zone, which would start for a suction about 600 kPa for the 70/30 mixture, large increases in suction lead to relatively small changes in water content. In this region the friction angle b barely changes and has a constant and low value (between 5 and 7 in the tests reported here). The nonlinearity in the shear strength versus matric suction relationship was observed by Escario & Sez (1986) for several clays tested under matric suctions of up to 1000 kPa, and has been since repeatedly reported, as recently by Huat et al. (2005) for a residual soil tested under suctions of up to 350 kPa, by Sun & Xu (2007) for a sandy
0.35 Volumentric water content (%) 0.30 0.25 0.20 0.15 0.10 0.05 1
transition zone residual z.
silt in a range of suctions of up to 35 kPa, and by Estabragh & Javadi (2008) for a silt tested under suctions of up to 300 kPa. The results could be also interpreted in terms of an effective stress of the Bishops type (Bishop & Blight 1963), and for that, an alternative way of expressing Equation 1 using the parameter could be used: f c [( ua) (ua uw)] tan c tan (2) The parameter value changes with degree of saturation. To obtain its value from the shear tests results, the following definition can be used:
(Sr )
tan b f (ua uw ) tan f ( ua )
(3)
Figure 10 shows the values of obtained using this Equation and considering for Sr the average value of those corresponding to the suctions taken to compute tan b. The computation has been made both for the peak failure values and for the ultimate values at the end of the tests. The dispersion of the results can be attributed to the fact that the slopes have been obtained in a discrete way, what results in large errors. An alternative way of computing these values is to use a function of the type (Sr) Srn, what leads to: f c [( ua) Srn (ua uw)] tan (4)
The experimental results at peak failure and at the end of the tests have been fitted to this Equation by least squares, and the values of c, n and
1.0
AEV
100 kPa 300 kPa 500 kPa 900 kPa
0.8 0.6

0.4 0.2 0.0 0.0 0.2 0.4 0.6 0.8 Degree of saturation Eq. 4 at peak Eq. 4 at the end 1.0
10 100 1000 10000 Matric suction (kPa)
Eq. 3 at peak Eq. 3 at the end
Figure 9. Idealised stages of the water retention curve for the unstressed 70/30 mixture compacted at 1.95 g/cm3 and results obtained at failure under different normal stresses (indicated in the legend).
Figure 10. Values of parameter at peak failure and at the end of the tests (ultimate) as determined from Equation 3 and by least squares fitting of Equation 4.
395
2000 Shear strength, f (kPa) 1600 1200 800 400 0 0 400 ns = 100 kPa ns = 500 kPa 800 1200 ns = 300 kPa ns = 900 kPa 1600 Matric suction, u a - u w (kPa)
Figure 11. Values of peak strength for the 70/30 mixture measured (symbols) and computed using Equation 4 and the parameters in Table I (lines).
the relation between suction and cohesion is linear. The friction angle with respect to suction is lower than the internal friction angle, and decreases with the level of suction until reaching a constant, low value for the suctions corresponding to the residual zone of the water retention curve. During shear the samples dilate, to a higher degree as the suction is higher. The results obtained can be reproduced using the Bishops parameter as a potential function of the degree of saturation with an exponent close to 2. The application of normal stress reduces the water retention capacity of the mixtures, much more significantly in the case of the 50/50 mixture, due to its more clayish nature. The effect of normal stress on retention capacity seems lower for low suctions. REFERENCES
Bishop, A.W. & Blight, G.E. 1963. Some aspects of effective stress in saturated and partly saturated soils. Gotechnique 13(3): 177197. Cui, Y.J. & Delage, P. 1996. Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique 46(2): 291311. Escario, V. & Sez, J. 1986. The shear strength of partly saturated soils. Gotechnique 36(3): 453456. Estabragh, A.R. & Javadi, A.A. 2008. Effect of suction on compressibility and shear behaviour of unsaturated silty soil. In: Doll, D.G.; Augarde, C.E.; Gallipoli, D. & Wheeler, S.J. (eds.): Unsaturated soils: Advances in Geo-engineering. Proceedings of the first European Conference on unsaturated soils, E-UNSAT 2008. CRC Press/Balkema. Taylor & Francis Group, London. 449452. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsaturated soils. Can. Geotech. J. 15(3): 313321. Huat, B.B.K., Ali, F.H. & Hashim, S. 2005. Modified shear box test apparatus for measuring shear strength of unsaturated residual soil. Am. J. Applied Sci. 2(9): 12831289. Sun, S. & Xu, H. 2007. Determining the shear strength of unsaturated silt. In: In: Schanz, T. (ed.): Experimental Unsaturated Soil Mechanics. Springer Proceedings in Physics, vol. 112. Springer, Berlin. 195206. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction. Can. Geotech. J. 33(3): 379392. Villar, M.V., Udas, A., Caamn, I. & Robles, J. 2006. Estudio experimental de la deformabilidad de un material geolgico: caracterizacin inicial e identificacin de parmetros. Informes Tcnicos CIEMAT 1080. Madrid, 45 pp. Wheeler, S.J. & Sivakumar, V. 1995. An elasto-plastic critical state framework for unsaturated soils. Gotechnique 45(1): 3553.
Table 1. Parameters obtained by fitting the experimental results to Equation 4. c (kPa) Peak Ultimate 0 0 n 1.695 2.439 52.7 48.7 Average error (kPa) 62 83
tan have been thus obtained (Table 1). The variation of with Sr obtained using these parameters is also shown in Figure 10. Figure 11 shows that the values of strength at peak failure measured (Figure 6, 70/30 mixture) and those computed using Equation 4 with the parameters in Table 1 are in good agreement. With respect to the volumetric behaviour during shear, after a short initial period of contraction, the samples steadily dilate, more pronouncedly as suction is higher. The value of the final shear stress is higher for the samples tested under suction than for those saturated, but the particular value of final shear stress reached for a given normal stress is rather independent of suction. The same was observed in the shear tests under triaxial conditions and suctions of up to 1500 kPa performed with a silt by Cui & Delage (1996). 5 CONCLUSIONS
Controlled-suction shear tests have been performed in sand/silt mixtures of proportion 70/30 and 50/50. The results obtained in the shear tests show the expected increase in strength with suction, particularly in cohesion, although the increase is less significant towards high suctions and for the high vertical stresses. In the range of stresses tested
396
Experimental study on K0 coefficient of Ningming unsaturated expansive soil

H.P. Yang, J. Xiao, G.F. Zhang & R. Zhang
Changsha University of Science & Technology, China
ABSTRACT: The static lateral pressure coefficient, K0, is usually assumed to be a constant in engineering practice, which in most cases is not in accordance with the actual condition. Therefore, systematic tests of K0 coefficient on two kinds of Ningming unsaturated expansive soils were carried out using a GDS triaxial apparatus. For the same kind of soil, the K0 value increases approximately linearly as confining pressure increases, but nonlinearly as water content increases. The increase rate of K0 value is higher when the water content is larger than 18%. A function expressing K0 in terms of water content and confining pressure is established. This function provides some important parameters for establishing a practical constitutive model for Ningming unsaturated expansive soil. Furthermore, the research results can be useful for practical engineering applications.
INTRODUCTION
In geotechnical engineering the static lateral pressure coefficient, K0, is a very important parameter used to analyze the state of stress in level ground and to determine the earth pressure on retaining structures subjected to very small lateral displacement, the pore pressure of the earth dam during the construction period and the force of friction on pile shaft. Poisson ratio, , is also an important geotechnical parameter, which can be indirectly determined from its relationship with K0. Therefore, a large number of laboratorial and field tests were carried out, and testing methods as well as empirical and theoretical formulas for K0 had been put forward. Jakys earth pressure expression (i.e., K0 1 sin ) is commonly used. A theoretical formula of K0 for non-cohesive clay has been derived by Shi Hong-Yan and Xie Ding-Yi. In addition, a formula of K0 for unsaturated soil, considering the change in volume of the soil mass, has been established by Li Shun et al. However, at present K0 coefficient, for numerical analyses, is often assumed to be a constant by experience. That is not in accordance with the actual condition in most cases. Therefore, it is not safe for the design of engineering. In order to carry out the western expansion soil project of Ministry of Communications and establish an applied constitutive model for engineering, the variation in K0 with water content and confining pressure is established according to K0 consolidation tests on Ningming expansive soil conducted using a GDS triaxial apparatus system.
K0 CONSOLIDATION TEST FOR UNSATURATED EXPANSIVE SOIL
2.1 Testing instrument and control method There is a built-in K0 module providing K0 consolidation and swelling in the GDS triaxial apparatus system, which can be controlled in two ways (i.e., back pressure volume controller and radial transducer controller). The continuous K0 test allows a rate of cell pressure increase to be defined and the required back pressure, and then the axial displacement is modified to ensure that the specimen diameter does not change. The diameter can be controlled by the calculated diameter of the test specimen by volume change measurement. Alternatively, in this paper the local measurement of the diameter (Hall Effect or LVDT local gauges (Fig. 1)) was used to control the diameter. 2.2 Preparation of specimens
The Ningming gray-white and gray-black expansive soils for this test were obtained from Nan-you
Figure 1. Hall effect and LVDT local gauges.
397
Table 1.
Characteristics of gray-white and gray-black expansive soils. Particle content % Effective content of Specific Dry Liquid Plasticity montmorillonite surface Natural water density limit index Specific (m2/g) content (%) (g/cm3) (%) (%) gravity 5 mm 2 mm (%) 1.53 1.58 61.8 57.3 30.9 26.3 2.66 2.75 50.08 52.61 47.12 26.31 19.07 17.6 170.90 161.11
Specimen
Gray-white 25.9 Gray-black 23.9
a (kPa)
highway in Guangxi Zhuang Autonomous Region. The characteristics of gray-white and gray-black expansive soils are shown in Table 1. The specimens were prepared by static compaction, the diameter and height of which were 50 mm and 100 mm, respectively. The dry densities of all specimens were 1.58 g/cm3. In order to avoid the influence of dry density on test results, the difference of densities between various specimens was less than 0.02 g/cm3. The initial saturation degrees of specimens varied with the initial water contents that were constant during the test. Twelve groups of tests on gray-white expansive soil were performed under three levels of confining pressures (50 kPa, 150 kPa and 200 kPa) and four different initial water contents (10.5%, 18.2%, 20.4% and 25.9%) corresponding to the initial saturation degrees (22.4%, 38.8%, 43.5% and 55.3%). Sixteen groups of tests on gray-black expansive soil were performed under four levels of confining pressures (25 kPa, 50 kPa, 150 kPa and 200 kPa) and four different initial water contents (10.8%, 18.3%, 23.0% and 25.1%) corresponding to the initial saturation degrees (23.0%, 39.0%, 49.1% and 53.6%). 2.3 Testing programs
700 600 500
Water content 10.5%
a (kPa)
400 300 200 100 0 0 50
r (kPa)
100
150
Figure 2. Variation in a with r at water content of 10.5% for gray-white expansive soil.
300 250 200 150 100 50 0 0 50
Water content 25.1%
The tests were carried out in accordance with the procedure of standard triaxial K0 consolidatedundrained test in Specification of Soil Test (SL2370291999). The selected rate of speed for confining pressure was 10 kPa/h, which was suitable for K0 consolidation condition. After the targeted confining pressure was achieved, the specimens were consolidated in continuous K0 consolidation condition until the deformation was less than 0.01 mm/h. The relationship between the axial stress (a) and the lateral stress (r) must be met: r / a K0. 2.4 Testing results and analyses
r (kPa) 100
150
Figure 3. Variation in a with r at water content of 25.1% for grey-black expansive soil.
The representative relationship curves of a and r for Ningming gray-white and gray-black expansive soils are shown in Figures 2 and 3, respectively. The static lateral pressure coefficient obtained from the test results are listed on Table 2 and 3.
The variation in static lateral pressure coefficient K0 with water content for Ningming gray-white and gray-black expansive soils under different confining pressures are shown in Figures 6 and 7, respectively. In order to obtain the function expression, the Poissons ratio and static lateral earth pressure coefficient K0 are assumed equal to 0.5 and 1.0, respectively, when the water content is equivalent to the liquid limit (i.e., the liquid limits of Ningming
398
Table 2. r (kPa) w (%) a (kPa) K0 Table 3.
Results of k0 consolidation test on grey-white expansive soil. 50 10.5 210 0.165 18.2 224 0.228 20.4 180 0.272 25.9 130 0.377 150 10.5 629 0.237 18.2 438 0.340 20.4 371 0.401 25.9 267 0.561 200 10.5 749 0.266 18.2 528 0.377 20.4 402 0.495 25.9 296 0.673
Results of k0 consolidation test on grey-black expansive soil. 50 150 200
r (kPa) 25
w (%) 10.8 18.3 23.0 25.1 10.8 18.3 23.0 25.1 10.8 18.3 23.0 25.1 10.8 18.3 23.0 25.1 49 41 210 142 90 78 551 393 252 223 670 485 309 276 a (kPa) 117 75 0.212 0.333 0.515 0.613 0.238 0.351 0.553 0.639 0.272 0.382 0.595 0.672 0.299 0.412 0.647 0.274 K0
0.8
0.6
10.50% 20.40%
18.20% 25.90%
0.8
0.6
K0
K0
0.4
0.4
0.2
0.2
0 0 50 100
50 kPa 150 kPa 200 kPa

10 15
r (kPa)
150
200
250
Figure 4. Variation in K0 with r at various water contents for grey-white expansive soil.
0.8
w (%)
20
25
30
Figure 6. Variation in K0 with water content at various confining pressures for grey-black expansive soil.
0.8
0.6
0.6
K0
0.4
K0
0.4
0.2
10.80% 23.00%
0 50 100
18.30% 25.10%
200 250
0.2
r (kPa)
150
0 10 15
25 kPa 50 kPa 150 kPa 200 kPa w (%)

20 25 30
Figure 5. Variation in K0 with r at various water contents for grey-black expansive soil.
gray-white and gray-black soils are 61.8% and 57.3%, respectively). Figures 8 and 9 are obtained by Origin7.5 software to fit the relationship curves of Figs. 6 and 7, respectively. It can be seen from Figures 4 and 5 that the relationships between K0 and confining pressure under various water contents for Ningming gray-white
Figure 7. Variation in K0 with water content at various confining pressures for grey-white expansive soil.
and gray-black expansive soils are approximately linear. The K0 value increases as confining pressure increases under various water contents, and it increases as water content increases at a given confining pressure. As can be seen from Figures 6
399
1.0 0.9 0.8 0.7 0.6 K0 0.5 0.4 0.3 0.2 0.1 0.0 0.0 10.0 20.0
50kPa y=1+(0.12-1)/(1+exp((x-32.79)/7.62)) 150kPa y=1+(0.18-1)/(1+exp((x-26.81)/6.18)) 200kPa y=1+(0.21-1)/(1+exp((x-24.00)/5.06))
Table 4. Values of a, b and c parameters in various confining pressures. Specimen Gray-white Confining pressure r (kPa) 50 150 200 25 50 150 200 a 0.12 0.18 0.21 0.18 0.21 0.24 0.27 b 32.79 26.81 24.00 24.60 24.25 23.85 23.12 c 7.62 6.18 2.06 4.30 4.06 3.90 3.52
Gray-black
30.0 40.0 w (%)
50.0
60.0
70.0
Figure 8. Variation in K0 with water content for greywhite expansive soil.

1.0 0.9
35 30 25 a,b and c 20 15 10 5 y = 0.0006x + 0.09 R2 = 1 (a) 0 50 100
y = -0.0588x + 35.703 R2 = 0.9998 (b)
a b c
0.8 0.7 0.6 K0 0.5 0.4 0.3 0.2 0.1 0.0 0.0 10.0 20.0 50kPa y=1+(0.21-1)/(1+exp((x-24.25)/4.06)) 150kPay=1+(0.24-1)/(1+exp((x-23.85)/3.90)) 200kPay=1+(0.27-1)/(1+exp((x-23.12)/3.52)) 30.0 40.0 w (%) 50.0 60.0 70.0
y = -0.0167x + 8.5114 R2 = 0.9861 (c)
150
200
250
Confining Pressure r (kPa)
Figure 9. Variation in K0 with water content for greyblack expansive soil.
Figure 10. Variation in a, b and c with confining pressure for grey-white expansive soil.
and 7 the K0 value increases nonlinearly with water content increasing under various confining pressures. The increment of K0 value is higher when the water content is larger than 18%. 3 THE FUNCTION EXPRESSION OF K0 COEFFICIENT
Figures 10 and 11 indicating the variation in a, b and c with confining pressure for Ningming graywhite and gray-black expansive soils are obtained from Table 4. As can be seen from Figs. 10 and 11, it is clear that the relationship between the parameters a, b and c with confining pressure is a good linear, namely: a pa r qa b pb r qb c pc r qc (2) (3) (4)
From the above analysis, it is known that the K0 value is a function of water content and confining pressure: K0 f (w, r). As can be seen from Figures 8 and 9, the K0 coefficient can be expressed as a function as follows: K 0 (w ) 1 (a 1) /(1 e
(w b ) c )
(1)
Then substitute the equations of (2), (3) and (4) to the equation of (1). The function expression of K0 coefficient with water content and confining pressure can be obtained as follows: K 0 (w, r ) 1 (pa r qa 1)/(1 e
(w pb r qb ) pc r qc
where: K0 is the static lateral earth pressure coefficient; w is the water content of the soil; a, b and c are the fitting parameters associated with the characteristic of the soil and confining pressure, the values of which are given in Table 4.
(5) where: For Ningming gray-white expansive soil, the values of parameters of pa, pb, pc, qa, qb and qc are
400
30 25
y = -0.0074x + 24.743 R2 = 0.9303 (b)
a b c
a, b and c
20 15 10 5 0 0 50 100
y = -0.0038x + 4.3453 R2 = 0.9028 (c) y = 0.0005x + 0.1763 R2 = 0.9542 (a) r (kPa)

150 200 250
2. The laboratory test results indicate that K0 coefficient of unsaturated expansive soil is not a constant but a function of water content and confining pressure. 3. A function of K0 coefficient in terms of water content is established for water content variation below the liquid limit (K0 1.0), which is more valuable for the engineering practice than a constant K0 coefficient. This function can provide some important parameters for establishing practical constitutive models for Ningming unsaturated expansive soil. REFERENCES
Li, S.Q. & Chen, H. et al. 2006. Expressions of coefficients of lateral earth pressure for unsaturated soils considering change of volume mass. Journal of Liaoning Technical University 25(2): 207210. (in Chinese). Li, X.P. & Zhao, Y.P. 2007. Discussion on the still lateral pressure coefficient and testing method. Journal of Railway Engineering Society, 8: 2022. (in Chinese). Shi, H.Y. & Xie, D.Y. 2001. A formula to determine the lateral pressure coefficient of cohesionless soil. Journal of Hydraulic Engineering 4: 8588. (in Chinese). The Ministry of Water Resources of the Peoples Republic of China. 1999. Specification of Soil Test. Beijing. (in Chinese). Yasuhiko, O. & Fumio, T. 1984. Some factors affecting K0: values of sand measured in triaxial cell. Soils and Foundations 24(3): 5268.
Figure 11. Variation in a, b and c with confining pressure for grey-black expansive soil.
0.0006, 0.0588, 0.0167, 0.09, 35.703 and 8.5114, respectively; For Ningming gray-black expansive soil, the values of parameters of pa, pb, pc, qa, qb and qc are 0.0005, 0.0074, 0.0038, 0.1763, 24.743 and 4.3453, respectively. 4 CONCLUSIONS
1. The K0 coefficient is mostly assumed to be a constant in engineering practice, which in most cases is not in accordance with the actual condition. Therefore, systematic laboratory tests on two kinds of Ningming unsaturated expansive soils were carried out using a GDS triaxial apparatus.
401
Experimental investigation of stress increment direction dependency of plastic flow in unsaturated soils
S. Zargarbashi & N. Khalili
Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering, University of New South Wales, Sydney, NSW, Australia
ABSTRACT: An experimental program has been conducted to investigate the stress path dependency of plastic strain rate on a low plasticity normally consolidated clay. The effect of unsaturation on the stress increment direction dependency of plastic potential is also examined. Stress probing is conducted at various degrees of unsaturation and at different stress directions on a series of identical samples. The experimental results are discussed and prevailing trends are highlighted. 1 INTRODUCTION 2 2.1 SOIL PROPERTIES AND TESTING PROCEDURE Test soil and sample preparation
There have been many constitutive models in the literature for the stress-strain behaviour of soils. They range from simple elasto-plastic models, involving a single yield surface and a single plastic potential, to sophisticated models based on the sub-yielding of the material within the virgin yield surface, multiple yield surfaces, kinematic hardening and bounding surface plasticity among others (Burland, 1967, Mroz et al., 1979, Dafalias, 1986, Inel & Lade, 1997, Khalili et al., 2008). The conventional elasto-plastic models form the basis for a majority of the contemporary models proposed in the literature. The more advanced models are used where the non-linearity and/or the cyclic response of the soil are concerned. Common to a majority of these constitutive models is the use of a stress direction independent plastic potential whilst it is not supported by the experimental observations; especially in stress states not close to the failure envelope (Lewin & Burland, 1970, Graham et al., 1983, Costanzo et al., 2006). This has rendered their numerical prediction for stress-strain behaviour of soils less than reliable particularly along complex non-proportional loading paths. Only limited experimental data is currently on the stress-path dependency of plastic strain increment in soils (Lewin & Burland, 1970, Costanzo et al., 2006). In this work, an experimental testing has being conducted to investigate the stress path dependency of plastic strain direction in a fine-grained soil in both saturated and unsaturated states. Stress increment direction dependency of the plastic potential on a normally consolidated sample of the soil. The results obtained are presented and discussed.
The soil used in the experiments was a low plasticity clay from Buffalo Dam in Victoria, Australia. The main properties of the test soil are summarized in Table 1. Prior to compaction, the soil was carefully wetted to the standard Proctor optimum moisture content and placed in a sealed plastic bag for at least 24 hours to allow for moisture equilibrium. Compaction was performed in five equal layers in a greased 50 mm diameter split mould. For each layer, a known mass of soil was compressed with a constant rate to a thickness of 20 mm to a target dry density of 18 kN/m3. The interface between the
Table 1. Soil Liquid limit, % Plastic Index, % Specific Gravity Maximum dry density*, kN/m3 Sample dry density*, kN/m3 Optimum Moisture Content*, % Air entry value**, kPa USCS Symbol Percent 4.74 mm Percent 75 m Percent 2 m Critical state friction angle, degrees Index properties of the test soil. Buffalo Dam Clay 39.6 15.9 2.6 18.2 18.0 16.2 60 CL 100 68.5 24 34.5
*Obtained using the standard proctor technique. **Obtained using the pressure plate technique at the void ratio of interest.
403
layers was scarified to minimize formation of weak zones in the specimens. 2.2 Testing equipment and procedure
Table 2. Test ID C1 C2 C3 C4 C5 C1-U200 C2-U200 C3-U200 C4-U200 C1-U350 C2-U350 C3-U350 C4-U350
Summary of test program. Origin C C C C C C C C C C C C C * (Deg) 45 135 225 315 90 45 135 225 315 45 135 225 315 Suction (kPa) 200 200 200 200 350 350 350 350 Remark Saturated Saturated Saturated Saturated Saturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated Unsaturated
All tests were conducted using a modified Bishop & Wesley-type triaxial cell capable of testing soils in an unsaturated state using axis translation technique. The cell is equipped with a computerized stress path controller and data acquisition system. This equipment controls hydraulic pressures and can run tests automatically for extended periods along any specified stress path through high-resolution, digitally controlled, pressure/volume controllers. Axial load is measured by an internal load cell. Cell and pore-water pressures are measured using pressure transducers. Air pressure is controlled using a regulator and measured by a pressure gauge. All pressures applied could be resolved and controlled to within 0.5 kPa. The deformations encountered during a loading process were estimated by measuring the changes in height and volume of the samples. The changes in height were measured by an external LVDT with a 25 mm range. It has been shown that external measurements are reliable to an axial strain of 0.05% [Costanzo et. al., 2006]. As such, no internal transducers were used in this work. For saturated samples, volumetric changes were measured by a digital volume change controller used to a resolution of 0.001 cc. The volume of unsaturated samples was measured using an automatic digital image capturing system. 3 TESTING PROGRAM
* is loading direction angle as shown in Figure 2.
Figure 1. Stress paths followed to initial stress states and the corresponding examined stress probes (saturated tests).
The experimental program consisted of 5 drained stress-probes on saturated samples and 8 constantsuction stress-probes on unsaturated specimens, starting from the initial stress states C, as summarized in Table 2, and shown in Figure 1. In addition to these test, supplementary tests were also performed to examine the repeatability of test results. The stress probes applied are depicted in deviatoric stress, q, and mean effectives stress, p, plane as shown in Figure 1. The stress state C was reached by stresscontrolled consolidation along a q/p 0.42 path to q 125 kPa. After initialization of the sample to the stress point of interest, the stress state was maintained for 48 hours to allow dissipation of any excess pore pressure generated during shearing. In the case of unsaturated samples, due to the contribution of matric suction to the mean effective stress, the loading path adopted was at a slightly different q/p ratio (Table 3) such that after application of the suction, the effect stress state traversed to the stress state C.
Table 3. Calculation of the stress ratio of the loading path adopted in unsaturated tests before stress probing. Test ID C1-U200 to C4-U200 C1-U350 to C4-U350 * 0.516 0.380 s (kPa) 200 350 s*** (kPa) 103.1 132.7 q/p 0.635 0.747
* is the effective stress parameter based on Khalili & Khabaz (1998) equation. **Contribution of matric suction to effectives stress.
The effective stress in the unsaturated soil was calculated using the relationship for the effective stress parameter proposed by (Khalili & Khabbaz, 1998). It should also be pointed out that several constantsuction one-dimensional consolidation tests were also conducted on this soil at different suctions; the
404
samples were prepared with the similar methods and at the same dry density as this study. No suction hardening effect was observed in these tests. Therefore, in this study, the size of the yield surface is considered to be independent of suction. Following the work of Costanzo et. al. (2006), the stress probes were applied at a constant stress rate of 2.5 kPa/h. Although, a small stress probe would be sufficient to investigate stress direction dependency of plastic flow rule, some of the stress probes were continued to attain critical state or a prescribed maximum allowable cell pressure so that the dependency of the plastic potential to stress ratio could also be investigated. 4 RESULTS AND DISCUSSION
C2
1
C5
C1
q [%]
C3
-1
C4
-2 -2 -1 0 1 2
p [%]
(a)
0.2
The strain paths obtained from the saturated and the unsaturated stress probes are shown in Figures 2 & 3, respectively. The results are presented in terms of changes of shear strain, q, versus volumetric strain, v. As shown in Figure 2, the stress probes in which the stress ratios increase faster, as expected, show larger strains. As can be seen from both figures strain increment direction is strongly dependent on the stress path increment. As stress paths C2 and C5 approach the CSL, as expected the slope of the plastic strain increment approaches infinity. On the other hand, along stress path C1, C3 and C4, a rather constant slope of the plastic strain increment is obtained (Fig. 2a). Comparing the observed trends shown in Figures 2 & 3, there appears to be no discernible effect of suction on the dependence of the strain increment directions, at least in range of suctions considered. Figure 4 shows the variation of strain path direction versus stress path direction in the range of 180 to 180 degree (measured from horizontal axis in counter clock wise direction). Although, the data are too limited to draw any definite conclusion, it may be noted that, for the stress state C (i.e. the virgin state), strain path direction has a harmonic variation with the loading path direction about line. This observation need to be confirmed by further experiments and at different stress ratios and stress histories. These types of observations could shed light on an approach for future plastic flow rules taking into account the loading direction. 5 REPEATABILITY OF RESULTS
C5 C2
0.1
C1
q [%]
C4 C3
-0.1
-0.2 -0.2 -0.1 0 0.1 0.2
p [%]
(b)
Figure 2. Strain path obtained from tests C1 to C5 (a) at the vicinity of point C.
0.2
0.1
C2-U200
C1-U200
q [%]
C4-U200
-0.1
-0.2 -0.2 -0.1 0 0.1 0.2
To assess the reliability of the results, it is important that repeatability of the responses obtained from the tests is evaluated.
p [%]
Figure 3. Strain path obtained from tests C1-U200 to C4-U200 at the vicinity of point C.
405
180
due to much stiffer material response which is expected (Costanzo et al., 2006). 6 CONCLUSION
Strain path direction,
90
-90
-180 -180 -90 0 90 180
Stress probe direction,
Figure 4. Strain path direction variation with loading direction initiated from stress-state C.
A systematic experimental program has being conducted to investigate the loading direction dependency of the strain rate direction. Several stress probes in saturated and unsaturated states were carried out on identical samples, initiated from a common stress state point on the wet side of CSL. The results showed that the strain increment direction strongly depended on the loading direction; however, this dependency gradually reduced as the stress state moves closer to the CSL. It was also observed that unsaturation, for the suction value considered, did not affect the strain rate direction. Early observations appear to indicate that strain rate direction is a harmonic function of stress probe direction about the stress path direction. REFERENCES
0.2
C5
0.1
C5-r
-0.1
-0.2 -0.2 -0.1 0 0.1 0.2
p [%]
Figure 5. Strain path obtained in a pair of duplicate test, C5 and C5-r.
Figure 5 shows the result of a pair of such repeated tests in term of strain path. Test C5-r duplicates test C5. As evidenced, the repeatability of results is fairly good for this case in which the material shows a relatively soft response. However because of viscous deformations, a larger difference between the results of duplicate tests is likely for those stress probes in which a full stress reversal occurs with respect to the previous loading history
Burland, J.B. 1967. Deformation of soft clays. PhD, Cambridge University. Costanzo, D., Viggiani, G. & Tamagnini, C. 2006. Directional response of a reconstituted fine-grained soilPart I: Experimental investigation. International Journal for Numerical and Analytical Methods in Geomechanics, 30, 12831301. Dafalias, Y.F. 1986. Bounding surface plasticity. I: mathematical foundation and hypoplasticity. Journal of Engineering Mechanics, 112, 966987. Graham, J., Noonan, M.L. & Lew, K.V. 1983. Yield states and stress-strain relationships in a natural plastic clay. Canadian Geotechnical Journal, 20, 502516. Inel, S. & Lade, P.V. 1997. Rotational Kinematic Hardening Model for Sand. Part II Characteristic Work Hardening Law and Predictions. Computers and Geotechnics, 21, 217234. Khalili, N., Habte, M.A. & Zargarbashi, S. 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Computers and Geotechnics, 35, 872889. Khalili, N. & Khabbaz, M.H. 1998. Unique relationship for for the determination of the shear strength of unsaturated soils. Geotechnique, 48, 681687. Lewin, P.I. & Burland, J.B. 1970. Stress-probe experiments on saturated normally consolidated clay. Geotechnique, 20, 3856. Mroz, Z., Norris, V.A. & Zienkiewicz, O.C. 1979. Application of an anisotropic hardening model in the analysis of elasto-plastic deformation of soils. Geotechnique, 29, 134.
q [%]
406
Water retention and hydraulic behaviour
Laboratory determination of Soil-Water Characteristic Curves for cracked soil

M.M. Abbaszadeh, S. Houston, C. Zapata, W. Houston & B. Welfert
Arizona State University, Tempe, Arizona, USA
K. Walsh
San Diego State University, San Diego, California, USA
ABSTRACT: In general, unsaturated hydraulic conductivities are more difficult to estimate than are saturated, but as the soil transitions from intact to cracked, the difficulty in estimating the hydraulic conductivity and the water storage properties increases. One critical step in determination of unsaturated flow hydraulic properties lies in the evaluation of the Soil-Water Characteristic Curve (SWCC). In this paper, the authors experience with a series of laboratory studies of direct measurements of cracked soil SWCCs is presented, including challenges associated with very low suction control. An oedometer-type SWCC apparatus, capable of suction and net normal stress control was used in these lab studies. This paper also addresses how the Air Entry Value (AEV) of the cracks can be calculated theoretically based on capillary theory, and the results are compared against AEVs determined from laboratory tests. 1 INTRODUCTION These laboratory results provide much needed data for consideration of the effect of soils cracks on unsaturated soil property models, which has not been considered extensively. 2 BACKGROUND
The problem of estimating ground surface flux is one of great interdisciplinary interests, and the literature is replete with related articles from disciplines including soil science, geotechnical engineering, environmental ecology, hydrology, water resources, forestry, landscape architecture, geology, and environmental engineering. Surface flux is related to complex interrelationships between the soil and atmosphere, and soil anomalies such as cracks must be appropriately considered in any surface flux model. However, there is little data available for assessment of the effect of cracks on unsaturated flow properties, such as the SoilWater Characteristic Curve (SWCC) and unsaturated hydraulic conductivity. The properties and behavior of unsaturated cracked soil are potentially quite different from those of intact soil, and the absence of direct test data on cracked soil properties leads to uncertainty in the evaluation of surface flux conditions, particularly expansive soils. It is important to develop an improved understanding of cracked soil properties because seasonal cracking of soil results in poor estimates of runoff and infiltration due to the changing soil storage conditions (Arnold et al. 2005). The primary objective of this paper is to present preliminary laboratory-scale SWCC test results on cracked soils. These data can be used, for example, to validate previously proposed bimodal models for fractured soils (Zhang & Fredlund 2004).
Relationships for estimating unsaturated hydraulic conductivity based on the Soil-Water Characteristic Curve (SWCC) are commonly used, but have not been thoroughly evaluated for cracked soils. The measurement of the hydraulic conductivity for an unsaturated soil is extremely difficult, and the existence of cracks further complicates the measurement. For this reason, the SWCC has been used to predict the hydraulic conductivity of a cracked material (Peters & Klavetter 1988, Mallant et al. 1997, Khne et al. 2002, Liu et al. 2004, Zhang & Fredlund 2004). Thus, one challenge for predicting the hydraulic conductivity of cracked soil is determining the SWCC. Once the SWCC is established for a cracked soil, it is likely that predictive models can be used to estimate the unsaturated hydraulic conductivity function. For example, Chertkov & Ravina (2000) studied the shrinking-swelling behavior of clay including a network of capillaries to represent cracks in the soil. The SWCC for the cracked soil was determined by using the total crack volume and the volume of water-filled cracks, and a generalization of the van Genuchten (1980)Mualem (1976) model was used for estimating the unsaturated hydraulic
409
100
90 80 70 60 50 40 30 20 10 0 0.1 1 10 100 1000 10000 100000 1000000
will cause the soil to exhibit a bimodal behavior. The first hump in the bimodal SWCC is the AEV associated with the cracked phase, while the second hump is associated with the intact phase. 3.1 Soil characteristics The soil used in this study was obtained from a site near San Diego, California. The basic index properties of the soil are presented in Table 1. 3.2 Sample preparation
Soil suction (kPa)
Figure 1. Typical Bimodal SWCC for Cracked Soil (Zhang & Fredlund 2004).
conductivity function. Liu and Bodvarsson (2001) studied the use of the van Genuchten (1980) and Brooks & Cory (1964) models for the hydraulic conductivity of a fractured rock. In this paper only the SWCC relationship is studied. On-going research by the authors will be directed towards validation of the use of existing SWCC-based predictive models for unsaturated hydraulic conductivity determination for cracked soils. Zhang and Fredlund (2004) discuss that a fractured rock will produce a bimodal SWCC with a matrix phase and a fracture phase. The SoilWater Characteristic Curve of the fractured rock was presented as the sum of the effects of the two phases, weighted according to their respective porosities. The combined matrix and fracture medium was treated as a continuum with the same suction applied to the combined material. A computed Soil-Water Characteristic Curve for the rock matrix, the fractures and the entire fractured rock mass is shown in Figure 1 (Zhang & Fredlund 2004). Taking a similar continuum approach to cracked soils, the Soil-Water Characteristic Curve takes on a bimodal character. Several mathematical models for the SWCC are available that allow for a bimodal SWCC (Durner 1994, Burger & Shackelford 2001, Gitirana & Fredlund 2004). 3 LABORATORY TESTING PROGRAM
Samples were initially compacted in three layers inside stainless steel rings having a height of 25 mm and diameter of 61 mm. Both samples were prepared with soil passing the No. 4 sieve and at 18% water content. The specimens were compacted to 90% of the standard Proctor maximum dry density. Following compaction, the cracks were created in the specimens using aluminum shims. These aluminum shims were varied in thickness to represent different crack widths. A total of 15 and 14 cracks were created for samples No. 1 and No. 2, respectively (Fig. 2). The depth of the cracks
Table 1. Soil characteristics. % Sand % Silt % Clay LL PI Optimum water content Max dry density (g/cm3) 2.72 63 30 7 SC 42 25 18% 1.74 115
Specific gravity Particle size analysis Unified classification system Atterberg limits Standard proctor
Expansion index (ASTM)
Laboratory-scale tests were conducted to determine the SWCC for two artificially cracked compacted soil specimens. Only the drying curve of the SWCC is presented in this paper, although the study of SWCC hysteresis for cracked soils is the subject of on-going research by the authors. The suction range for these tests was 0.1 1300 kPa. An oedometer-type pressure plate device (Perez-Garcia, et al. 2008) was used to obtain the SWCC. A hanging manometer technique was used to apply very low matric suctions required to capture the bimodal nature of the SWCC. Theoretically, the existence of the cracks in the soil matrix
Figure 2. Artificially cracked sample No. 1 with 3.2% crack volume. (The crack volume for sample No. 2 was 4.4%).
410
ranged between 10 to 15 mm. The total volume of cracks was measured as a percentage of the overall specimen volume to be about 3 to 5%. A companion study was conducted with the same soil to evaluate the extent of cracks forming naturally due the drying and wetting cycles. From that study, it was concluded that the total volume of cracks was about 35% of the overall volume of the sample. One challenge in dealing with cracks is the uncertainty associated with the contributions of various factors controlling the cracks behavior. For example, the extent of crack healing during wetting is not entirely predictable. To capture the SWCC bimodal behavior, the cracks must be initially fully saturated, but should be wide enough so they will not heal completely upon saturation. Further, the cracks should be small enough to prevent desaturation by gravity drainage before any suction (air pressure) is applied using the pressure plate device. Based on preliminary observations, for the study soil, the cracks should be about 1.11.3 mm wide if a bi-model SWCC is to be observed in the laboratory. For this soil, cracks of 1.11.3 mm width were observed to shrink to 0.70.8 mm, after saturation. Although soil crack patterns have been broadly documented (Wells et al. 2003, Scott et al. 1986, Konrad & Ayad 1997, Yesiller et al. 2000, Velde 1999, Young 2000, Vogel et al. 2005, Tang et al. 2008), there is not yet any widely dominant pattern of crack formation identified for field conditions. Despite some restrictions on creating the cracks artificially, however, we have tried to make the specimen crack patterns reasonably consistent with crack patterns found in nature. 3.3 Test procedure, including aspects of testing at very low values of matric suction
suction constant, close monitoring is required on regular basis, becoming cumbersome in consideration of the lengthy test times required for equilibration, especially for highly plastic soils. Another point which makes tests at low suctions challenging is the fact that it is not possible to fully saturate the intact portion of the specimen because back-pressure saturation techniques are not easily employed in pressure plate testing. It was observed, at some very low suction stages of the SWCC test, that even after the specimen (likely the cracks in the specimen) released water at a prior, lower suction stages, the sample would tend to absorb water from the tube as the suction was increased (though still quite low). It is believed that this behavior is a result of the intact matrix part of the soil not having been fully saturated, even when the cracks were filled with water and extensive time for saturation of the specimen under submergence conditions had preceded the SWCC test. In other words, at early stages of the test, when the cracks are still full of water, the fractured phase of the soil dominates the behavior, while at later, higher suction stages, as the cracks dewatered, the intact soil matrix (not 100% saturated) governs the response. 4 THEORETICAL RELATIONSHIP BETWEEN WIDTH OF CRACK AND CAPILLARY RISE IN A CRACK
A hanging manometer technique was used to apply very low suctions. This method involves creating a negative pore water pressure (uw), while keeping the pore air pressure (ua) constant and equal to atmospheric pressure. This will result in matric suction being equal to the value of the negative pore water pressure because the matric suction is defined as ua uw. As the water elevation inside the water tubes is positioned lower than the base elevation of the oedometer, the sample experiences suction equals to uw. For instance, to apply 0.1 kPa, one has to create the elevation difference equal to 1.0 cm between the water level in the tubes and the cell base (where the sample sits). One of the major difficulties associated with setting a fixed low suction value is the continuous elevation change of the water that occurs inside the tube as the specimen seeks equilibration with the applied suction. Thus, the applied suction changes as the water elevation of the tube changes, and to keep the small applied
Should one idealize a soil crack as a capillary, there would always be a relationship between the width and depth of cracks which are capable of holding water inside of them. If the crack is too deep or too wide, the water cannot develop enough tension to overcome the self-weight of the water inside it, and as a result, the water will flow out of the crack, assuming the water pressure at the base of the crack is essentially zero. First, the relationship will be derived between height of capillary rise, hc, which will be assumed to be the crack depth, and the crack width, wc. A continuous rectangular section throughout the crack depth is assumed, which is equivalent to saying the crack has a constant width from top to bottom, as shown in Figure 3. This assumption is of interest because cracks of this shape were generated in the laboratory. Assuming the meniscus is fully developed and tangent to the side wall of the crack, the surface tension forces can be assumed to be vertical at the crack walls (Fig. 4) and equal to: Upward forces Ts 2 cm (1)
where, Ts surface tension force per unit of length (73 105 N/cm); and 2 cm is the total
411
1 cm 1 unit length of crack
hc
water-filled crack
wc
Figure 3.
Schematic of constant width crack.
Ts
hc
W 1 cm wc
Figure 4. Free Body Diagram (FBD) of unit length water element in crack.
nevertheless potentially useful as a rough guide in estimating the AEV of the cracks. Equation 4 was used to generate the results shown in Table 2, wherein the depth of crack ranges from 7 mm to 13 mm. This range in crack depth was chosen because, for this type of soil, naturally formed cracks, formed in laboratory condition, were about 10 mm deep, or slightly more. For each crack depth shown, it is assumed that the crack is full of water, the meniscus is fully developed and the surface tension at the top of the crack is just sufficient to balance the weight of water in the crack. Because the suction is given by ua uw and the preceding derivation was made for the case of ua 0, the only component of suction is the water tension, uw. Thus, the corresponding suction column in Table 2 can be thought of an equivalent AEV. At the suction shown, the dewatering is just commencing and air is starting to enter the crack. At crack sizes smaller than those shown in Table 2, the capillary model would predict that dewatering due to gravity alone would not occur. However, if ua were elevated to a value above zero, then ua would generate an additional downward force which, together with the weight, could be made to overcome the surface tension forces. The derivation can be repeated along the same lines as before, but a new force due to ua must be added to the free body diagram (Fig. 5). Again, for a unit length of crack, Equation 1 remains unchanged, but Equation 2 can be rewritten as follows: Downward forces hc wc 1 cm w ua (wc 1 cm) (5)
length over which surface tension acts, for a 1 cm segment. The downward forces are equal to the weight of water and therefore, equal to the volume times the unit weight of water: Downward forces hc wc 1 cm w (2)
For equilibrium in the vertical direction, and solving for ua (with hc and wc in cm) we get: ua 1.46 10 3 9.807 10 3 hc wc (6)
where, hc crack height; wc crack width; 1 cm 1 unit length of crack; and w specific weight of water (9.807 103 N/cm3). For equilibrium in the vertical direction, upward forces are equal to the downward forces: Ts 2 cm hc wc 1 cm 9.807 103 Solving for hc (with hc and wc in cm): hc 0.149 0.15 wc wc (4) (3)
Table 2. Depth of crack and corresponding suction for which cracks of various widths will just start to dewater due to gravity alone. Depth of crack, hc cm 0.7 1.0 1.3 mm 7.0 10.0 13.0 Corresponding suction hc w kPa 0.069 0.098 0.127 Width of crack that dewaters due to gravity alone cm 0.21 0.15 0.11 mm 2.1 1.5 1.1
It should be noted that Equation 1 is based on several simplifying assumptions which are at variance with actual field conditions, but it is
412
Ts ua ua
ua ua
Table 3c. Relationship between hc, wc, and ua for commencement of crack dewatering based on Equation 6, for hc 1.3 cm. wc ua mm 1.0 0.75 0.50 0.10 N/cm2 0.0018 0.0067 0.0164 0.133 kPa 0.018 0.067 0.164 1.33 uw hc w kPa 0.128 0.128 0.128 0.128 Suction (ua uw) kPa 0.146 0.195 0.292 1.46
hc
W 1 cm wc c
cm 0.1 0.075 0.05 0.01
Figure 5.
FBD including downward forces due to ua.
Note that hc controls uw, which is given by uw hc w. Also, hc controls the volume and weight of water in the crack. Due to these compensating effects, ua uw is insensitive to hc, as shown in Tables 3a, 3b, and 3c. Figure 6 shows that ua is also somewhat insensitive to hc. Now that different derivations were obtained for constant width crack, the same procedure can be used to derive new relationships for any shape
Table 3a. Relationship between hc, wc, and ua for commencement of crack dewatering based on Equation 6, for hc 0.7 cm. wc cm 0.2 0.15 0.1 0.075 0.05 0.01 mm 2.0 1.5 1.0 0.75 0.50 0.10 ua N/cm2 0.00044 0.0029 0.0077 0.0126 0.022 0.139 kPa 0.0044 0.029 0.077 0.126 0.22 1.39 uw hc w kPa 0.0686 0.0686 0.0686 0.0686 0.0686 0.0686 Suction (ua uw) kPa 0.073 0.098 0.146 0.195 0.289 1.459
Figure 6. Relationship between hc, wc, and ua for commencement of crack dewatering.
of crack. For instance, in case of a V-shape crack, the analysis remains the same, except the volume of the unit length of water becomes one half of that obtained in the case shown above.
Table 3b. Relationship between hc, wc, and ua for commencement of crack dewatering based on Equation 6, for hc 1.0 cm. wc cm 0.15 0.1 0.075 0.05 0.01 mm 1.5 1.0 0.75 0.50 0.10 ua N/cm2 0.0 0.0048 0.00967 0.0194 0.1362 kPa 0.0 0.048 0.0967 0.194 1.362 uw hc w kPa 0.098 0.098 0.098 0.098 0.098 Suction (ua uw) kPa 0.098 0.146 0.195 0.292 1.46
Quantification of a bimodal SWCC for cracked soil is difficult to accomplish for a variety of reasons, not the least of which is the difficulty in controlling the required extremely small suctions. It was possible to bracket the AEV on a test specimen as follows. An SWCC test specimen was prepared with cracks of an initial width, wc, of about 1 mm and a depth, hc, of about 10 mm. After wetting to essentially saturation a slight closing of the cracks was observed, to a width of about 0.75 mm. A value of ua of about 0.07 kPa was applied and the specimen was subsequently observed after time for equilibrium. A few of the cracks showed signs of starting to dewater. The test chamber was resealed and a new ua value of 0.1 to 0.2 kPa, say 0.15 kPa, was applied and allowed to equilibrate. Subsequent examination of the cracks showed that they were more or less completely dewatered. Thus, the ua causing initiation of dewatering was somewhere
413
between 0.07 kPa and 0.15 kPa, experimentally, and perhaps closer to 0.07 kPa. Of course, soil suction is given by ua uw, but it is common practice in both SWCC testing and triaxial shear testing, when using the axis translation technique, to neglect uw and simply equate suction to the imposed ua. However, for the case at hand, the ua values are so small that uw values are not necessarily negligible. Accordingly, ua uw was calculated for Table 3 using non-zero values in general for both ua and uw. For the purpose of comparing the experimental results with the theoretical results, it is most convenient to simply assume the uw values in the laboratory test specimen were the same as for the matching boundary conditions in the capillary model (Table 3). Then the comparison amounts to simply comparing the ua applied in the lab to the value of ua computed and entered in Table 3. If the wc is 0.75 mm and values from Table 3b are used, for which hc 1.0 cm, a value of ua 0.1 kPa is obtained. This value compares very well with the measured range of 0.07 to 0.15 kPa. For the meager amount of experimental data available to date, the capillary model compares very well with experimental data. Interestingly, the SWCCs for intact and cracked samples were found reasonably close after crack desaturation, as can be seen in Figure 7 where the curves are tending to merge at higher suctions. According to the tests conducted for cracked samples, bimodal behavior of the cracked soil is observed. The major difference between the cracked and intact curves is the initial volumetric water content, which should be expected, since the 3 to 4% crack volume significantly impacted the overall void ratio. These crack voids were almost completely saturated during the wetting process. Consequently, the volumetric water content would be higher for cracked sample compared to an
intact sample. As the suction increases from zero to 0.1 kPa, for these particular sizes of the cracks, nearly all of the cracks dewater, and the first break in the curve corresponds to the AEV of the cracks. The second break in the curve occurs somewhat above 10 kPa and is believed to represent the AEV of the intact matrix. 6 CONCLUSION
More comparisons would be needed before elevating the capillary model to a predictive model that should be used in lieu of measurements. But the comparison above strengthens the initial assumption that the capillary model can be used to obtain at least a rough estimate of AEV for cracks. Experimental results to date show that the SWCC for a cracked soil can be represented by a bimodal curve. However, the AEV of the cracks is very low, even for the relatively small width cracks considered in the laboratory study. Dewatering of larger field cracks would be expected to occur at extremely low suction values, and perhaps to dewater under gravity alone. ACKNOWLEDGEMENT This study was supported by the National Science Foundation (NSF) under grant number CMMI0825089. The opinions, conclusions, and interpretations expressed in this paper are those of the authors, and not necessarily of NSF. REFERENCES
Arnold, J.G., Potter, K.N., King, K.W. & Allen, P.M. 2005. Estimation of soil cracking and the effect on surface runoff in a Texas Blackland Prairie Watershed. Hydrol. Process 19(3): 589603. Brooks, R.H., & Corey, A.T. 1964. Hydraulic properties of porous media. Colorado State University Hydrology Paper, Fort Collins Nr. 3, Vol. 27, March. Burger, C.A. & Shackelford, C.D. 2001. Soil-Water Characteristic Curves and dual porosity of sanddiatomaceous earth mixtures. J. of Geotechnical and Geoenvironmental Engineering, ASCE 127(9): 790800. Chertkov, V.Y. & I. Ravina. 2000. Shrinking-swelling phenomenon of clay soils attributed to capillary-crack network. Theoretical and Applied Fracture Mechanics 34: 6171. Durner, W. 1994. Hydraulic conductivity estimation for soil with heterogeneous pore structure. Water Resources Research 30(2): 211223.
Figure 7. Comparison between cracked and intact SWCC for the same samples.
414
Gitirana, Jr. G. & Fredlund, D.G. 2004. Equations for the soil-water characteristic curve based on meaningful and mathematically independent parameters. ASCE J. of Geotechnical and Environmental Engineering 130(2): 209212. Konrad J.M. & Ayad, R. 1997. Desiccation of a sensitive clay: field experimental observations. Can. Geotech. J. 34: 929942. Khne, J.M., Khne, S. & Gerke H.H. 2002. Estimating the hydraulic functions of dual-permeability models from bulk soil data. Water Resources Research 38(7): 11211132. Liu, H.H. & Bodvarsson, G.S. 2001. Constitutive relations for unsaturated flow in a fracture network. J. of Hydrology 252: 1625. Liu. H.H., Bodvarsson, G.S. & Finsterle. S. 2004. A note on unsaturated flow in two-dimensional fracture networks. Technical Note. Earth Sciences Division. Lawrence Berkeley National Laboratory. Uni. of Calif., Berkeley. Mallant, D., Tseng, P.H., Torde, N., Timmerman, A. & Feyen, J. 1997. Evaluation of multimodal hydraulic function in characterizing a heterogeneous field soil. J. of Hydrology 195: 172199. Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12: 593622. Perez-Garcia, N., Houston, S., Houston, W., Padilla, J.M. (2008). An Odometer-Type Pressure Plate SWCC Apparatus, ASTM Geotech. Testing J., March. Peters, R.R. & Klavetter, E.A. 1988. A continuum model for water movement in unsaturated fractured rock mass. Water resources research 24(3): 416430.
Scott, G.J., Webster R. and Nortcliff S. An analysis of crack pattern in clay soil: its density and orientation, J. Soil Sci. 37 (1986), pp. 653668 Tang, C., Shi, B., Liu, C., Zhao, L., & Wang, B. 2008. Influencing factors of geometrical structure of surface shrinkage cracks in clayey soils. Engineering Geology 101: 204217. Van Genuchten, M. Th. 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44: 892898. Velde, B. 1999. Structure of surface cracks in soil and muds. Geoderma 93: 101124. Vogel, H.J., Hoffmann, H. & Roth K. 2005. Studies of crack dynamics in clay soil I. Experimental methods, results, and morphological quantification. Geoderma 125: 213223. Wells, R.R., DiCarlo, D.A., Steenhuis, T.S., Parlange, J.Y., Romkens, M.J.M. & Prasad, S.N. 2003. Infiltration and surface geometry features of a swelling soil following successive simulated rainstorms. Soil Science Society of America Journal 67(5): 13441351. Yesiller, N., Miller, C.J., Inci G. & Yaldo, K. 2000. Desiccation and cracking behavior of three compacted landfill liner soils. Engineering Geology 57: 105121. Young, I.M. 2000. An empirical stochastic model for the geometry of two-dimensional crack growth in soil (with Discussion). Geoderma 96: 263289. Zhang, L. & Fredlund, D.G. 2004. Characteristics of water characteristic curves for unsaturated fractured rocks. The Second Asian Conference on Unsaturated Soils, Unsat-Asia, Osaka, Japan: 425428.
415
The void ratio dependency of the retention behaviour for a compacted clay
C. Air Farulla
Dipartimento Ingegneria Strutturale, Aerospaziale e Geotecnica, Universit degli Studi di Palermo, Palermo, Italy
A. Battiato A. Ferrari
Agroscope Reckenholz-Tnikon Research Station ART, Tnikon, Ettenhausen, Switzerland Soil Mechanics Laboratory LMS, cole Polytechnique Fdrale de Lausanne, Lausanne, Switzerland
ABSTRACT: The paper presents the results of a comprehensive experimental programme aimed to characterize the retention properties of a compacted scaly clay by determining the retention curves in terms of main wetting and main drying branches at different values of void ratio, ranging between 0.45 and 0.80. Different experimental techniques were combined in order to explore the retention properties in a wide range of suction. Vapour equilibrium technique was used to impose total suction in the range 2 110 MPa, while the air overpressure technique was applied to control matric suction in the range 0.01 0.80 MPa. Collected results clearly pointed out a strong dependence of the retention curves on the void ratio. The dependences of the air entry value and of the hysteresis on the compaction state of the material are analysed. These aspects are discussed in the light of the microstructural arrangement resulting from the compaction process.
INTRODUCTION
A proper understanding of the water retention properties is a fundamental requirement in the analysis of unsaturated soils. Retention behavioural features have been in fact recognized to play a fundamental role both in the hydraulic and in the mechanical response of soils subjected to suction variations. Retention properties are stated by quantitative relationships between the suction and the amount of water which can be retained within soil pores, i.e. relationships between suction and water content, or suction and degree of saturation. Hysteretic features are observed in such relationships, resulting in different values of the degree of saturation attainable for a given suction. Moreover, recent developments pointed out that, especially for compacted clays, these relationships can significantly evolve with the void ratio variations (e.g. Romero & Vaunat 2000; Birle et al. 2008). These issues must be necessarily taken into account when the mechanical response is analysed in terms of the effective stress for unsaturated soils (Jommi & Di Prisco 1994; Wheeler et al. 2003;
Nuth & Laloui 2008), which requires the product between suction and degree of saturation to be computed. Moreover, when flow problems are considered, the dependency of the Darcy coefficient on the degree of saturation implies a non-unique relationship between suction and permeability. In spite of the recognized importance of these topics, the determination of the complete retention behaviour is not a routine practice and few experimental programs have been presented in order to focus on the quantification of the hysteretical features. When dealing with compacted clays, the main limitations are related to the wide suction ranges to apply in order to investigate the retention response for significant variations of the degree of saturation, and the high volume changes which can occur as a consequence of the suction application. With reference to these issues, the paper presents the results of a comprehensive experimental programme aimed to characterize the retention behavioural features of a clay compacted at different dry densities. Different experimental techniques were combined in order to explore the retention behaviour in a wide range of suction.
417
2 2.1
EXPERIMENTAL PROGRAMME Tested material
Tested samples were prepared using a stiff and highly fissured clay outcropping near Palermo (Sicily). The clay is subdivided by a thick network of discontinuities in stiff fragments (scales) whose size ranges from a few millimetres to a centimetre. It is a kaolinitic-illitic clay with liquid limit wl 58% and plasticity index Ip 30%. The specific gravity is Gs 2.78. The air-dried clay with a hygroscopic water content wh 0.05 (at a relative humidity of approximately 50%) was disaggregated by a rubber pestle, and the fraction passing at n 10 ASTM sieve (mesh aperture of 2 mm) was selected. The air dried clay was added with distilled water and carefully mixed. After a curing time of 2 or 3 days, samples were compacted statically into a small rigid cylinder mould, having height of 12 mm and diameter of 15 mm. Compacted clay is characterized by a double porosity network (Air Farulla & Jommi 1995; Air Farulla et al. 2007). More than 130 samples were prepared with different initial physical characteristics. Initial water content (w), dry unit weight (d), void ratio (e) and degree of saturation (Sr) of compacted samples varied in the limits: w 0.135 0.245; d 1.42 1.87 g/cm3; e 0.49 0.95; Sr 0.63 0.88. 2.2 Experimental techniques Two experimental techniques were combined in order to explore the retention properties of the compacted clay in a wide range of suction. Vapour equilibrium technique was used to impose total suction in the range 2 110 MPa, whereas axis translation technique (air overpressure method) was applied to control matric suction in the range 0.01 0.80 MPa. In the vapour equilibrium technique, total suction is imposed by controlling relative humidity in the soil pore gaseous phase. To this end samples are allowed to come in equilibrium with the relative humidity of a closed (i.e. at constant mass) reference system (Romero 2001). Psychometric law enables to convert relative humidity values in the corresponding total suction () values (Fredlund & Rahardjo 1993). The relative humidity in the reference system was controlled by means of aqueous solutions of sodium chloride (NaCl) with concentration variable between 0.45 and 6.08 molNaCl/kgH2O, corresponding to relative humidity values respectively of 0.98 ( 2 MPa) and 0.75 ( 38 MPa) at 20C (Romero 2001). Higher values of total suction were achieved by air drying samples at a temperature of 20C and a relative humidity varying between 0.44 0.48
corresponding to total suction in the range of 110 100 MPa. Vapour equilibrium technique was applied using airtight glass jars (capacity of 500 ml), half filled up with aqueous sodium chloride solutions. Using an appropriate support the compacted clay samples were set into the jars above the solution level, allowing water vapour transmission by simple diffusion. During the equalization steps, jars were preserved within a closed polystyrene box placed in a controlled temperature room at 20C in order to avoid any even small temperature change. Each set of samples was subdivided in two groups, one of them followed a drying path, whereas the other one followed a wetting path after a preliminary air-drying stage. Both main drying and main wetting paths were followed allowing each sample to equalize at only one suction value in a free shrinkage condition in drying and in a free swelling condition in wetting. The reproducibility and the reliability of the technique were assessed by comparing results from two or three samples subjected to the same suction value. In general the results differed by a negligible amount, even for the lower suctions. Sample equalization evolution was controlled weekly by monitoring sample weight and volume until a constant mass condition was achieved. Weights were measured with a resolution of 0.0001 grams. Height and diameter of the samples were measured with a resolution of 0.001 mm by using a digital micrometer. Suction values lower than 2 MPa were controlled by means of the axis translation technique (air overpressure method) in two suction controlled oedometers. In the context of a wider research programme devoted to investigate the hydromechanical behaviour of the compacted unsaturated clay (Air Farulla et al. 2007), results related to main drying and main wetting paths were selected in the suction range 0.01 0.8 MPa. 3 ANALYSIS OF RESULTS
Equilibrium conditions of all compacted clay samples tested in main drying and wetting paths by vapour equilibrium technique are reported in the w-e plane (Fig. 1). Curves at equal value of suction interpolating experimental data (least square method) and curves at equal value of degree of saturation are also depicted. Dry density influences the retention properties of the compacted clay in different ways. For higher suction values, equal suction curves are vertical or nearly vertical, indicating that void ratio does not exert any appreciable influence on the retention behaviour. A unique correspondence exists between suction and water content irrespective of the dry density of the compacted clay. For lower
418
Figure 2.
Water retention curves for e 0.45 0.60.
Figure 1. Water content versus void ratio at suction controlled main drying and wetting equilibrated conditions (vapour equilibrium technique).
suctions (water content higher approximately than 0.11) different water contents are measured at constant suction as void ratio varies. The observed behaviour is similar to the one reported by other authors for different compacted clays (e.g. Romero 1999; Tarantino & De Col 2008). Hydraulic hysteresis effects are also evident as different water contents are measured at the same suction for wetting and drying processes. Equal suction curves related to the wetting paths are always shifted to the left compared with the corresponding drying curves, indicating that water contents measured in wetting are lower than corresponding values in drying at the same suction value. These effects, moderate for the higher suctions, became very significant as suction decreases and void ratio increases. Effects of void ratio on the evolution of retention properties can be better appreciated and discussed with reference to the main drying and main wetting retention curves represented in the plane suction-water ratio for e 0.45 0.60 in Fig. 2. Water ratio, ew, is the ratio of water volume to solid volume and it is assumed to represent the amount of water retained in the soil pores because of its similarity to void ratio. Moreover void ratio and water ratio are work-conjugate variables of net stress and suction, respectively (Houlsby 1997; Romero & Vaunat 2000). Main retention curves are derived from experimental data represented in Fig. 1 for total suction values in the range 2 110 MPa (vapour equilibrium technique) and from suction controlled oedometric test results for matric suction in the interval 0.01 0.8 MPa (axis translation technique).
Retention curves in Fig. 2 were obtained by least squares fitting of the experimental data using a modified form of the van Genuchten equation in a similar way as proposed by Romero and Vaunat (2000): ew ewh 1 C (s ) n e ewh 1 ( s ) where s ln 1 a C (s) 1 ln(2 )
m
(1)
(2)
where s represents suction value. The parameters 1/, n and m are related respectively to the air entry value or water entry value, the slope of the curve at inflection point and the residual water ratio. The residual water ratio is equal to the hygroscopic water ratio ewh 0.111, corresponding to the suction value a 110 MPa (Fig. 2). The previous equation refers to isochoric conditions. For each of the selected void ratios (0.45, 0.50, 0.55 and 0.60) experimental data differing 0.025 from the considered value were fitted. Parameters , n and m, determined by a least squares fitting method, are collected in Table 1. Parameter varies about linearly in a wide range, whereas parameters n and m vary in narrow ranges. Calculated parameters , n and m were used to obtain the main retention curves for e values between 0.65 and 0.80, where no data from the axis translation technique were available. To this aim, the parameter was assumed to vary linearly with the void ratio, whereas parameters n and m were assumed to be constant and equal to their average values.
419
Table 1. Parameters n, m and 1/ for main drying and main wetting reention curves. n e 0.45 0.50 0.55 0.60 drying wetting drying wetting drying wetting drying wetting 1.312 0.927 1.400 1.037 1.400 1.037 1.445 1.053 m 0.248 0.400 0.212 0.321 0.204 0.321 0.201 0.318 1/ [MPa] 2.849 1.837 1.538 0.909 0.909 0.606 0.667 0.435
Figure 3.
Water retention curves for e 0.650.80.
The predicted curves, represented in Fig. 3, are matched to data collected from the equal suction curves in Fig 1. A good agreement is observed in the suction ranges where experimental data are available. Analysis of data represented in the diagrams of Fig. 2 and Fig. 3 suggest that the water ratio ew 0.30 (corresponding to w 0.11) could be assumed to subdivide the plots in two regions in which different evolution of main retention curves with void ratio can be observed. For water ratio lower than 0.30 experimental points collect, independently of the void ratio considered, along two linear (semilog plot) envelopes related to drying and wetting paths respectively, stemming from the point at hygroscopic water ratio (ewh) (where drying paths end and wetting paths start). The two linear envelopes are very close but distinct. Suction values corresponding to ew 0.30 are about 10 and 6 MPa, respectively, for the main drying and the main wetting retention curves. For water ratio values greater than 0.30, as void ratio increases main drying and main wetting
retention curves move to the right and depart each other. The dominium of attainable states delimited by the main drying and wetting retention curves expands, as consequence of the greater influence of hysteresis. The observed behaviour can be interpreted by considering the double porosity features of the material structure at microscopic level and the different water retention mechanisms in micropores (i.e. pores within particle aggregates) and in macropores (i.e. pores between individual particle aggregates or assemblages) (Romero & Vaunat 2000). For ew 0.30 (w 0.11) it can be supposed that water is present in the micropores and macropores are completely dry, as void ratio change (due mainly to a reduction of macropore volume) does not induce any variation of retention properties. Water adsorption by clay particles inside aggregates becomes the dominant mechanism in the retention behaviour. Adsorption mechanism depend on mineralogical clay properties as specific surface and plasticity. This result has been validated on the basis of a MIP test on a freeze-dried compacted scaly clay sample. Assuming an entrance pore size of approximately 2 m as a schematic separation between micro and macropores, a microstructural void ratio (cumulated intrusion void ratio in the microporosity region) em0 0.30 was obtained (Ferrari et al. 2009). Of course, em0 and ewm are coincident as micropores are considered saturated. For ew 0.30 (w 0.11) water content is high enough to partly fill the inter-aggregate voids and to form menisci at aggregate contacts. The dominant mechanism of water storage in this region is capillarity which is dependent on void ratio and mechanical actions which modify macropore volume. This interpretation can be supported on the basis of data represented in the diagram of Fig. 4 showing evolution versus time of water ratio and void ratio of a compacted sample in a free shrinkage air drying test, during which sample weight and volume were monitored. Soon after air exposition, sample dries rapidly and shrinks greatly until a water content of about 0.11, corresponding to ew 0.30 is achieved. Water loss and shrinkage volumetric strain in the successive phase (ew 0.30) evolve at greatly reduced rate, and furthermore void ratio reduction are also very low. The drying process in the first phase takes place in a few hours, while many days are necessary for the successive loss of water and volume shrinkage. Different mechanism evidently operate for ew greater or lower than 0.30 in determining both water loss (evaporation) rate and intensity of corresponding volumetric strains.
420
Figure 4. Water ratio and void ratio evolution in a free shrinkage air-drying test.
Figure 5. Void ratio variation versus water ratio in free swelling at different suction values.
For ew 0.30 water fill macrovoids, as bulk water and water in the menisci at particle contacts in the air filled voids. Because of the difference in relative humidity between pore voids and environment, water can leave pore voids rapidly as consequence of both high gradient in pore void and air suction difference and the relative great pore sizes. As suction in the soil increases volumetric strains develop as effective stresses in the bulk water zone and curvature of menisci in the air filled voids increase (Wheeler et al. 2003). For ew 0.30 macrovoids are empty while microvoids are still saturated. Water evaporation at increasing suction is now contrasted by adsorption attractive bonds and slows significantly. Moreover volumetric strains as consequence of aggregate shrinkage are of very moderate intensity. Increments of void ratio (swelling strains are considered negative) versus water ratio measured along wetting paths on samples of the compacted clay are represented in Fig 5. Also in this case, as a first approximation, a bilinear evolution of the rate of the wetting processes can be detected. On the basis of the collected data, the water ratio value ew 0.30 can be assumed with confidence in fixing the boundary of the region ruled by the discussed different retention mechanisms. Void ratio effects on Air Entry Value (AEV) and Water Entry Value (WEV) are highlighted in the diagram of Fig. 6, in which collected results were recovered from the retention curves of Fig. 2 and
Figure 6. Air entry value and water entry value at varying void ratio.
Fig. 3 at Sr 0.95. Experimental points were fitted to curves with equations: AEV 0.0574e3.467 [MPa] WEV 0.0176e3.468 [MPa]
2
(3) (4)
with R equal to 0.986 and 0.992, respectively. In order to characterize the hydraulic hysteresis loop an adimensional hysteresis index, HI, is defined as: HI 1 Areawetting /Areadrying (5)
where Areawetting and Areadrying represent the area subtended by main wetting curve and main drying
421
Equations to evaluate AEV and WEV values (recovered from main retention curves at Sr 0.95) are also presented. On the basis of the collected experimental data, retention properties of the compacted clay can be considered known, or at least reliably estimated, for a very wide interval of void ratio values. These results can contribute to elaborate a model of the coupled hydro-mechanical behaviour of the tested material. REFERENCES
Air Farulla, C. & Jommi, C. 2005. Suction controlled wetting-drying cycles on a compacted scaly clay. In Proceedings of International Conference on Problematic Soils. Edited by H. Bilsen and Z. Nalbatoglu. Eastern Mediterranean University, Famagusta, N. Cyprus. 2527 May 2005, Vol. 1, pp. 229238. Air Farulla, C., Ferrari, A. & Romero, E. 2007. Mechanical behaviour of compacted scaly clay during cyclic controlled-suction testing. In Experimental Unsaturated Soil Mechanics. Edited by T. Schanz. Springer Proceedings in Physics 112, pp. 345354. Birle, E., Heyer, D. & Vogt, N. 2008. Influence of the initial water content and dry density on the soilwater retention curve and the shrinkage behavior of a compacted clay. Acta Geotechnica 3: 191200. Ferrari, A., Air Farulla, C. & Romero, E. 2009. On the volumetric response of a compacted clay subjected to wetting and drying cycles. 4th Asia-Pacific Conference on Unsaturated Soils, 2325 November 2009, Newcastle, Australia. Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics for unsaturated soils. J. Wiley, New York. Houlsby, G.T. 1997. The work input to an unsaturated granular material. Geotechnique, 47(1): 193196. Jommi, C. & Di Prisco, C. 1994. Un semplice approccio teorico per la modellazione del comportamento meccanico dei terreni granulari parzialmente saturi (in Italian). Atti del Convegno sul Tema: Il Ruolo dei Fluidi nei Problemi di Ingegneria Geotecnica, Mondov, 167188. Nuth, M. & Laloui, L. 2008. Effective stress concept in unsaturated soils: Clarification and validation of a unified framework. International Journal for Numerical and Analytical Methods in Geomechanics. Romero, E. 1999. Characterisation and thermo-hydromechanical behaviour of unsaturated Boom clay: an experimental study. Ph.D. thesis, Universitat Politecnica de Catalunya, Barcelona, Spain. Romero, E. 2001. Controlled-suction technique. In Proceedings of the 4 Simpsio Brasileiro de Solos No Saturados. Edited by W.Y.Y. Gehling and F. Schnaid. Porto Alegre, Brasil, pp. 535542. Romero, E. & Vaunat, J. 2000. Retention curves of deformable clays. Experimental Evidence and Theoretical Approaches in Unsaturated Soils, Trento, pp. 91106. Tarantino, A. & De Col, E. 2008. Compaction behavior of clay. Geotechnique 58(3): 199213. Wheeler, S.J., Sharma, R.S. & Buisson, S.R. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Geotechnique 53(1): 4154.
Figure 7.
Hysteresis index evolution with void ratio.
curve respectively. The evolution of the hysteresis index with void ratio is reported in Fig. 7 together with the fitting straight line (R2 0.997): HI 0.0791e 0.0575 (6)
Calculated data evidence a linear dependence of HI versus e, for the range of void ratio values considered. It is also evident that plotted points related to the fitted (Fig. 2) and modelled (Fig. 3) retention curves match very well. 4 CONCLUSIONS
Collected results point out a strong dependence on void ratio of the retention properties of the compacted clay. An interpretation of the different mechanisms which regulate the water content changes in the soil and corresponding volumetric strains is provided on the basis of the microstructure characteristics of the material. For ew ewm water is inside microvoids, held by adsorption bonds to the particle surface. Drying or wetting processes evolve at very low rate and corresponding volumetric strains are of moderate intensity. For ew ewm water is into macropores and capillarity is the dominant retention mechanism. Changes of water volume as suction varies are relatively fast and corresponding volumetric strains attain greater values. Main retention curves in wetting and drying are represented by a modified form of the van Genuchten equation. Parameter recovered by fitting experimental points in the e 0.45 0.60 range are used to model retention main wetting and drying curves for variable e values enabling to represent retention properties in a very wide void ratio range: e 0.45 0.80.
422
Comparison between the in situ and laboratory water retention curves for a silty sand
A. Askarinejad, F. Casini, P. Kienzler & S.M. Springman
Institute for Geotechnical Engineering, ETH Zrich, Switzerland
ABSTRACT: After an extreme rainfall event in May 2002 a series of landslides occurred in Ruedlingen in Canton Schaffhausen, North Switzerland. A 38 steep slope has been chosen in this area beside the river Rhine to carry out an artificial rainfall experiment to investigate the dependence between rainfall, suction, saturation and shear resistance. Two sprinkling experiments were conducted to represent an extreme rainfall event, the second of which resulted in failure of 130 m3 of the slope. Several cycles of wetting and drying were applied to the soil and suction and volumetric water content were measured at different depths in three locations of the slope, by which in-situ Water Retention Curves (WRC) can be derived. The WRC of an undisturbed sample was also determined from laboratory test. The in situ and laboratory WRCs are compared in this paper and the differences will be discussed. 1 INTRODUCTION 2 TEST FIELD
The stability of slopes depends directly on the distribution of pore pressures inside its mass. This distribution can be strongly influenced by rainfall due to the infiltration of water into the slopes (Alonso et al. 1995). The description, quantification and study of soil water flow processes leads to a consideration of the hydraulic properties of unsaturated soils. This involves the relationship between the soil water content, the hydraulic conductivity, and the soil water retention function, which relates the soil water content to the capillary suction (Braddock et al. 2001). Soil water retention can be determined both in the laboratory and in the field. In the former, both values of suction and water content are measured in the same sample, while in the field the measurements are done at two different points and over different soil volumes. The differences between insitu and laboratory WRCs have been investigated and attributed to inadequate representation of large pores in laboratory, scale effects in the sample, disturbance, the spatial variability of the soil, hysteresis and overburden pressure (Field et al. 1985; Shuh et al. 1988). Several cycles of wetting and drying were applied to the soil, suction and volumetric water content were measured at different depths in three locations of the slope, by which in-situ Water Retention Curves (WRC) of the soil were derived. The WRC of undisturbed sample was also determined from laboratory tests and will be compared with the insitu results in this paper.
The selected experimental area is a steep, forested slope in Ruedlingen, North Switzerland, where 42 surficial landslides occurred (Fischer et al. 2003) in spring 2002, after an extreme rainfall (100 mm fell in 40 minutes). To investigate the hydrological and mechanical behaviour of the slope, a sprinkling experiment was carried out in autumn 2008. This test was followed by a triggering experiment in spring 2009. This paper is based on the data from the first test. 2.1 Geometry
The selected experimental area is located on an east facing slope on the banks of the river Rhine (Springman et al. 2009). The altitude is about 350 masl. The average gradient of the slope was determined using a total station theodolite to be 38. During the preliminary investigations, sandstone and marlstone were located at a depth of between 0.5 m to more than 5 m along the 30 m 7.5 m plan section. 2.2 Infiltration characteristics
Infiltration characteristics were evaluated with combined sprinkling and dye tracer tests. The tests were replicated across three different locations near to the investigated area (TP 3, TP 4, TP 5; Figure 1). The experiments revealed high infiltration capacity of the soil. No overland flow was observed during the sprinkling at any of the three locations. After
423
is typical for many rainfall-induced landslides (Springman et al. 2003). However, stained fractures below the subsoil revealed that substantial drainage might occur into the bedrock, which may prevent complete saturation and failure of the instrumented slope (Springman et al. 2009). 2.3 Laboratory classification
The behaviour of soil was investigated under saturated and unsaturated conditions (Springman et al. 2009, Casini et al. 2010a & b). The average saturated permeability of soil was determined to be ksat 1 * 107 m/s. The grain-size distributions at different depths of TP2 are shown in Figure 3. The silt and clay fractions increase with depth, while the sand fraction decreases. The soil can be classified as MediumLow plasticity silty sand (ML) according to USCS. The soil Water Retention Curve (WRC), which links suction s ua uw and volumetric water content w Vw /V for a natural sample is presented in Figure 4. The WRC was obtained under suction
Figure 1. Test site plan showing dimensions and position of Test Pits (TP).
Figure 3. of TP2.
The grain size distribution at different depths
Figure 2.
Dye pattern in the middle section of TP5.
excavation, flow paths of the infiltrated water were clearly visible by the dye patterns (Figure 2). In summary, the dye patterns showed complete and homogeneous saturation of the whole soil profile with only little preferential drainage and with perched saturation above the bedrock. With regard to the planned landslide triggering experiment, these results confirmed the high vulnerability of the slope to failure and led to expectations of a more deep-seated failure above the transition to bedrock rather than a more surficial failure that
Figure 4. Water retention curve of an undisturbed sample z 120 cm (TP2).
424
controlled conditions with a Fredlund Apparatus (Perez Garcia et al. 2008). As the first step, a suction of s 70 kPa was imposed to the sample and some water was expressed from the soil. Therefore, the suction in the natural sample was less than 70 kPa. In the continuation step, the sample absorbed some water, following a scanning curve until reaching the main wetting and the main drying during desaturation (Figure 4). The experimental points were fitted with the Van Genuchten (VG) model (1980): w wres 1 wsat wres 1 ( s )n
m
where , n, m are parameters while 0, m 0, n 1 and m 1 1/n. wsat and wres are the saturated and residual volumetric water content, respectively. The values for the wetting and drying branch are reported in Table 1. The air entry value is saev 2.7 kPa for the drying curve, which is consistent with the WRC expected for sandy soils (Lu and Likos, 2004). 2.4 Instrumentation An extensive instrumentation plan was designed to measure hydrological and geo-mechanical responses of the slope. Detailed measurements of soil suction, water level and soil volumetric water content were measured combined with an investigation of subsurface flow at the lowest part of the slope by means of tracer experiments. Deformations were monitored during the experiment, both on the surface via photogrammetrical methods and within the soil mass, using a flexible probe equipped with strain gauges at different points and two axis inclinometers on the top (Askarinejad 2009). The instruments were installed mainly in three clusters over the slope. The instruments included jet-fill tensiometers, TDRs, Decagons, piezometers, soil temperature sensor, deformation probes, earth pressure cells, acoustic sensors and rain gauges (Figure 5). The tensiometers were installed at depths of 15, 30, 60, 90, 120, and 150 cm below the ground surface in each cluster. Decagons were installed at shallow depths of
Table 1. van Genuchten parameters for Ruedlingen silty sand. wsat [-] Wetting Drying 0.44 0.44 wres [-] 0.20 0.25 [kPa-1] 0.90 0.37 N [-] 1.53 1.72
Figure 5.
Instrumentation plan and section.
15 to 60 cm every 15 cm, and TDRs from 60 cm to 150 cm, with aspacing of 30 cm. All the instruments were calibrated and checked in the laboratory for proper functioning before installation in the field. The hydrological responses of the soil were measured during the experiment with logging interval of 5 minutes. 3 3.1 EXPERIMENTAL PROCEDURE AND INSITU MEASUREMENTS Applied rainfall
The artificial rainfall was applied by means of 10 sprinklers with the same spacing located on the middle longitudinal line of the field. The lower sprinklers experienced higher hydraulic heads as the water was supplied from the upper part of the slope. Therefore, the rainfall was not uniformly distributed. The average applied rain intensity is shown in Figure 6 a. As indicated in this figure the rain was applied with an average intensity of 35 mm/hr for 3 hours as the first wetting phase (W1), followed by a 20 hour stop allowing the soil to drain (first drying phaseD1). Afterwards, the slope was sprinkled for 1.5 days with an average intensity of 17 mm/hr, which increased with a first shock of 45 mm/hr to average value of 30 mm/hr for another 1.5 days (second wetting phaseW2) and then the second drying phase started (D2).
425
Rain intensity [mm/hr]
0 10 20 30 40 50 6 5 4 3 2 1 0 5.5 5 4.5 0.3
D1
D2 W
Suction VWC 0.5 Volumetric water content [-]
D2
Suction [kPa]
0.4
0.2 0.5
D2
0.4
4 3.5 3 0.2 2.5 2 8 6 0.4 4 2 0 0.2 -2 0.3
Volumetric water content [-]
Suction [kPa]
measurements respond without significant time discrepancy in all 3 depths. This verifies the high infiltration capacity of the soil obtained from the earlier hydrological investigations. At some points during the second wetting phase of the rainfall, the changes in VWC and suction are not contemporaneous. This can be due to either the hysteretic effect of the soilwater retention curve or the difference in the location of the sensors at each depth. In figures 6 b and c, at the beginning of the second drying phase (D2), the VWC drops as the suction increases, but in cluster 2 at depth of 120 cm (Figure 6 d), the suction increases up to 2.5 kPa and only after this point the VWC starts to decrease. This point can be described as the air entry value of the WRC which is consistent with the value derived from the laboratory test (saev 2.7 kPa). Figures 7 to 9 show the insitu WRC derived from Figures 6 b, c, & d respectively, compared with laboratory curve. The first wetting branch (W1) shows higher values of suction at the same VWC compared with first drying curve (D1), which is in contrast with the
6 5
Suction [kPa]
insitu Wetting 1 Insitu Drying 1 insitu Wetting 2 Insitu Drying 2 Lab WRC
Drying Lab.
0.1 0.5 Volumetric water content [-]
Wetting Lab.
3 2 1 0 0.3 0.32 0.34 0.36
D2 W2 D1
W1
Suction [kPa]
D2
0.3
0.38
0.4
0.42
0.44
VWC [-]
Figure 7. Insitu water retention curve of cluster 1 at depth of 60 cm compared with laboratory results.
6
Wetting lab.
D2
d
-4 10/31/08 10/27/08 10/28/08 10/29/08 10/30/08 11/1/08 0.1
insitu wetting 1
Drying lab.
insitu drying 1 insitu wetting 2 insitu drying 2 Lab WRC
W1
Suction [kPa]
Figure 6. (a) Average applied rain intensity, (b) Variations of volumetric water content and matric suction in cluster 1 at depth of 60 cm, (c) cluster 2 at depth of 30 cm, and (d) cluster 2 at depth of 120 cm.
W2
3
D1
3.2
Insitu water retention curves
Figures 6 b, c, and d show the variations of Volumetric Water Content (VWC) and matric suction at three different depths of 30, 60, and 120 cm in clusters 1 and 2. After changes in the intensity of applied rainfall, the suction and volumetric water content
0 0.2 0.25 0.3 0.35 VWC [-] 0.4 0.45 0.5
426
6 5 4
Suction [kPa]
D2
Drying lab.
W2
W1
insitu wetting 1 insitu drying 1 insitu wetting 2 insitu drying 2 Lab WRC
3 2
D1 Wetting lab.
1 0 0.3 0.35 0.4

VWC [-]
0.45
0.5
Figure 10. Drying curve of insitu WRC for cluster 1 compared with laboratory results.
unsaturated-soils hysteresis theory (Mualem 1973). There are two hypotheses for this observation. i. As the soil is loose and the overburden pressures are low, therefore, the volume of the soil could increase during the wetting due to reduction in effective stress (swelling), and on the other hand during the drying period, the soil volume decreases because of the build up of suction and increase in effective stresses. ii. As the soil is granular and its response to the applied rainfall is relatively fast (medium to high permeability), small heterogeneity in the flow paths from the surface to the tensiometers and TDRs, like small channels due to dead roots, can affect the response time. Therefore, especially during the intense wetting phases, the readings of sensors at each time are maybe not exposed to the same conditions. Also, it should be noted that tensiometers measure suction in a thin soil layer around a ceramic cup of 7 cm length and a diameter of 2.2 cm, whereas a TDR measures soil volumetric water content in a cylindrical soil volume with a diameter and length of 5 cm and 15 cm, respectively. There is also a possibility that the branches of W1, D1, and W2 in the insitu WRCs are not the main drying curves but are scanning curves. Accordingly, in order to be sure that the changes in suction and volumetric water content are on a main branch of the curve, the last drying part of the curves (D2) of clusters 1 and 2 at different depths has been compared with the drying branch of the curve obtained from the laboratory test. The results are presented in Figures 10 & 11. In Figure 10, the laboratory curve is in quite good agreement with the insitu measurements. In cluster 1, the curves related to the depths of 120 cm and 90 cm show air entry values of around 2.2 kPa. This value is lower than the one implied by the laboratory WRC, which can be related to
Figure 11. Drying branch of insitu WRC for cluster 2 compared with laboratory results.
the void ratio difference between the soil at these depths and the void ratio of the natural sample tested in the laboratory (e 0.8). The insitu void ratio of the soil can be calculated from the volumetric water content under saturated conditions and using Equations 1 and 2.
Sr .n, Sr 1 s 1 ns
e n 1 n
(1) (2)
The void ratios calculated from this approach are 1.04 and 0.96 at depths of 90 cm and 120 cm, respectively. Similarly, in cluster 2 (Figure 11) at a depth of 120 cm, the air entry value is marginally higher than the one at 150 cm, while the soil has a lower void ratio at 120 cm compared with the one at 150 cm. The data from figures 10 and 11 suggest that the trend lines of the insitu drying branches are less steep than the line from the laboratory results.
427
This observation can be attributed to the fact that the laboratory curve has been derived from a one dimensional drainage path, while the insitu condition is a three-dimensional situation and the water can flow in different directions rather than only vertically, also it can be due to the different porosity of insitu and laboratory test. Therefore, the differential water capacity defined as d /ds is higher from field data than from laboratory data in the range of low soil water matric potentials. This should result in slower infiltration predictions based on the sorptivity from the laboratory data. These and similar observations indicate the need to pay more attention to the effects of scale on soil hydraulic properties (Pachepsky et al. 2001). 4 FURTHER TEST PLANS
ACNOWLEDGEMENTS This research was funded by the Competence Centre for Environment and Sustainability (CCES) within the framework of the TRAMMProject. Marco Sperl and Ernst Bleiker helped with the fieldwork. We are grateful to the Gemeinde of Ruedlingen and their President Mrs. Katy Leutenegger for giving permission to carry out this experiment on their land.
REFERENCES
Alonso, E. Gens, A. & Lloret, A. 1995. Effect of rain infiltration on the stability of slopes. In Proc. of the 1st Intern. Conf. on Unsaturated Soils, Paris: 241249. Rotterdam: Balkema. Askarinejad A. 2009. A method to locate the slip surface and measuring subsurface deformations in slopes. 4th Intern. Young Geotechnical Engineers Conf., Alexandria, Egypt: 171174. Braddock, R.D. Parlange, J.-Y. & Lee, J. 2001. Application of a soil water hysteresis model to simple water retention curves. Transport in Porous Media, 44: 407420. Casini, F. Jommi, C. Springman, S.M. 2010a. A laboratory investigation on an undisturbed silty sand from a slope prone to landsliding. Granular Matter 12 (DOI: 10.1007/s10035-010-0182-y). Casini, F., Minder, P., Springman, S.M., 2010b. Shear Strength of an unsaturated silty sand (this issue). Field, J.A. Parker, J.C. & Powell, N.L. 1985. Comparison of field- and laboratory measured and predicted hydraulic properties of soil with macropores. Soil Sci. 138: 385396. Fischer, C. Lpez, J. Springman, S.M. 2003. Remediation of an eroded steep slope in weathered sandstone after a major rainstorm. Intern. Conf. on Landslides, Hong Kong: 878883. Lu, N. Likos, W. 2004. Unsaturated soil mechanics, Wiley, New Jersey. Mualem, Y. 1973. Modified approach to capillary hysteresis based on a similarity hypothesis, Water Resour. Res., 9: 13241331. Pachepsky, Y. Rawls, W.J. & Gimnez, D. 2001. Comparison of soil water retention at field and laboratory scales. Soil Sci. Soc. Am. J. 65: 460462. Perez-Garcia, N. Houston, S.L. Houston, W.N. Padilla, J.M. 2008. An oedometertype Pressure Plate SWCC Apparatus. Geotechnical Testing Journal 31: 19. Springman, S.M. Jommi, C. Teysseire P. 2003. Instabilities on moraine slopes induced by loss of suction: a case history. Gotechnique, 53 (1): 310. Springman, S.M. Kienzler, P. Casini, F. Asakrinejad, A. 2009. Landslide triggering experiment in a steep forested slope in Switzerland. 17th Intern. Conf. of Soil Mech. & Geot. Eng. Alexandria, Egypt: 16981701. Shuh, W.M. Cline, R.D. & Sweeney, M.D. 1988. Comparison of a laboratory procedure and a textural model for predicting in situ water retention. Soil Sci. Soc. Am. J. 52:12181227.
In order to investigate the effect of cycles of wetting and drying on the void ratio of the soil, a series of tests will be performed on undisturbed samples from different depths of the slope profile, loaded to the same overburden pressures as in the field. Also, more laboratory WRC tests will be done on undisturbed samples from different depths with higher resolution at higher VWCs and suctions lower than 15 kPa. 5 CONCLUSION
Two cycles of wetting and drying has been applied on a natural, steep, forested slope in northern Switzerland. At 3 different points of the slope, and at 6 depths, the values of volumetric water content and suction have been measured. Partial agreement was observed comparing the insitu values with a laboratory derived WRC on an undisturbed sample. In some of the insitu WRCs, it was observed that at the same water content, the wetting branch shows higher values of suction than the drying branch. This can be due to the change in void ratio of the soil during wetting and drying, to the heterogeneity of the soil along the flow path from the surface to the measurement depth of different sensors, and the difference in response time of different types of devices. The hypothesis of volume change and effect of initial porosity of the soil on the WRC will be investigated by testing samples with different void ratios. The trend lines of the drying branches of insitu WRCs are less steep than the line from laboratory results. This observation is attributed to the fact that the laboratory curve has been derived from a one dimensional drainage path, while three-dimensional flow is possible under situation insitu condition or also this can be related to the effect of sampling and needs to be verified by further laboratory tests.
428
On the inverse problem solution for laboratory determination of unsaturated water hydraulic conductivities
K.V. Bicalho
Federal University of Espirito Santo, ES, Brazil
J.-M. Fleureau Y.-J. Cui
Ecole Centrale Paris & CNRS, Chtenay-Malabry, France Ecole des Ponts ParisTech, UR Navier-CERMES, France
C. Romanel
Pontifical Catholic University of Rio de Janeiro, Brazil
ABSTRACT: Soil hydraulic properties are essential in describing unsaturated flow in natural or compacted soils that are often used to construct liners and covers for waste containment. This paper provides an overview of the use of an approach called inverse problem solution for laboratory determination of unsaturated water hydraulic conductivity. In this approach the experimental data are viewed as the solution of the boundary value problem for which the governing equation and the initial and boundary conditions are known, but the material functions and the parameters in the governing equations are unknown. Since the inverse solution problems are usually mathematically ill posed, the quality of the experimental data is essential for the success of using this approach. 1 INTRODUCTION specific restrictions exist with respect to sample size; no linearization of the flow equation is required; and one has the freedom to select any desired functional model for the hydraulic parameters. Nonetheless, the identifiability or reliability, uniqueness and stability of the results have been still questioned. This paper provides an overview of the use of IPS techniques for laboratory determination of unsaturated water hydraulic conductivities. Details as to the procedures for conducting the laboratory tests are not presented. They are given in the various papers cited in the references. The focus is on the inherent advantages and shortcomings of IPS techniques as well as the examination of the underlying assumptions. Since unsaturated water flow analysis has numerous geophysical and geotechnical applications, the paper brings some soil physics and vadose zone hydrology literature into the field of civil and geotechnical engineering. 2 THEORY
An accurate soil physical description of the combined Soil Water Retention Curves (SWRC) and unsaturated hydraulic conductivity (kw) functions over a wide range of soil water contents is essential prerequisite for describing unsaturated flow in natural or compacted soils that are often used to construct liners and covers for waste containment. Determining kw in a direct fashion has been generally found to be difficult in laboratory and an alternative to direct estimates is the use of inverse problem solutions (Zachman et al. 1982, Durner et al. 1999, Hopmans et al. 2002). In this approach the experimental data are viewed as the solution of the boundary value problem for which the governing equation and the initial and boundary conditions are known, but the material functions and the parameters in the governing equations are unknown. Advances in numerical modeling and increased computing capacity now allow numerical inversions of saturated-unsaturated flow problems on a common personal computer. Inverse Problem Solution (IPS) techniques are interesting because retention and conductivity curve parameters can be simultaneously derived from a single experiment; no
Inverse problem theory is the mathematical theory describing how information about a parameterized
429
physical system can be derived from observational data, theoretical relationships between model parameters and data, and prior information. Solving the inverse problem is usually defined as the estimation of parameters by calibrating a model against observed data. The application of the inverse approach to the modeling of hydraulic soil parameters consists in using hydraulic models and optimization algorithms for determining material hydraulic characteristics by inverse simulation of transient outflow/ inflow experiments. This technique has been common practice in the field of groundwater modeling over years as presented by Yeh (1986), and it is now routinely used in the field of unsaturated water flow (Wildenschild et al. 1997, Simunek et al. 1998, Hopmans & Simunek 1999, Romano & Santini 1999, Bicalho 1999, Hopmans et al. 2002, Znidarcic et al. 2004). It has been also used in more complicated situations involving unsaturated swelling soils (Masrouri et al. 2008). By using the IPS approach it is also possible to estimate the material hydraulic characteristics in the positive pore water pressure range where the pore air phase is discontinuous (Bicalho 1999). In the inverse modeling of a transient flow experiment, for example, soil hydraulic properties are assumed to be described by a mathematical model with yet unknown parameter values. A number of mathematical functions have been suggested for describing the soil hydraulic properties (Dane & Topp 2002). Abdallah & Masrouri (2002) state that the hydraulic conductivity (kw) seems to be the most sensitive parameter of a model simulating the transient phase during in situ single ring permeability tests since small variation of kw produce important changes of the objective functions. The quality of the experimental data is essential for the success of using the inverse approach. An experiment is setup under controlled conditions with prescribed initial and boundary conditions. During the experiment one or more flow controlled variables are measured. Subsequently, the governing equation (i.e., Richards equation) is solved numerically using the parameterized hydraulic functions with initial estimates for their parameters. These parameters are optimized by minimization of an objective function containing the sums of squared deviations between observed and predicted flow variables, using repeated numerical simulation of the flow process. This iterative inversion of the flow equation is in contrast to direct inversion techniques as used in analytical solutions (Yeh 1986). The IPS approach has been applied in the laboratory using outflow experiments such as the one-step (Kool et al. 1985, van Dam et al. 1992) and multi-step outflow approach (van Dam et al. 1994,
Zurmhl & Durner 1998) where the measured outflow and soil water (in some cases) were used as fitting criteria in estimating the soil hydraulic parameters. A classical transient flow experiment for determining material hydraulic characteristics is the one-step method, where a stepwise change in the water potential is imposed on one boundary of a soil column. A problem of the one-step method is that the quick change of the boundary condition does not represent natural conditions and may lead to nonequilibrium in the soil sample. Also, contact problems at the soil- porous plate interface may occur as a result of a large pressure drop, and the sensitivity of the method can be low if, as a result of a large pressure drop, a very thin drained soil layer next to the porous plate controls the total flow rate (Durner et al. 1999). McLaughlin & Townley (1996) presented a review of the available numerical methods which have been used for IPS. Several kw estimations have been done using data obtained with the evaporation method (Santini et al.1995, Ciollaro & Romano 1995, Simunek et al. 1998, Peters & Durner 2008). These laboratory methods provide information about the drying paths of the SWRC. Fleureau & Taibi (1995) pointed out that the hysteresis between the drying and wetting paths is very important, both for water and air hydraulic conductivities. Therefore, the effect of hysteresis on the kw functions in a large degree of saturation domain remains uncertain, and an additional scaling factor may be required for relating the soil parameters derived from the drying curve to that corresponding to the wetting curve. Bicalho (1999) demonstrated the feasibility of the IPS approach to obtaining the hydraulic characteristics for unsaturated soils. The transient water flow measurements were accomplished by using a standard triaxial or flexible wall permeameter cell. A flow pump was used to control the water flow in and out of the sample and the induced pressure difference was measured by a precision transducer. A PC based data acquisition system was used to continuously record the test progress. A numerical solution for the Richards equation for one-dimensional transient water flow was used to numerically simulate the experiment. The hydraulic conductivity function was adjusted until a good agreement between the experiment and the analyses was obtained. The comparison between the numerical model and the experimental data is presented in Figure 1. The results show that good agreements between the predicted and measured curves were achieved by using the hydraulic constitutive functions proposed for this soil sample. The major disadvantage of laboratory testing techniques is the lack of representativeness of the tested soil samples due to a possible
430
Figure 1. Comparison of the numerical results with the measurements.
scale-dependency of hydraulic conductivity. One should also be aware that undisturbed soil samples are very difficult to obtain and remolded soil samples have been often used for evaluating the kw functions. Moreover, significant difference exists in the soil microstructure and macrostructure between different soil samples. For these reasons the laboratory experiments performed to measure kw functions may not be applied to field situations. In this regard, IPS techniques are interesting because they can be applied to field experiments; the retention and conductivity curve parameters can be derived from one single experiment. Advantages of the IPS approach include: potential time and cost savings since steady state flow conditions are not required; greater flexibility in the design of experiments as no specific boundary and initial conditions are required for the solution; the lack of inherent constraints on the form or complexity of the model selected to represent the hydraulic characteristics of the soil (Gribb et al. 2004).
UNIQUENESS AND STABILITY ANALYSIS
minima or by the same global minimum at more than one point in the parameter space (Carrera & Neuman, 1986). The instability problem is caused by the high sensitivity of the parameters and the non-uniqueness problem, on the other hand, occurs when the parameters under study have low sensitivity to the criteria being investigated. This could happen in the solution because at some point in the search space, the parameter(s) may not be sensitive any longer. This problem is also caused by local optima and is a threat to gradient-dependent search algorithms. Kool et al. (1985) applied the inverse approach by numerical solution of the Richards equation for the transient one-step outflow process. They concluded that uniqueness problems are minimized if the experiment is designed to cover a wide range in water content. Moreover, they also determined that initial parameter values must be reasonably close to their true values, and that outflow measurement errors must be small. Prior information about the parameters is an effective method to solve the problem of nonuniqueness. For example, independent measurements of the saturated hydraulic conductivity, the water content and pressure head can be added to measurement of transient cumulative flow rate for inverse parameter identification (Toorman et al. 1992). Sensitivity to errors and the solution uniqueness of inverse methods with one-step tests have been further investigated (e.g., Parker et al. 1985, van Dam et al., 1992). Parker et al. (1985) experimentally applied the one-step method to four different soils of distinct texture, and indicated that the inverse problem for a one-step outflow experiment is non-unique when the number of unknown parameters is more than two and only outflow volume is measured. It was also found that the optimized soil hydraulic functions could be extrapolated to a water content range beyond that achieved with the single pressure step by including an independently measured point of the soil water retention curve in the objective function. 4 CONCLUSIONS
Whereas it appears easy to judge the suitability of a constitutive function when measured data points are fitted, the problem is more subtle when an inverse approach is used. In general, most of the literature about the inverse approach always emphasized the ill-posed nature of the mathematical solution. This is caused mainly by the nonuniqueness problem and instability in the solution. Non-uniqueness leads to more than one set of parameters, yielding minimum values for the objective function (deviations between measurements and simulations) determined by local
The high variability and the large number of factors influencing the kw for different types of soils poses a problem for engineers to predict kw functions. An alternative to direct estimates of soil hydraulic properties is the use of IPS which is attractive because the approach can be applied to field experiments and both retention and conductivity curve parameters can be derived from a single experiment. Nonetheless, the identifiability (reliability), uniqueness and stability of the results are often questioned. Prior information about the
431
parameters (i.e., saturated hydraulic conductivity and water content or suction variations during the unsaturated flow process) is an effective method to solve the problem of non-uniqueness. The quality of the experimental data is essential for the success of using the inverse approach. Therefore, a considerable part of research on inverse modeling of transient flow experiments should therefore focus on determination and evaluation of the experimental data of the tested soil samples.
REFERENCES
Abdallah, A. & Masrouri, F. 2002. Numerical simulation of infiltration in situ tests: applications to unsaturated hydraulic parameters identification. Unsaturated Soils, Balkema, 1, 4550. Bicalho, K.V. 1999. Modeling Water Flow in an Unsaturated Compacted Soil. Ph.D. dissertation, CUBoulder, Co. Carrera, J. & Neumann, S.P. 1986. Estimation of aquifer parameters under transient and steady state conditions. 2. Uniqueness, stability, and solution algorithms. Water Resources Research, 22, 211227. Ciollaro, G. & Romano, N. 1995. Spatial variability of the soil hydraulic properties of a volcanic soil, Geoderma, 65, 263282. Dane, J.H. & Topp, G.C. 2002. Methods of Soil Analysis: Part 4, Physical Methods. SSSA Book Series No 5. Soil Science Society of America, Madison, WI. Durner, W., Schultze, B. & Zurml, T. 1999. State-ofthe-art in inverse modeling of inflow/outflow experiments, Proc. Int. Workshop on Charac. and Measurement of the Hydraulic Properties of Unsaturated Porous Media, ed. van Genuchten, Leij, and Wu, CU-California, CA, 661681. Fleureau, J.M. & Taibi, S. 1995. Water-air permeabilities of unsaturated soils. First Int. Conf. Unsaturated Soils, Paris. Gribb, M.M., Kodesova, R. & Ordway, S.E. 2004. Comparison of soil hydraulic property measurement methods. J. Geotech. Geoenviron. Eng., 130:10, 10841095. Hopmans, J.W. & Simunek, J. 1999. Review of inverse estimation of soil hydraulic properties, Proc. Int. Workshop on Characterization and Measurement of the Hydraulic Properties of Unsaturated Porous Media, ed. van Genuchten, Leij, and Wu, CU-California, CA. 643659. Hopmans, J.W., Simunek J., Romano, N. & Durner, W. 2002. Inverse methods, in Methods of Soil Analysis, Part 4, Physical Methods, ed. J. Dane and G. Topp, SSSA Book Series N.5, Madison, WI, USA. 9631008. Kool, J.B., Parker, J.C. & van Genuchten, M.Th. 1985. Determining soil hydraulic properties from one-step outflow experiments by parameter estimation: I. Theory and numerical studies. Soil Sci. Soc. Amer. J., 49, 13481354.
McLaughlin, D. & Townley, LR. 1996. A reassessment of the ground-water inverse problem. Water Resour Res. 32(5):11331161 Masrouri, F., Bicalho, K.V. & Kawai, K. 2008. Laboratory Hydraulic Testing in Unsaturated Soils. Geotech Geol Eng, DOI 10.1007/s10706-008-9202-7. Parker, J.C., Kool, J.B. & van Genuchten, M.Th. 1985. Determining soil hydraulic. properties from one-step outflow experiments by parameter identification II, Experimental studies. Soil Sci. Soc. Am. J. 49, 13541359. Peters, A. & Durner, W. 2008. Simplified evaporation method for determining soil hydraulic properties. Journal of Hydrology, 356(12), 147162. Romano, N. & Santini, A. 1999. Determining soil hydraulic functions from evaporation experiments by a parameter estimation approach: experimental verifications and numerical studies. Water Resour. Res. 35, 33433359. Santini, A., Romano, N., Ciollaro, G. & Comegna, V. 1995. Evaluation of a laboratory inverse method for determining unsaturated hydraulic properties of a soil under different tillage practices, Soil Sci., 160, 340351, 1995. Simunek, J., Wendroth, O. & van Genuchten, M.Th. 1998. A parameter estimation analysis of the evaporation method for determining soil hydraulic properties, Soil Sci. Soc. Am. J. 62, 894905. Toorman, A.F., Wierenga, P.J. & Hills, R.G. 1992. Parameter estimation of hydraulic properties from one-step outflow data, Water Resour. Res., 28(11), 30213028. van Dam, J.C., Stricker, J.N.M. & Droogers, P. 1992. Inverse Method for det. soil hydraulic functions from one-step outflow experiments, Soil Sci. Soc. Am. J., 56, 10421050. van Dam, J.C., Stricker, J.N.M. & Droogers, P. 1994. Inverse method for det. soil hydraulic functions from multistep outflow experiments. Soil. Sci. Amer. J., 58, 647652. Wildenschild, D., Jensen, K.H., Hollenbeck, K.J., Illangasekare, T.H., Znidarcic, D., Sonnenborg, T. & Butts, M.B. 1997. A two-stage procedure for determining unsaturated hydraulic characteristics using a syringe pump and outflow observations, Soil Sci. Soc. Am. J., 61, 347359. Yeh, W. W-G. 1986. Review of parameter identification procedure in ground-water hydrology: the inverse problem. Water Resour. Res. 22(2), 95108. Zachman, D.W., Duchateau, P.C. & Klute, A. 1982. Simultaneous Approx. of Water Capacity and Soil Hydraulic Conductivity by Parameter Identification, Soil Sci., 134, 157163. Zhang, J. & Yeh, T.-C.J. 1997. An iterative inverse method for steady flow in the vadoze zone. Water Resources Research, 33 (1), 6371. Znidarcic, D., Hwang, C. & Bicalho, K.V. 2004. Experimental determination of hydraulic characteristics for unsaturated soils. Unsaturated Soils, Balkema, 3, 11371141. Zurmhl, T. & Durner, W. 1998. Determination of parameters for bimodal hydraulic functions by inverse modeling. Soil Sci. Soc. Amer. J., 62, 874880.
432
Water retention curve and shrinkage of a waste from the paper industry
J.C.O. Bizarreta & T.M.P. de Campos
Pontifcia Universidade Catlica do Rio de Janeiro, RJ, Brazil
ABSTRACT: The paper presents results of physical-chemical and mineralogical characterization analysis performed on a waste material from the paper industry and discusses results of tests performed to obtain the water retention curve and evaluate shrinkage characteristics of the waste. The waste comprises 30% of organic fibers and 70% of minerals; the predominant minerals being calcite and aragonite. The filter paper technique was used to determine the water retention curve. The presence of fungi and sample preparation are studied in relation to the results of the water retention curve. It is shown that the material is quite different from others found in the literature, presenting a fairly high air-entry value (order of 1.000 kPa) and high shrinkage without cracking. 1 INTRODUCTION 2 MATERIAL AND METHODS
Waste Paper Sludges (WPS) are being studied by different authors (e.g. Zimmie et al. 1993, Mo-Young & Zimmie 1996, Kraus et al. 1997, Quiroz et al. 1998, Cabral et al.1999, 2004, Burnotte et al. 2000, Kamon et al. 1999, 2002a, Iberio 2007) as an alternative material to be employed in cover systems of sanitary landfills, such as oxidative or capillary barriers (e.g. Cabral et al. 2000, Diaz et al. 2000, Parent & Cabral 2006, Parent et al. 2007, Lacroix Vachon et al. 2007, Massood et al. 2008). In a municipal solid waste landfill, the main function of such cover is to reduce the ingress of water and gas output. The reduction of lost gas to atmosphere favours the use of methane as a source of energy as well as contributes to decreasing emissions of greenhouse effect gases. Two aspects are relevant to the study of the landfill cover: the Soil Water Retention Curve (SWRC) and the shrinkage potential of the material without cracking. As it is well known, the SWRC can be correlated to the permeability function, which controls the rain water infiltration in the landfill cover. The occurrence of cracking under drying provides undesirable or uncontrolled gas loss or water infiltration through the cover (e.g. Andersland & Al-Moussawi 1987). This paper presents and discuss aspects related to water retention and shrinkage of a WPS. Possible interference of fungi in the technique employed to obtain the SWRCs is examined. The obtained results are compared to those found in the current literature.
A WPS from a Paper Industry from Rio de Janeiro, Brazil, was tested in the Geotechnical and Environmental Laboratory of PUC-Rio. Such industry produces special types of paper, some of then used to manufacture cigarettes in different parts of the world. Fresh samples of the waste, collected just after its production, were used in the tests. They were kept, in the laboratory, inside a sealed container, under a temperature of 20 1C. A summary of the testing types employed is given next. Details of the methodologies used can be found in Bizarreta (2009). 2.1 Characterization
The waste was characterized through physical chemical-mineralogical and morphological analysis. Conventional geotechnical characterization tests were used in the physical characterization. Chemical tests comprise the determination of organic matter content as well as pH, C, H, N and S. Mineralogical and morphological analyses were performed by X-Ray Diffraction (XRD) and Scanning Electron Microscopy (SEM). 2.2 Water retention
The SWRC was obtained using the filter paper technique (e.g. Marinho 1994, Moncada 2008). Whatman N 42 filters were used and the calibration curve put forwards by Chandler et al. (1992) was employed to obtain the suction from the measured water content of the specimens.
433
Both drying and wetting paths were followed using specimens with different initial water content prepared by static compaction into rings of 47.5 mm of diameter and 20 mm of height, refer to Bizarreta (2009) for details. To obtain drying curves, different specimens (at a given initial humidity) were left to dry under room temperature for different periods of time. After that, each specimen was duly protected and left at rest for 2 days for internal moisture equalization. In sequence, the dimensions (diameter and height) of each specimen was measured using a caliper and air dried filter papers were positioned at their top and bottom. Then, the adequately sealed specimens were left again at rest for 15 days. To obtain wetting curves, specimens compacted at different initial moisture content were left to air dry under a temperature of circa 25C. Then, different amount of distilled water was added to different specimens. To get specimens close to saturation, vacuum was applied in specimens inserted in a desiccator. After having the specimens under different saturation conditions, the same procedure above described was followed (2 days for moisture equalization, measurement of dimensions and 15 days of final equalization). 2.3 Shrinkage
It has to be observed, however, that owing to the presence of fibers, both the liquid and plastic limits of the residue, as observed by other authors (e.g. Zimmie et al. 1993, Kamon et al. 1999, Cabral et al. 2002) were fairly difficult to be obtained using the conventional Casagrande testing proposals. The mineralogical composition of WPSs varies with the type of papers being produced and the manufacturing methodology employed. In general, the WPS is formed by volatile (organic) and nonvolatile (mineral) compounds. The volatile material consists of cellulose fibers, hemicellulose and lignin (e.g. Teixeira 2001). The non-volatile material, in general, comprise minerals such as calcite, metakaolin, talc and rutile (e.g. Cabral et al 1999, 2002; Teixeira 2001), or kaolinite (e.g. Kamon et al. 2002a, Iberio 2007). As shown in Table 2, the WPS herein considered has calcium carbonate minerals (calcite and aragonite) as dominant, non volatile minerals (64.3%). Kaolinite and anatase are components with limited participation. Figure 1 presents an electron microscope image with 2000X increase of the studied WPS. It can be observed that the cellulose fibers are inserted within a mineral matrix, that looks like a weakly cemented material comprising, essentially,
Table 2. Type Mineral WPS mineralogical composition. Material Calcite Aragonite Kaolinite Anatase Formula CaCO3 CaCO3 Si2Al2O5(OH)4 TiO2 (%) 59.9 4.4 3.5 2.2
Shrinkage tests were performed on specimens statically compacted at initial water contents of 171, 128 and 61% in a ring with 71,5 mm diameter and 20 mm height. During air drying under a temperature of circa 20C, the weight, diameter and thickness of the specimens were monitored. 3 3.1 RESULTS AND DISCUSSION WPS characterization
Physical characteristics of the WPS are presented in Table 1. The low value of specific gravity of the residue (2.08) is compatible with its high organic matter content (30%). Such values are within the limits found in the literature, that is, Gs 1.62 to 2.30 and OM 25% to 86% (e.g. Mo-Young & Zimmie 1996; Teixeira 2002). The tested material showed plasticity, with a plastic index of 94%.
Table 1. Gs 2.08 Physical characterization of WPS. wL (%) 172 wn (%) 310340 wot (%) 70 OM (%) 30 Figure 1. SEM image of the WPS with 2000X magnification.
wP (%) 78
Gs, specific gravity; wP, plasticity limit; wL, liquid limit; wn, natural water content; OM, organic matter content; wot, Proctor optimum moisture content
434
calcium carbonate. It is notable the difference in morphology and size between the cellulose fiber and the surrounding matrix. The use of EDS technique comproved the absence of calcium carbonate in the cellulose fibers. 3.2 Soil Water Retention Curve (SWRC)
Fungi colonies
3.2.1 Contraction, arching and fungi occurrences In the determination of the SWRC, large specimen contractions were observed under drying (Figure 2). This did not introduce any problem regarding the contact between the filter papers and the specimens. In some cases, however, it was observed the occurrence of arching in the central portion of the specimens, refer to Figure 3. This was particularly true for the samples with the highest initial water content (wi 162 to 172%). In these cases, the use of the top filter paper with a diameter nearly equal to that of the specimen (condition A in Figure 3) provided bad contact and, consequently, poor results. This was solved by using a filter paper with a diameter circa of 15% smaller than the initial diameter of the specimens (condition B in Figure 3). Considering the whole range of initial water content of the specimens used to obtain the SWRC, it was observed the formation of colonies of fungi in the top and bottom of the samples. These were quite easily noticeable by changes in colour at different spots of the filter paper (greenish, yellowish and deep white coloured spots). Some of such spots are highlighted in Figure 4. The occurrence of microorganisms and fungi in tests performed to obtain
Figure 4. Presence of fungi in the WPS sample.

80
75 70 65 60 55 50 1 10 100 1000
Wo = 130%, Static Compaction, 4 days Wo=125%, Static Compaction, 2 days
Suction (kPa)
Figure 5. fungi.
Water retention curves with the presence of
Figure 2. Specimens (a) without and (b) with contraction.
Figure 3.
Arching of the sample and the filter paper.
the SWRC in WPS materials has been observed by other authors (e.g. Cabral et al. 1999, Teixeira 2001, Teixeira 2002). To avoid the occurrence of such type of potential problem, this last author used a fungicide in his investigations. It is not known if the use of a fungicide affects the calibration curve of the filter paper or the suction pattern developed in any given material. Thus, such alternative was not adopted, being outside the scope of the present work to discuss further aspects related to that. It is, however, worthwhile to mention that tests performed on samples kept sealed for 2 and 4 days did not show any relevant change in the general pattern of the SWRC, refer to Figure 5, although the specimens kept longer under anaerobic conditions presented much larger growth of fungi colonies. Also, as will be seen next, results of tests performed on samples with the lowest initial water content (that showed the lesser, if any, growth of fungi colonies) matched quite well all other obtained results. Based on that, eventual effects of fungi growth on the determination of the
435
SWRC were considered to have been of minor, if any, relevance in what refers to the present work. 3.2.2 Testing results and comparisons Figures 6 and 7 show the data obtained following, respectively, the wetting and the drying paths. Figure 8 shows these same data including curves fitted using Fredlund & Xing (1994) SWRC modeling proposal. Similar results were obtained by using the model proposed by van Genuchten (1980).
100
Table 3. WPSs.
AEV and maximum obtained suction of Maximum suction (kPa) 300
Author
Method
sat 0,70,8
AEV (kPa) 2535
90 80 70 60 50 40 30 20 10 0 1 10 100 1000 10000 100000 wo = 130% wo =76% wo = 162%
Cabral et al. Pressure (1997) plate and filter paper Cabral et al. Pressure (1999) plate Teixeira Suction (2002) and pressure plates Cabral et al. Pressure (2004) plate Parent Pressure (2006) plate
0,7
2530
200 400
0,850,90
0,75
1525
40 500
0,720,75 2090
Suction (kPa)
Figure 6.
100
Obtained data under drying path.
90 80 70 60 50 40 30 20 10 0 1 10 100 1000 10000 100000
wo = 162% wo = 130% wo = 76%
Suction (kPa)
Figure 7.
110
Obtained data under wetting path.
100 90 80 70 60 50 100 WETTING a=1064.8 n= 1.27 m=1.86 Yr=10000 kPa DRYING a=1655.7 n= 2.25 m=1.37 Yr=10000 kPa 101 102
AEV = 800 kPa
Suction(kPa)
103
104
Figure 8. Obtained SWRCs by fitting data using the model of Fredlund and Xing (1994).
Unlike what has been found by other authors (e.g. Parent 2006), the degree of saturation suction plots in Figures 6 and 7 indicate that unique relationships prevail for both wetting and drying paths, regardless the initial moisture content of the WPS specimens. These figures also indicate that the filter paper technique is quite useful in obtaining SWRCs for WPS materials. As shown in Table 3, such technique has not been fully considered when studying such type of residue. In this table sat is the saturated volumetric water content and AEV is the air entry value obtained following drying paths. Maximum suctions values reflect maximum capabilities related to the equipments employed by the different authors in the determination of the SWRCs. The Fredlund & Xing model parameters that provided a best fit to the results herein considered are included in Figure 8, which shows that the SWRC for the tested WPS is not unique, i.e., it varies depending on the moisture path being followed. Regardless the path considered the residual saturation value obtained for the studied WPS is of circa 4%. Considering, as usual, the air entry value related to the drying path, an AEV of circa 800 kPa is obtained. Such value is typically more than 10 times greater than those available in the current literature (refer to Table 3). Figure 9 and Table 4 present comparisons between results obtained in the present work and by Parent (2006). It is apparent in Figure 8 that the SWRC herein obtained has simply been displaced to the right. The data included in Table 4 suggest that both the much lower content of volatile material (organic matter) and much higher content
436
100 90
220 200 180
Normal shrinkage Residual shrinkage
80 70 60 50 40 30 20 10 0 1 10 wo = 162%, e = 3.47 wo = 130%, e=2.59 Parent (2006), e=3.67 Parent (2006), e=2.71 100 1000 10000
Volume per 100 g of dry soil (cm3)
160 140 120 100 80 60 40 20 0 0
Zero shrinkage
w= 65% w=127%
Air
GAE
Water Solid
w=170% S=100%
Suction (kPa)
Figure 9. Comparison of obtained retention curve data and that produced by Parent (2006). Table 4. Influence of calcium carbonate and volatile material (OM) in the AEV of WPS materials. AEV (kPa) 15 50 700 to 800 Calcium Carbonate (%) 29.2 64.3
20
40
60
80
100 120 140 160 180
Water content (%)
Figure 10. WPS.

3,6
Shrinkage pattern followed by the tested
Author Parent (2006) Present paper
eo 3.67 2.71 3.47 2.59
OM (%) 60.3
Void ratio
Wo =76% - D= 4,75cm
3,2 2,8 2,4 2,0
Wo = 125% - D=4,75cm Wo=162% - D=4,75cm S= 100% Wo=65% - D=7,15cm Wo=127% - D=7,15cm Wo=170% - D=7,15cm
30
of calcium carbonate found in the residue herein considered may be responsible by its much higher air entry value. 3.3 Shrinkage
1,6 1,2 0 20 40 60
Water content (%)
80
100
120
140
160
180
Figure 10 shows results of drying tests performed on samples initially saturated (w 171% and 128%) and unsaturated (w 65%) plotted according to Marinho (1994). As defined by Haines (1923), the drying process initially follows the 100% saturation line (region named as normal shrinkage). The curve that follows comprises the region named as residual shrinkage, that ends at the point of zero shrinkage (which, in the case, corresponds to w 0%, i.e., the studied material presents contraction down to oven drying under 110C, with the volume per 100 g of dry soil being of 117 cm3). The point separating the normal and the residual shrinkage regions, named as General Air Entry (GAE) by Marinho (1994) corresponds to a moisture content of circa 90%. Figure 11 includes data corresponding to drying of specimens used to obtain the SWRCs, in a plot of water content void ratio. A larger scatter is observed, probably resulting from the difference in diameter of the different sets of tested
Figure 11.
Drying curve of the WPS.
specimens, which implies in different accuracies in volume determination. Nonetheless, this figure indicates that the voids ratio corresponding to the zero shrinkage condition of the WPS is of circa 1.4. Also, the GAE value above mentioned can be obtained from this type of plot. Even though large volume changes occurred under drying, following, apparently, a linear pattern with initial water content (Figure 12), no fissures detectable under naked eye were observed in none of the specimens submitted to drying, regardless their initial water content (Figure 13). This is quite remarkable particularly when it is considered the large change in morphology between the specimens initially saturated (w 171% and 128%) and those unsaturated (w 65%). As indicated in Figure 13, the initially saturated WPS showed, after drying, a smooth finishing, typical of fine grained, low plasticity clay, while the initially unsaturated residue
437
Shrinkage volumetric change (%)
50 45 40 35 30 25 20 15 10 60 70 80 90 100 110 120 130 140 150 160 170 180
fungi appears to not have affected obtained results. This, however, has to be further investigated. ACKNOWLEDGMENTS This work was developed as part of the project PRONEX-Rio E-26/171.544/2006, financed by FAPERJ/CNPq. The first author thanks CAPES for the support during his M.Sc. at PUC-Rio. REFERENCES
Andersland O.B. & Al-Moussawi H.M. 1987. Crack formation in soil landfill covers due to thermal contraction. Waste Management & Research, 5, 445452. Bizarreta J.C.O. 2009. Evaluation of the potential use of waste paper industry in capillary barrier systems. M.Sc. Thesis. Catholic University of Rio de Janeiro (in portuguese). Burnotte F., Lefebvre G., Cabral A., Audet C. & Veilleux A. 2000. Use of deinking residues for the final cover of a MSW landfill. In: 53rd Canadian Geotechnical Conference, volume 1, pages 585591, Montral, Canada. Cabral A.R., Lefebvre G., Burnotte F., Panarotto C.T. & Pastore E.L. 1999. Use of Pulp and Paper Residues as an Alternative Cover Material to Landfill and to Acid Generating Tailings. Segundo Seminrio de Meio Ambiente em Indstria de Processo. So Paulo, Jun pp. 5670. Cabral A., Racine I., Burnotte F. & Lefebvre G. 2000. Diffusion of oxygen through a pulp and paper residue barrier. Canadian Geotechnical Journal, Vol. 37, pp. 201217. Cabral A.R., Lefebvre G., Burnotte F. & Panarotto C.T. 2002. Geotechnical characterization of a pulp and paper (deinking) residue used in cover systems. Proceedings of the 4th International Conference on Environmental Geotechnics. vol. 1. Balkema, Rio de Janeiro, Brazil, pp. 207212. Cabral A.R., Planchet L., Marinho F.A. & Lefebvre G. 2004. Determination of the Soil Water Characteristic Curve of Highly Compressible Materials: Case Study of Pulp and Paper By-product. Geotechnical Testing Journal, ASTM, Vol. 27, N2. Chandler R.J., Crilly M.S. & Montgomery-Smith G. 1992. A Low-cost Method of Assessing Clay Desiccation for Lowrise Buildings. Proceedings, Institute of Civil Engineering, Vol. 92, No. 2, pp. 8289. Daz A., Vilches p., Salas O., Torrea J. & Zulueta A. (2000). Paper sludge hydraulic barriers: use as construction material in landfill. Ingeniera Qumica, Ao XXXII (367), pp. 109114, ISSN: 02102064 (in spanish). Fredlund D.G. & Xing, A.Q. 1994. Equations for the soil-water characteristic curve. Canadian Geotechnical Journal, 31: 521532. Haines, W.B. 1923. The volume changes associated with variations of water content in soil. Journal of Agricultural Science, 13, 296311.
Compaction water content (%)
Figure 12. content.
Shrinkage volume change versus water
wi = 171%
wi = 128%
wi = 65%
Figure 13. Typical condition of specimens after drying.
showed morphology somewhat granulated, with lumps of probably calcium carbonate minerals forming large grains. Investigations performed by authors such as Kraus et al. (1997), Moo-Young & Zimmie (1996), & Kamon et al. (2002b) indicated that WPS behaves like high plasticity clays, presenting cracks under drying. The particular mineralogy, associated to the fiber/mineral dosage of the studied WPS may be related to its apparently unusual behaviour on drying. 4 CONCLUSIONS
The studied WPS showed characteristics suggesting that it may be useful as an alternative material to be employed in cover systems of sanitary landfills. Unlike other waste paper sludges, the WPS investigated appears to be low sensitive to variation on initial compaction water content, presents a fairly high air entry value and does not crack under drying. The relatively high proportion of calcium carbonate minerals and low content of volatile material may explain the differences in behaviour found when considering results available for similar materials in the current literature. The use of the filter paper technique proved to be quite useful in the determination of suctionwater content relationships. The occurrence of
438
Iberio L.S. (2007). Characterization of a residue from the paper industry to use as geotechnical material. M.Sc. Thesis, Federal University of Rio Grande do Sul (in portuguese). Kamon M., Katsumi T., Rajasekaran G. & Inazumi S. 1999. Potential application of paper mill sludge as landfill cover, Third Japan Nat. Symp. Environ. Geotechnology, JGS, pp. 8590. Kamon M., Inazumi S., Rajasekaran G. & Katsumi T. 2002a. Evaluation of waste sludge compatibility for landfill cover application, Soils and Foundations, Vol. 42, No. 4, pp. 1327. Kamon M., Inazumi S. & Katsumi T. 2002b: Performance evaluations of landfill cover systems with sludge barriers, Geotechnical Engineering, Journal of the Southeast Asian Geotechnical Society, Vol. 33, No. 3, pp. 113132. Kraus J.F., Benson C.H., van Maltby C. & Wang X. 1997. Laboratory and field hydraulic conductivity of three compacted paper mill sludges. Journal of Geotechnical and Geoenvironmental Engineering, 123: 654662. Lacroix Vachon B., El-Ghabi, B. & Cabral, A.R. 2007. valuation prliminaire de lefficacit du recouvrement avec double effet de barrire capillaire install au site de st-tite-des-caps, qc. In, Ottawa. Marinho F.A.M. 1994. Shrinkage behavior of some plastic soils. Ph.D. Thesis, Imperial College, London. Massood A.A., Lacroix B.V. & Cabral A. 2008. Hydraulic barrier and its impact on the performance of cover with double capillary barrier effect. GeoEdmonton08. Moncada M.P.H. 2008. Evaluation of hydraulic properties of unsaturated tropical soils. D.Sc. Thesis. Catholic University of Rio de Janeiro (in portuguese). Moo-Young H. & Zimmie T. 1996. Geotechnical properties of paper mill sludges for use in landfill covers. Journal of Geotechnical Engineering. ASCE. Vol. 122, N 9, pp. 768776.
Parent S.E. 2006. Hydraulic and geotechnical aspects of capillary barrier design using a highly compressible recycled material, Ph.D Thesis, Univerisit de Sherbrooke, Quebec. Parent S.E. & Cabral A. 2006. Design of inclined covers with capillary barrier effect. Geotechnical and Geological Engineering, 24, 689710. Parent S.E., Cabral A. & Zornberg J. 2007. Water retention curve and hydraulic conductivity function of highly compressible materials. Canadian Geotechnical Journal, Vol 44, pp. 12001214. Quiroz J.D. & Zimmie T.F. 1998. Paper mill sludger landfill cover constrution. Recycled materials in geotechnical applications, Vipulanandan, C. and Elton, D.J. (eds.), Geotechnical special publication No. 79, ASCE, pp. 1936. Teixeira C.E. 2001. volution biochimique des rsidus de dsencrage dans un contexte de valorisation comme barrire de recouvrement. Ph.D Thesis, Sherbrooke: Universit de Sherbrooke. Teixeira P.F. 2002. Research on the geotechnical characteristics of waste paper. M.Sc. Thesis, Polytechnic School, University of So Paulo. (in portuguese). van Genuchten M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal, 44: 892898. Zimmie T. F., Moo-Young & LaPlante 1993. The use of waste paper sludge for landfill cover material. Gren 93: Waste disposal by landfill; Symp. Of Geotechnics related to the environment, R.W. Sasby, Wd., Balkema, Rotterdam, 487495.
439
Soil-water characteristic curves obtained through the wetting paths for a tropical soil profile
B.H. Bonder
Universidade Estadual de Campinas and Construtora Norberto Odebrecht, Santos-SP, Brasil
M.G. Miguel
Universidade Estadual de Campinas, Campinas-SP, Brasil
ABSTRACT: This research presents the soil water characteristic curves (SWCC) for a tropical soil profile, typical of 14% of the region of Campinas, in the state of So Paulo, Brazil. This profile consists of colluvial, lateritic and collapsible silty clay, with an approximate 6.5 m thickness and present in a clayed silt soil. The water level is approximately at 17 m. To determine the curves, the filter paper and axis-translation method with wetting paths were used. Impregnated thin-layer plates were prepared with collected undisturbed samples, so as to obtain their structures in the field along the profile, until a 4.5 m depth. Geotechnical characterization tests and chemical-mineralogical tests were carried out on the disturbed samples. The SWCCs for the soil profile until a 4.5 m depth were the bimodal type, typical of tropical soils with macro and micro-aggregated structures. There were differences between the curves obtained with the different methods, mainly due to the way the wetting paths were carried out on the specimens. These differences are very clear for suctions below 100 kPa, where the influence of macropores is more expressive. 1 INTRODUCTION order to understand the behavior of the shearing, compressibility and permeability of these soils. This variable is called suction and can be matric or total. First, in order to evaluate changes in the suction of a tension state of an unsaturated soil, it is necessary to determine the Soil-Water Characteristic Curve (SWCC). This curve is a graphical representation of the suction relation (matric or total) and quantity of water that can be represented by the gravimetric moisture content (w), volumetric moisture content () or the degree of Saturation (Sr). SWCC describes a different pathway in terms of soil drying or wetting. Because of this, the curves obtained by the drying or wetting paths do not coincide, giving rise to a phenomenon called curve hysteresis and which is a characteristic of soil suction (Bonder, 2008). Hysteresis caused by drying and wetting paths can be attributed to non-uniformity of the voids, to the air bubbles captured by the voids of the soil during the wetting and structure alteration due to the expansion or contraction of the soil (Calle, 2000). According to Rojas (2002), the real value of suction depends not only on the degree of saturation, but also on the initial state of the soil and the whole history of drying and wetting until that moment.
1.1 Lateritic soils Lateritic soils are tropical soils with certain peculiarities that make them present physical, mechanical and hydraulic behavior that differ from the Classical Soil Mechanics models. The formation of lateritic soils occurs through the action of the pedological alteration processes that act later, or together with the mechanisms of disaggregation and decay caused by physical and chemical weathering, with a more intense chemical action. The pedological alteration processes that contribute most significantly to lateritic soil formation is leaching, which is the intense migration of particles under the action of infiltration and evaporation, resulting in a layer of porous soil (high permeability), consisting of more stable minerals (quartz, kaolinite and magnetite) and a low degree of saturation. As this process is very slow, it occurs in the well-drained superficial layers, located above the water level, therefore, not saturated. (Committee on Tropical Soils of ISSMFE, 1985; Gidigasu, 1976). 1.2 Soil-water characteristic curves The mechanics of unsaturated soils have a significantly important variable state of stress in
441
1.3
Methods of measurements of suction
There are several methods to determine soil suction, both in the field, as well as in the laboratory. According to Soto (2004), suction measurements in the laboratory and in the field, come up against a number of experimental problems and usually must resort to different methods in order to know if the suction, within a wide range, is sufficient for practical use. 1.4 Filter paper method
The filter paper method has been widely used to obtain SWCC as it is easy to use, inexpensive and covers a relatively large suction range. The disadvantage of this method is the need for extreme caution in its implementation. The advantage of this method is that the paper filter is adapted to the value of the soil suction, instead of the soil adjusting to the applied suction value. The test consists of placing a filter paper of known retention characteristics in a hermetic environment along with a soil sample. Due to the contact of the paper with the soil, which is able to retain moisture content, there is water migration until potential equilibrium is reached, thus obtaining matric suction (Gardner, 1937). If water in the soil is not in direct contact with the filter paper, total suction can be obtained, after potential equilibrium. This equilibrium time is being studied by several authors. Feuerharmel et al., (2006) suggest a seven day equalization time for suction values above 10,000 kPa and four days for values lower than 10.000 kPa. Marinho (1997) argues that equalization time is related to the type and level of suction. This author suggests a seven day period, when measuring matric suction, regardless of the level of suction. According to ASTM (2004), the minimum equalization time between the filter paper and the soil is seven days. Equilibrium time is a very important factor to obtain the correct suction value. The procedure of this method is quite simple; however it requires great care in determining the filter paper moisture content. According to Chandler and Gutierrez (1986), filter paper weighing time should be approximately 30 seconds to prevent that it gains or loses moisture to the atmosphere. According to Marinho (1994), the filter paper transfer time for a closed capsule or a zip lock plastic bag must be 5 seconds at most. 1.5 Axis-translation method
reference from atmospheric conditions is changed. Assuming that this produces the same increase in water pressure (uw), the difference (ua uw) remains (where ua is the air pressure), or better, the suction is maintained and the pressure in the water is increased, taking it to measurable values. In principle, the axis translation method can be used either connected to other equipment for measuring other properties (resistance, compressibility, etc.), controlling the suction during these tests, as well as to impose suction in soil samples. Special porous plates that allow the flow of water but not that of air are needed for this method. Maximum suction to be applied depends on the input pressure of air on the porous plates, which can reach up to 1,500 kPa (Soto, 2004). This is not recommended for suction measurement below 10 kPa (Libardi, 1995). Air permeability is practically zero in an unsaturated soil with high value degree of saturation (Boocking and Fredlund, 1980), as well as requiring relatively expensive equipment. 1.6 Soil Characterization This study was carried out with soil from the Experimental Field Soil Mechanics and Foundations (EFSMF), the State University of Campinas. EFSMF is a geological-geotechnical profile which is a trait of 14% of the region of Campinas, located in the central-east portion of the state of Sao Paulo in Brazil. The subsoil is made up of a layer of approximately 6.5 m thick colluvial soil. This layer consists of a lateritic silty clay (Miguel et al., 2007). Below it, there is a layer of non-lateritic clayey-sandy silt up to a depth of 19 m. Between 6.5 and 7 m depth, the profile shows a layer of concretioned material, consisting of a little compact yellowish brown sandy silt. The water table is 17 m, approximately. Several field tests were carried out at EFSMF such as SPT (Standard Penetration Test), CPTU (Cone penetration test), DMT (Dilatometer test) and others (Cavalcante et al., 2007). The fact of the superficial layer of EFSMF consisting of lateritic clay has been the subject of research because of its geotechnical behavior peculiarities. X-ray diffraction tests carried out on samples collected throughout the profile indicated the presence of minerals such as kaolinite, gibbsite, hematite, goethite and quartz. Table 1 presents natural specific values (nat), field moisture content of (wfield), void ratio (e) and degree of saturation (Sr) of this EFSMF layer for the depths studied (Miguel et al., 2007). Figure 1 shows the granulometric curves for the same samples with and without the use of defloccullant (CD and SD, respectively) during the sedimentation and depth function phase (z) of the samples.
The axis translation method has been developed in order to prevent that the water from the measuring system and interstitial water work with pressure values below that of the atmosphere (Hilf, 1956 cited in Marinho, 1997). In other words, where the atmospheric pressure is increased, the water pressure
442
Table 1. Physical index values of undisturbed samples (Miguel et al., 2007). Prof. (m) 1.5 2.5 3.5 4.5 5.5 6.5 nat (kN/m3) 11.80 13.71 13.96 14.38 14.11 14.68 field (%) 24.2 23.7 22.3 27.2 22.2 22.4 e () 2.14 1.74 1.68 1.72 1.66 1.55 Sr (%) 33.7 44.3 40.7 48.6 41.1 44.2
(capillary impregnation), using a vacuum desiccators, and then with methylene blue in order to obtain pore distribution. After sample hardening, they samples were sliced, to make the thin plates. The interpretation of these specimens was carried out using a petrographic microscope with 2.5X, 10.0X, 25.0X, 40.0X and 50.0X objective lens, associated with data from X-ray Diffraction and X-ray Fluorescence. 2.3 SWWCs by filter paper method
100% 90% 80% 70%

CD -z = 1.5 m CD -z = 2.5 m CD -z = 3.0 m CD -z = 4.3 m CD -z = 5.0 m CD -z = 6.0 m SD -z = 1.5 m
60% 50% 40% 30% 20% 10% 0% 0.001
SD -z = 2.5 m SD -z = 3.0 m SD -z = 4.3 m SD -z = 5.0 m SD -z = 6.0 m
0.01
0.1
10
Grain-size (mm)
Figure 1. Granulometric curves with and without the use of deflocculant.
The aim of this study was to determine SWCCs in wetting paths, from the lateritic and colluvial surface layer soil of EFSMF up to a 4.5 m depth, by means of filter paper and the axis translation methods in the laboratory, and analyze these curves in terms of the chemical-mineralogical structure and composition, considering their formation processes by interpreting the thin and impregnated plates in order to better understand the behavior of unsaturated tropical soil. 2 2.1 MATERIALS AND METHODS Samples
The soil samples studied were undisturbed and disturbed, and were taken from the colluvial and lateritic surface layer of the EFSMF profile at 1.5 m, 2.5 m, 3.5 m, 4.5 m depths, by excavating a 1 m diameter pit. 2.2 Thin plates
Thin plates were prepared and impregnated by means of undisturbed samples collected at 1.5 m, 2.5 m, 3.5 m and 4.5 m depths. The plates were impregnated with small quantities of acrylic resin
So as to determine the SWWCs due to wetting paths, several specimens from undisturbed samples were made. These specimens were molded in 2.0 cm and 5.0 cm diameter metal rings, by means of samples collected at 1.5 m, 2.5 m, 3.5 m and 4.5 m depths. Two specimens were used at each depth level. For the wetting paths, the specimens were placed in the furnace for 24 hours to start the tests at a very low humidity level, between zero and 1%. The test was carried out in a temperature-controlled room at a temperature of 20C. The filter papers used were Whatman No 42, taken out of the box and immediately placed in contact with the specimen. The specimens receive a PVC plastic film at the bottom of the metal ring, attached to the sides of the rings, to avoid material loss during the handling of the specimens. Thereafter, each set (specimen and filter paper) was protected by aluminum foil, as suggested by Feuerharmel et al., (2006), and an overload (aluminum capsule cap) was placed to ensure that there would be total contact between the paper and soil. The whole set was protected again, but now with the PVC film and placed inside a sealed plastic bag, which in turn was placed inside a plastic box which was also closed, to ensure that the variation in air humidity would not interfere in the equilibrium of filter paper with soil. The entire procedure was done as quickly as possible to prevent the filter paper from reaching equilibrium with the air, that is, with relative air humidity. The specimens were left in the plastic box for seven days so that the equilibrium time between the filter paper and soil would be reached. After seven days, the specimens were taken from the box and plastic bag, following the steps: the PVC film and aluminum protection was removed, along with the overload, and then after that the filter paper in contact with the soil. After that, the paper was placed inside an aluminum capsule, the weight of which was known, to be weighed on a scale with 0.1 mg resolution. This procedure, from the time of the removal of the protection layers until the placing of the filter paper inside the capsule was carried out in about 6 seconds, so as to prevent the
Percent
443
filter paper losing or gaining water. After weighing the capsule and the paper filter set, the filter paper was taken to a furnace oven for 24 hours. To determine other values of moisture content equilibrium of the wetting paths of the specimen, a certain amount of distilled water was added uniformly with the help of a dropper, on the surface of the specimen, waiting for two hours so that the water could infiltrate the specimen (equilibrium) and not accumulate on its surface. After that, the process described above was repeated. In all the testing phases, the handling of the filter paper was carried out with a metal clamp, to avoid altering the characteristics of the paper. The moisture content equilibrium rates of the filter paper and soil were calculated, in which each gravimetric moisture content corresponds to a matric suction estimated from the moisture of the filter paper, using the calibration curve for Whatman paper No 42, obtained by Chandler et al., (1992). These suction and moisture content values correspond to a SWCC point. 2.4 SWWCs by axis-translation method
4.5 m depth, one can note that they are made up of very angular quartz grains and, with undulating extinction, typical of sedimentary material. The surroundings of the grain were covered by ferruginous reddish material, made up of hematite and at times goethite. Some of these grains are fractured and filled with the same material that is around it. In the thin plates with the 1.5 m and 2.5 m depth specimens, there is a predominance of interconnecting pores (Figure 2). Unlike what was observed in the plates with the 4.5 m depth specimens (Figure 3). In the thin plates with 3.5 m depth specimens, there is a equilibrium between communicating and noncommunicating pores. 3.2 SWCC obtained by filter paper method
Figures 4, 5 and 6 present SWCCs obtained with filter paper and axis translation method due to the wetting paths.
In these tests, three specimens of undisturbed soil samples for each one of the 1.5 m, 2.5 m, 3.5 m, 4.5 m depths, were taken to the Richards apparatus and placed on a saturated porous plate with low air-entry value (for suctions between zero and 100 kPa) and on a porous stone saturated with high air-entry value (for suctions between 100 kPa and 1.000 kPa). First, the specimens were air dried for 12 hours and their initial matric suctions were measured using the filter-paper method. The initial matric suction value was near 450 kPa. After that, the specimens were put on the porous plate to reach mass equilibrium. Next, a new decreasing suction value was applied until new mass equilibrium was reached. This process was continuously repeated always applying a lower suction value until reaching 0 kPa (Miguel and Vilar, 2009). The stages of applied suction for the wetting procedure were 450 kPa, 400 kPa, 300 kPa, 250 kPa, 150 kPa, 100 kPa, 80 kPa, 70 kPa, 40 kPa, 30 kPa, 20 kPa, 10 kPa, 6 kPa, 4 kPa and 0 kPa. At the end of the test, specimens were saturated determining the moisture content for each pressure stage (matric suction), which led to the pair: matric suction and gravimetric moisture content, for the construction of SWCCs. 3 3.1 PRESENTATION AND ANALYSIS Structure and mineralogical composition
Figure 2. Hexagonal particles of hematite, pores of variable dimension and micro-aggregates (1.5 m depth). Miguel and Vilar (2009).
In the interpretation of thin plates impregnated with undisturbed soil samples collected up to a
Figure 3. Difference of behaviour and porosity of the hematitic material, when it borders the grains and when it cements them (4.5 m depth). Miguel and Vilar (2009).
444
50 45 Gravimetric Moisture Content (%) 40 35 30 25 20 15 10 5 0 0.1 1 10 100 1000 10000 100000 Matric Suction (kPa)
Filter paper Axis-translation
50 45 Gravimetric Moisture Content (%)

Filter paper
40 35 30 25 20 15 10 5 0 0.1 1 10 100
Axis-translation
1000
10000
100000
Figure 4. SWCCs, in wetting paths, obtained with filter paper and axis-translation methods for 1.5 m depth.
50 45 Gravimetric Moisutre Content (%) 40 35 30 25 20 15 10 5 0 0.1 1 10 100 1000 10000 100000 Matric Suction (kPa)
Filter paper Axis-Translation
Matric Suction (kPa)
50 45 Gravimetric Moisture Content (%) 40 35 30 25 20 15 10 5 0 0.1 1 10 100 1000 10000 Matric Suction (kPa) 100000
Axis-translation
Filter paper
of thin plates, specimens from shallower depths show a predominance of interconnecting pores, unlike those of greater depths. That is, the smaller the depth, the greater the voids, and the greater the difference between the curves in the initial pathway up to 100 kPa. The difference between the curves obtained by wetting paths, based on the axis translation and filter paper method is mainly due to the wetting procedure used in the filter paper method, in other words, water is placed on the specimens, with the aid of a dropper. Based on this, the soil carried out passive work, because it is being induced to wet. However, in the axis translation method, the soil really has to work to reach moisture equilibrium with the suction imposed in the Richards chamber. This difference is very clear for suctions below 100 kPa. For higher suctions, the two methods provide almost identical values. Another factor that provides this difference is the occluded air that is present in the pores of this specimen due to the procedure of the axis translation method, as from the moment that suction is applied, the lack of effective contact with the saturated porous plate will allow the air to enter the pores, leaving the air occluded. It can also influence, due to the cycle that can occur during the wetting of the specimen (wetting-drying-wetting), the carrying out of the tests, because when you put water in the specimens (wetting), there was a short waiting period before placing the filter paper, and that during this waiting period there was a moisture loss in the specimen. 4 CONCLUSION
Looking at Figures 4 to 7, there is one major difference between the curves, especially on the initial pathway between 0.1 to 100 kPa. According to Table 1, samples from shallower depths have larger voids. As indicated by the interpretation
Both the filter paper and axis translation methods have some limitations. The filter paper cannot absorb water after 175% moisture content of the
445
paper, which corresponds to a 3 kPa suction, but suctions up to 5 kPa are also not reliable. The axis translation method has several mitigating circumstances. The specimen equilibrium time with the porous plate is long, the contact between the plate and the soil must be guaranteed as much as possible and, due to the imposition of matric suction, the formation of occluded air in the specimen is almost inevitable. It is important to understand how they obtained the SWCCs by wetting paths. As has been seen, there is a difference in the curves obtained by different methods, due to the procedure used in the specimen wettings. The SWCCs by wetting paths can be obtained by adding water directly to the specimens, as is done in the filter paper methods, or causing the specimens to carry out work to wet itself, as what occurs in the axis translation method. ACKNOWLEDGEMENT The authors would like to thank Fundao de Amparo Pesquisa do Estado de So Paulo (Fapesp). REFERENCES
ASTM D 5298-03. 2004. Test method for measurement of soil potential suction using filter paper, Volume 04.08. Bonder, B.H. 2008. Curvas de reteno de gua de um solo coluvionar de Campinas obtidas mediante tcnicas de laboratrio e de campo. Dissertao de Mestrado. Universidade Estadual de Campinas, Brasil. 130p. (in Portuguese). Boocking, and Fredlund, D.J. 1980. Limitation of axis translation technique, 4o International Conference on Expansive Soils, pp.117135, Denver, EUA. Calle, J.A.C. 2000. Anlise de ruptura de talude de um sSolo no saturado, Tese de Mestrado, EESC-USP/ So Carlos. (in Portuguese). Cavalcante, E.H.; Giachetti, H.L.; Danziger, F.A.B.; Coutinho, R.Q.; Souza, A.; Kormann, A.C.M.; Belincanta, A.; Pinto, C.S.; Branco, J.C.M.C.; Ferreira, C.V.; Carvalho, D.; Marinho, F.A.M.; Cintra, J.C.A.; Dourado, K.C.A.; Moraes, L.S.; Albuquerque Filho, L.H.; Almeida, M.S.S.; Gutierrez, N.H.M.; Albuquerque, P.J.R.; Chamecki, P.R.; Cunha, R.P.; Teixeira, R.S.; Menezes, S.M.; Lacerda, W.A. 2007. Campos experimentais brasileiros. Revista Luso Brasileira de Geotecnia. 111, 99205 (in Portuguese).
Chandler, R.J. and Gutierrez, C.I. 1986. The filter paper method of suction measurement. Geotechnique, (36): 265268. Chandler, R.J.; Crilly, M.S.; Montgomery-Smith, G. 1992. A low-cost method of assessing clay desiccation for low-rise buildings. Proc. of the Institution of Civil Engineering. 92 (2), 8989. Committee on Tropical Soils of ISSMFE. 1985. Peculiarities of Geotechnical Behavior of Tropical Lateritic and Saprolitic Soils Progress. Edile: So Paulo/SP. Feuerharmel, C.; Gehling, W.Y.Y.; Bica, A.V.D. 2006. The use of filter-paper and suction-plate methods for determining the Soil-Water Characteristic Curve of undisturbed colluvium soils. Geotechnical Testing Journal. 29 (5), 419425. Gardner, W.R. 1937. A method of measuring the capillary tension of soil moisture over a wide moisture range. Soil Science. 43 (4), 277283. Gidigasu, M.D. 1976. Laterite Soil Engineering: Pedogenesis and Engineering Principles. Developments in Geotechnical Engineering, 9. Elsevier Scientific Publishing Company. Libardi, P.L. 1995. Dinmica da gua no solo. Edio do Autor, ESALQ, Departamento de Fsica e Meteorologia, 497 p. (in Portuguese). Marinho, F.A.M. 1994. Tcnica do papel filtro para medio de suco. Encontro sobre Solos No Saturados. Porto Alegre/RS. 1, 112125. (in Portuguese). Marinho, F.A.M. 1997. Medio de suco em solos. III Simpsio Brasileiro de Solos No Saturados. Rio de Janeiro/RJ, 1, 373397. (in Portuguese). Miguel, M.G.; Albuquerque, P.J.R.; Azevedo, G.S.; Silva, G.S.V.; Carvalho, D. de. 2007. Lateritic behaviour of the colluvial soil from Campinas-SP-Brasil. XIII Pan American Conference on Soil Mechanics and Geotechnical Engineering, Isla de Margarita. CDRom. (in Portuguese). Miguel, M.G. and Vilar, O.M. 2009. Study of the water retention properties of a tropical soil. Canadian Geotechnical Journal. 46, 10841092. Rojas, E. 2002. Modeling the Soil Water Characteristic Curve during wetting e drying cycles, Proc. of the Third International Conference on Unsaturated Soil, 1, 215219, Recife, Brazil. Soto, M.A. 2004. Comparao de mtodos de imposio e de controle de suco em ensaios de solos no saturados. Tese de Doutorado, Escola Politcnica da Universidade de So Paulo, So Paulo. (in Portuguese).
446
An investigation on the effect of initial water content and dry density on drying soil-water characteristic curve of a cohesionless soil
Malaya Chetia & S. Sreedeep
Department of Civil Engineering, Indian Institute of Technology, Guwahati, Assam, India
ABSTRACT: Soil-Water Characteristic Curve (SWCC) is important for understanding different behavior of unsaturated soil such as strength, permeation, compressibility etc. Each soil has got a peculiar SWCC that may get influenced by different factors. This paper describes a few experimental results for obtaining continuous drying SWCC for a cohesionless soil with different initial water contents and dry densities using a tensiometer and a volumetric water content sensor. The main objective of this paper is to investigate the effect of the above mentioned parameters on SWCC of cohesionless soil. The initial specimen density was controlled by changing compaction energy and water content. The details of the methodology adopted are discussed in this paper. The test results indicates that the SWCC obtained in terms of volumetric water content is more sensitive to the changes in initial water content than the initial dry density. 1 INTRODUCTION measuring the wetting curve, only the drying curve is usually measured (Hillel 1980). Following key points are relevant for SWCC: 1. The volumetric water content at saturation, s, describes the water content at which the soil is completely saturated and typically depicts the initial state for the evaluation of the drying path. 2. The air-entry value, a, is determined as the point where the two tangent lines of the drying curve intersect each other. This value corresponds to the matric suction required to remove water from the largest pores (Brooks & Corey 1966). 3. The water content corresponding to the asymptote of the SWCC at low degree of saturation is called the residual water content (r). The soil suction corresponding to the residual water content is called the residual soil suction, r. 4. The water-entry value, w, on the wetting SWCC, is defined as the matric suction at which the water content of the soil starts to increase significantly during the wetting process. It is observed that each SWCC is influenced by different physical, chemical, mineralogical and mechanical properties. Among these properties, the influence of water content and density on SWCC are much discussed by the researchers. However, there are different contrasting studies pertained to the effect of dry density on the SWCC (Birle et al. 2008; Sun et al. 2007; Marinho & Stuermer 1999;
During the last few decades several researchers and practitioners have stressed the importance and need of adopting unsaturated soil mechanics in the analysis of problems associated with compacted soils, swelling clays and other problematic soft soils (Fredlund & Rahardjo 1993; Houston et al. 2006). To accomplish this, the determination of a characteristic curve known as the Soil-Water Characteristic Curve (SWCC), which is a graphical relationship between soil suction () and gravimetric water content, (w), volumetric water content, (), or the degree of saturation, Sr becomes essential. Numerous research works have demonstrated that the SWCC is central to the behavior of an unsaturated soil (e.g., Fredlund & Rahardjo 1993; Barbour 1998). The SWCC can be related to other properties describing the behavior of the soil, such as the unsaturated coefficient of permeability (Fredlund et al. 1994) and the shear strength (Vanapalli et al. 1996). In summary, SWCC is most desirable for implementation of unsaturated soil theory into practice (Miao et al. 2006). A typical SWCC is S-shaped and is hysteretic; that is, for a given water content, higher matric suctions exist for drying than wetting process. Hysteresis is caused by size differences between the primary pores and the interconnecting pore throats, changes in the contact angle during wetting and drying, and trapped air (Tinjum et al. 1997). Because of experimental difficulties inherent in
447
Krahn & Fredlund 1972). Therefore, there is a need to further explore its influence on different soil type. Hence, this paper investigates specifically the influence of initial water content and dry density on SWCC of a poorly graded sandy soil. For this purpose, drying SWCC corresponding to different initial densities were obtained using a tensiometer and a volumetric water content sensor. For presenting SWCC, the volumetric water content (defined as the volume of water in soil divided by the total volume of the soil) is chosen because the volumetric water content can represent the effect of water content, degree of saturation and void ratio of the sample (Miao et al. 2006). It is also felt that is a better representation of volumetric properties of porous media. 1.1 Role of initial water content on SWCC
Tinjum et al. (1997) demonstrated the influence of compaction of the soil sample at optimum moisture content, wet of optimum and dry of optimum. The SWCCs obtained varied with the compaction water content. The higher the compaction water content, the steeper the SWCC. It has been opined that the variation in SWCC corresponding to different compaction water content is mainly due to its influence on micro and macro fabric of the compacted soil. Curves with different initial water contents tend to converge at high suction values. Marinho & Chandler (1993) have observed linear SWCCs independent of the initial water content. Marinho & Stuermer (1999) have studied the influence of compaction energy on the SWCC of a residual soil. Soil samples were prepared using the modified Proctor, the standard Proctor and the nonstandard energy. It was observed that at a particular energy of compaction the initial water content affects the SWCC. The different initial paths are followed by the drying SWCCs and the later portions of the curves coincide. Birle et al. (2008) have studied the influence of different initial water contents on the SWCC of compacted clay. From the observations it has been concluded that the initial water content has a much larger influence on the SWCC than the initial dry density. Ng & Pang (2000) have studied the influence of compaction water content on the SWCC of a volcanic soil under various vertical stresses. Their results show that the SWCC of compacted fine-grained soil is strongly influenced by the initial water content. Li et al. (2009) have observed that the air entry value of soil sample of lower initial water content is lower. 1.2 Role of initial density on SWCC
for a silty sand with a high initial compacted density had a higher air entry value than that of a low initial compacted density. The rate of desorption was found to be less at higher initial density. Birle et al. (2008) have investigated the influence of dry density on SWCC of compacted clay. Their results show that the SWCC in terms of gravimetric water content is independent of the initial dry density. The same was observed by other researchers (Box & Taylor 1962; Olson & Leonard 1965; Campbell & Gardner 1971; Krahn & Fredlund 1972; Erzin & Erol 2007). However, expressing the SWCC in terms of the degree of saturation the influence of different dry densities is highlighted. Sreedeep & Singh (2005) have stated that the dry density will not affect the soil suction until a certain critical value of dry density has been reached. Sun et al. (2007) have performed a series of controlled-suction tests to study the collapse behavior of unsaturated compacted clay with different initial dry densities and suctions. It has been found that the SWCC of a compacted soil during wetting is mainly dependent upon the current density. The SWCC in terms of suction and degree of saturation shifts upwards with increasing specimen density. The authors have mentioned that it is more difficult to identify the influence of the initial density on the SWCC expressed in terms of water content. But it is relatively easy to note the upward shift of the log-Sr curves with increasing initial density. Sun et al. (2006) have investigated the effect of the initial density on SWCC of unsaturated compacted specimens. The test results indicate that the initial density has major influence on the SWCC. They have observed the upward shift of SWCC with increasing specimen density. Miao et al. (2006) have found that the air-entry value of the SWCC increases with the dry density, but the variation of the residual water content is small over the dry density range (1.42 gm/cm3, 1.48 gm/m3 and 1.54 gm/cm3) considered. Yang et al. (2004) have studied the factors affecting drying and wetting SWCCs of sandy soils. It has been observed that soils with a low dry density have a lower air-entry value and residual matric suction than soils with a high dry density. Li et al. (2009) have found that the air entry value and slope of SWCC vary with specimen density. The slope of the SWCC is more when the density of the sample is smaller. 2 EXPERIMENTAL DETAILS
In the literature several investigations on the influence of soil density on the SWCC are found. Croney & Coleman (1954) have reported that
A Perspex box illustrated in Figure 1, equipped with T5 tensiometer (UMS GmbH, Munich) and ECH2O-TE volumetric water content sensor (Decagon Inc., USA), was used to measure suction and volumetric water content, respectively, of dynamically compacted poorly graded sandy soil in the perspex mold. Sensors were secured in place to ensure proper contact with the soil.
448
Sensors support Tensiometer To data logger Perspex box ECH2O-TE sensor To data logger
Table 1. Physical properties classification of the soil SA. Property Specific gravity
and
Soil SA 2.66 0 38 60 2 0 0.09 3.90 0.83 SP 0.92 0.67
Perspex mold 11 cm
Compacted sample 16 cm
60 cm
Figure 1. Schematic layout of suction and volumetric water content measurement box.
Particle size characteristics Coarse sand (4.75-2 mm) Medium sand (2-0.425 mm) Fine sand (0.425-0.075 mm) Silt size (0.075-0.002 mm) Clay size (0.002 mm) Effective diameter (D10 mm) Uniformity coefficient (CU) Coefficient of curvature (CC) IS classification* Maximum void ratio (emax) Minimum void ratio (emin) *IS 1498: 1970.
The tensiometer used in this study can measure matric suction (m) up to 100 kPa and characterized by very short response time. The small dimension of the T5 tensiometer ensures minimal disturbance of the soil sample during its insertion. This electronic pressure transducer tensiometer is high resolution measuring device for the continuous measurement of water tension in the soil. The changes in water tension is captured electronically and converted to pressure units in hPa (hecto Pascal) or kPa. For logging the suction measurement automatically, the tensiometer is connected to a computer via. DL6 data logger. EC-TE capacitance-type sensors determine based on the dielectric constant or permittivity of the material in which they are inserted. The sensor works in the frequency range of 70 MHz. The basic principle behind the working of the sensor is the dielectric permittivity, , of soil medium changes with its water content. is dependent on the capacitance property of the soil mass. The sensor measures the capacitance property and gives the output in milliVolt (mV) (Decagon devices Operators Manual 2006). This mV output designated as Raw is converted to , based on the general calibration equation given by the manufacturer (Decagon Inc. USA). A detailed description of the technology of this sensor is reported in Bogena et al. (2007) and Kizito et al. (2008). 3 MATERIALS AND TEST METHODS
Part 14: 1983). The details of the characterization are listed in Table 1. It can be noted that the soil is poorly graded sand (SP) with no fine fraction. The air-dried sand was mixed with required amount of distilled water and compacted into the perspex mold of 12 cm diameter and 15 cm height, by giving different number of blows to achieve different packing dry density (d). The soil sample was then saturated by ponding it with distilled water and further removing the excess water on top. After saturating the soil sample, ECH2O-TE and T5 sensors were inserted into the packed soil sample. The soil sample was then allowed to air dry and the suction and volumetric water content of the sample continuously recorded using the respective data loggers. These data were used to develop SWCC of the soil. 4 RESULTS AND DISCUSSION
A locally available river sand designated as SA is used in this study for investigating the effect of initial water content and initial density on drying soilwater characteristic curve. The soil is characterized for its specific gravity, grain size distribution, and maximum and minimum void ratios by following the guidelines presented in the literature (IS 2720: Part III: Sec 2: 1980; IS 2720: Part 4: 1985; IS 2720:
The data obtained from the T5 and ECH2O-TE measurements were used to plot the SWCCs corresponding to different compaction states for the soil SA, as depicted in Figure 2. It can be noted from the figure that different initial paths are followed by the curves corresponding to different compaction state. The initial portion of the SWCC shifts upwards with the increase in initial compaction water content (w). Further, it can be noted that all measured SWCCs coincide for matric suction, m 10 kPa. Marinho & Stuermer (1999) have attributed this coinciding tendency of SWCC to the water located into pores of similar size/shape. One of the advantage of this study over those reported in the literature is that the volumetric water contents are measured ones and not estimated.
449
0.6 0.5 0.4 0.3

Continuous measurements w = 18.10 %, d = 1.59 gm/cm3 w = 18.10 %, d = 1.48 gm/cm3 w = 13.34 %, d = 1.31 gm/cm3 w = 8.34 %, d = 1.39 gm/cm3 w = 7.27 %, d = 1.31 gm/cm3
0.6
w = 18.10 %, d = 1.48 gm/cm3
0.5 0.4 0.3

w = 18.10 %, d = 1.59 gm/cm3

0.2 0.1 0.0 0 10
0.2 0.1
10
1
10
m(kPa)
10
0.0 100
101
m (kPa)
102
103
Figure 2. Continuous drying SWCCs for specimens at different initial water contents and initial densities.
0.6
w = 13.34 %, d = 1.31 gm/cm3
Figure 4. Influence of initial dry density on continuous drying SWCC.
0.5 0.4 0.3 0.2 0.1 0.0 100

w = 7.27 %, d = 1.31 gm/cm3
According to Birle et al. (2008), the tendency of the SWCCs to coincide at higher values of m may be due to the unimodal pore size distribution. However, further investigations are required to ascertain such possibility. 4.2 Effect of initial density on SWCC Figure 4 shows the comparison of continuous drying SWCCs for two soil samples at same initial compaction water content but at different initial dry density. It can be noted that both the drying SWCCs matches perfectly. Therefore, it can be summarized that the initial dry unit weight of soil SA does not have significant influence on the SWCCs. This observation is in agreement with the results reported by Birle et al. (2008). 5 CONCLUSIONS
101
m (kPa)
102
103
Figure 3. Influence of initial water content on continuous drying SWCC.
4.1
Effect of initial water content on SWCC
Figure 3 depicts the SWCCs of the samples at similar compaction dry density but with different initial water contents. From the figure it can be observed that there is an initial difference in both, which gradually reduces and converge at m close to 10 kPa. At m greater than 10 kPa, both the curves matches well. Therefore, it can be summarized that the initial water content influences the initial portion of drying SWCCs. In the high m range the SWCC is independent of the initial compaction conditions. This observation is in agreement with studies reported in the literature (Tinjum et al. 1997; Vanapalli et al. 1999; Tarantino & Tombolato 2005; Birle et al. 2008).
The study deals with an investigation on the effect of initial water content and initial density on the continuous drying Soil-Water Characteristic Curve (SWCC) of a sandy soil. Continuous drying soilwater characteristic curves were obtained using a T5 tensiometer and ECH2O-TE/EC-TM volumetric water content sensor. The study indicates that the initial dry density has negligible influence on SWCCs obtained using continuous measurement. However, it has been noted that the initial compaction water content influences initial potion of continuous SWCC close to full saturation. Such an influence becomes negligible towards higher matric suction.
450
REFERENCES
Barbour, S.L. 1998. Nineteen Canadian geotechnical colloquium: The soil-water characteristic curve: A historical perspective. Canadian Geotechnical Journal 35: 873894. Birle, E., Heyer, D. & Vogt, N. 2008. Influence of the initial water content and dry density on the soil-water retention curve and the shrinkage behavior of a compacted clay. Acta Geotechnica 3: 191200. Bogena, H.R., Huisman, J.A., Oberdorster, C. & Vereecken, H. 2007. Evaluation of a low-cost water content sensor for wireless network applications. Journal of Hydrology 344: 3242. Box, J.E. & Taylor, S.A. 1962. Influence of soil bulk density on matric potential. Soil Science America Proc. 26: 119122. Brooks, R.H. & Corey, A.T. 1966. Properties of porous media affecting fluid flow. Journal of the Irrigation and Drainage Division 92: 6189. Campbell, G.S. & Gardner, W.H. 1971. Psychrometric measurement of soil water potential: temperature and bulk density effects. Soil Science Society of America Proc. 135: 812. Croney, D. & Coleman, J.D. 1954. Soil structure in relation to soil suction (pF). Journal of Soil Science 5(1): 7584. Decagon Devices, Operators manual version 2 2006. ECH2OTE/ECTM. Water Content, EC and Temperature Sensor Manual. Decagon Devices: WA99163. Erzin, Y. & Erol, O. 2007. Swell pressure prediction by suction methods. Engineering Geology Soils. 92: 133145. Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics for unsaturated soils. John Wiley and Sons, Inc.: New York. Fredlund, D.G., Xing, A. & Huang, S. 1994. Predicting the permeability function for unsaturated soils using the soil-water characteristic curve. Canadian Geotechnical Journal 31: 533546. Hillel, D. 1980. Fundamental of soil physics. Academic Press, Inc.: San Diego: Calif. Houston, W.N., Dye, H.B., Zapata, C.E., Perera, Y.Y. & Harraz, A. 2006. Determination of swcc using one point suction measurement and standard curves. Unsaturated Soils 2006. IS 1498: 1970. Indian Standard on classification and identification of soils for general engineering purposes. Bureau of Indian Standards. New Delhi. IS 2720: Part III: Sec 2: 1980. Indian Standard on test for soilsPart III: determination of specific gravity-sec 2: fine, medium and coarse grained soils. Bureau of Indian Standards. New Delhi. IS 2720: Part 14: 1983. Indian Standard on methods of test for soils-part 14: determination of density index (relative density) of cohesionless soils. Bureau of Indian Standards. New Delhi. IS 2720: Part 4: 1985. Indian Standard on methods of test for soils-part 4: grain size analysis. Bureau of Indian Standards. New Delhi.
Kizito, F., Campbell, C.S., Campbell, G.S., Cobos, D.R., Teare, B.L., Carter, B. & Hopmans, J.W. 2008. Frequency, electrical conductivity and temperature analysis of a low-cost capacitance soil moisture sensor. Journal of Hydrology 352: 367378. Krahn, R.L. & Fredlund, D.G. 1972. On total, matric and osmotic suction. Soil Science 114: 339348. Li, X. & Zhang, L.M. 2009. Characterization of dualstructure pore-size distribution of soil. Canadian Geotechnical Journal 46(2): 129141. Li, J.Y., Yang, Q., Li, P.Y. & Yang, Q.L. 2009. Experimental research on soil-water characteristic curve of remolded residual soils. Electronic Journal of Geotechnical Engineering Bundle L 14: 112. Marinho, F.A.M. & Chandler, R.J. 1993. Aspects of the behavior of clays on drying. Unsaturated Soils. ASCE Geotechnical Special Publication 39: 7790. New York. Marinho, F.A.M. & Stuermer, M.M. 1999. The influence of the compaction energy on the SWCC of a residual soil. Advances in Unsaturated Geotechnics-Geotechnical Special Publication 10: 125141. Miao, L., Jing, F. & Houson, S.L. 2006. Soil- water characteristic curve of remolded expansive soils. Unsaturated Soils 2006: 9971004. Ng, C.W.W. & Pang, Y.W. 2000. Influence of stress state on soil-water characteristics and slope stability. Journal of Geotechnical and Geoenvironmental Engineering 126(2): 167188. Olson, R.E. & Leonard, J.L. 1965. Pore water pressures in unsaturated soils. Journal of the Soil Mechanics and Foundation, ASCE 91: 127150. Sreedeep, S. & Singh, D.N. 2005. A study to investigate influence of soil properties on its soil suction. Journal of Testing and Evaluation 33: 16. Sun, D.A., Sheng, D.C., Cui, H.B. & Li, J. 2006. Effect of density on the soil-water retention behaviour of compacted soil. Unsaturated Soils 2006: 13381347. Sun, D., Sheng, D. & Xu, Y. 2007. Collapse behaviour of unsaturated compacted soil with different initial densities. Canadian Geotechnical Journal 44: 12521264. Tarantino, A. & Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Geotechnique 55(4): 307317. Tinjum, J.M., Benson, C.H. & Blotz, L.R. 1997. Soil-water characteristic curves for compacted clays. Journal of Geotechnical and Geoenvironmenal Engineering 11: 10601069. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Cliffton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction. Canadian Geotechnical Journal 33: 379392. Vanapalli, S.K., Fredlund, D.G. & Pufahl, D.E. 1999. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Geotechnique: 49(2): 143159. Yang, H., Rahardjo, H., Leong, E.C. & Fredlund, D.G. 2004. Factors affecting drying and wetting soil-water characteristic curves of sandy soils. Canadian Geotechnical Journal 41: 908920.
451
A transfer function of a soil water characteristic curve model for sands

Emke Imre
Szent Istvn University and BME Geotechnical Department, Budapest, Hungary
Kalman Rajkai
Hungarian Academy of Sciences, Research Institute for Soil Science and Agricultural Chemistry, Budapest, Hungary
R. Genovese & C. Jommi
Politecnico di Milano, Milano, Italy
ABSTRACT: A transfer function describes a relation between the grading curves and a physical property of the soils. Its direct interpolation needs exponentially many data in terms of the number of the fractions in the grading curve N. A methodbased on the grading entropy concept, a double interpolation technique and c N 2 data as a maximumis used to interpolate a transfer function between the parameters of the Fredlund-Xing soil water characteristic curve equation and the grading curve for well-graded sands in this paper. The method of the measurement and the measured data are also briefly presented. 1 INTRODUCTION first stage of a three-stage research (Imre et al., 2003 to 2008). The detailed validation including the data of the second and third stages will follow in some subsequent papers. 2 GRADING ENTROPY
The soil water-retention curve and the hydraulic conductivity function are the primary input for the assessment of infiltration processes in unsaturated soils. The soil water-retention curve is often estimated rather than measured. It is generally approximated by some parametric functions and the hydraulic conductivity function is generally derived from this. A transfer function describes the relation between the grading curves and a given physical property (function) of the soils. The construction of a transfer function with direct interpolation needs exponentially many data in terms of the number of the fraction in the grading curve N. which is too many. To overcome this difficulty, a general transfer function construction method is suggested with a double approximation technique, based on the grading entropy concept, needing maximum c N2 data which can be decreased to c N (Imre et al., 2009a). The method is applied to the parameters of the van Genuchten water retention curve equation of sands in a previous paper. The first results can be used for a class of sands with basically continuous grading curve (Imre et al., 2009b). In this paper the method is applied to the parameters of the Fredlund-Xing (1994) soil water characteristic curve equation. The measured water-retention curve data are originated from the
The grading entropy concept of Lrincz (1986) is presented and completed by some comments among othersconcerning the transfer function construction method and the meaning of the grading entropy coordinates (Imre et al., 2009a). 2.1 Statistical cells for the grading entropy
The grading entropy concept of Lrincz (1986) is a kind of statistical entropy (Korn and Korn, 1975). Two statistical cell systems are used. The larger statistical cells (fractions) are defined by successive multiplication with a factor of 2, starting from an arbitrary elementary cell width d0 on the pattern of the classical sieve hole diameters. For convenience, the height of the SiO4 tetrahedron (222 mm, Imre, 1995) is adopted here. The relationship between the diameter range and the fraction absolute serial number j ( j 1, 2 ... see Table 1): 2jd0 d 2j1d0, 2 j22 mm d 2j23 mm (1)
where d0 is the elementary cell width.
453
Table 1.
Fraction data. 1 222221 0 ... 23 12 22 24 24 23 25 48 24
Fraction number j d limits in [mm] S0i []
where the relative frequency of the elementary cell in fraction i is:
i
xi , Ci
(7)
The smaller cell system is defined with equal width of d0. The number of the elementary cells Ci in fraction i is equal to: Ci 2i d0 2i 1 d0 2i 1. d0 (2)
where xi is the relative frequency of fraction i and Ci is the number of the elementary cells in fraction i. From this, the grading entropy of the k-th fraction S0k is (Table 1) is the serial number minus 1: Sok lnCk ln 2 k 1 k 1. ln 2 ln 2 (8)
2.2
Simplex representation of the grading curves
The variable called number of the fractions N is introduced as the difference of the serial numbers of the finest and coarsest fractions: N jmax jmin 1. (3)
Being directly related to the particle size in fraction k. The grading entropy of a mixture of the fractions can be split as follows: S S0 S, (9)
The base entropy So is the mean fraction entropy S0i, being related to the mean particle size: S0 xi S0i .
xi 0
A space of the possible grading curves consists of such grading curves that have a specified N value and a specified minimum grain diameter dmin value (being larger than or equal to d0) moreover, the distribution is uniform within a fraction. Concerning the fractions, N 1 relative frequencies xi (i 1, 2, 3, ..., N) are independent as follows: xi 1,
i1 N
(10)
The entropy increment S is specific statistical entropy of the grading curve in terms of the fractions: S 1 xi ln xi , xi 0. ln 2 xi 0 (11)
xi 0,
N 1.
(4)
This equation defines an N 1 dimensional closed simplex. There is a one-to-one relationship between the points of an N 1 dimensional closed simplex and a space of the possible grading curves. The simplex is frequently used in the description of the grading entropy concept. 2.3 Grading entropy A possible definition of the specific entropy (Korn and Korn, 1975): 1 i ln i ln 2 i 0 (5)
The relative base entropy A is equal to: xi (Soi So min ) N 1

N
A = i 1
i 1
xi (i 1) N 1
(12)
and the normalized entropy increment B is equal to: B S ln N (13)
2.4 Grading entropy map, entropy diagram The entropy map [So, S] or the normalized entropy map [A,B] is defined between the N 1 dimensional open simplex and the two dimensional space of the (normalized) grading entropy coordinates, for a given value of N and dmin. These maps are continuous on the open simplex and can continuously be extended to the closed simplex for a given value of N. The image of the simplexthe entropy diagram (Fig. 1)is compact.
where i is the relative frequency of the cell i. The grading entropy S is defined as the specific statistical entropy of the grading curve in terms of the elementary statistical cells for the open simplex: S x x 1 Ci i ln i ln 2 i 0 Ci Ci (6)
454
The constant multiplied form of the normalized entropy increment B is the following logarithm of the generalized weighted geometrical mean of the relative frequencies xi (i 1, 2, 3, ..., N), denoted by B0: B0 1 1 xi ln xi ln xi xi N ln N xi 0 N ln N xi 0 (14) The B function is concave with a maximum at the simplex gravity centre where the curve length is a minimum in semilog scale. The grading curve with maximum B for a specified A and N is also with minimum curve length in semilog scale since its distribution is fractal. It follows that optimal grading curve with a maximum B for a fixed A and N is a mean point of the constant A section of the simplex. Since the grading curves with constant A and N have the same sub-graph area, deviating from each-other in such a way that equal areas are found below and above, it follows that the grading curves with constant B are with about the same curve length on the constant A section of the simplex. The distance of two grading curve in terms of the absolute value of the difference of an entropy coordinate is a pseudo-metric. The A is a distance from the finest fraction where A is zero and the Bmax B is a distance from the foregoing mean point of the constant A section of the simplex. 2.6 The transfer function construction method The transfer function F: K is defined between the N 1 dimensional simplex (defined by N and dmin) and the one dimensional space of a soil parameter K. In the suggested method the transfer function is constructed in two steps. First a mean transfer function is interpolated over the nonnormalised entropy diagram using the data of some optimal points which map into the maximum S lines. Then the mean transfer function is extended into the inverse image of the entropy diagram points. In the simplest possible case the extension is made with the constant function, the same soil physical value is considered for those grading curves that map into the same non-normalized grading entropy diagram point. This method implies the following assumption. The maximum S linesthe image of the optimal lines with various N valuesmay intersect eachother. It is assumed, that the scalar values related
(a)
2.8
N=7 N=6 N=5 N=4 N=3
maximum lines minimum lines
Entropy increment,S [-]
2.4 2.0 1.6 1.2 0.8 0.4 0.0
0 .0
2 .0 4 .0 6 .0 Base entropy, So [-]
8 .0
(b)
Figure 1. Entropy diagrams. (a) Normalized diagrams for simplexes with N 1 7. (b) Non-normalized diagram for a simplex with N 7, S0 min 1 representing its continuous sub-simplexes with N 1 7.
There exists a minimum and a maximum value of B for every specified A. The minimum line of the normalized entropy diagram is generally approximated by the image of edge 1 N (Fig. 1). A single grading curveoptimal grading curvemaps into the maximum B for a given A which can be determined by conditional optimisation. The maximum B value for a fixed A is a critical value of the entropy map. The inverse image of a regular value or entropy diagram point is an N 3 dimensional manifold, being centred to the optimal point. 2.5 The meaning of the entropy coordinates
The base entropy So and the relative base entropy A are the weighted mean of the (reduced) fraction entropies, which are one-to-one related to the fraction diameter. The value of A expresses the proportion of the larger grains. A is a kind of stability number (i.e. if A 2/3 then the structure of the soil is stable, Lrincz, 1986). The grading curves with the same A and N have the same sub-graph area in semilog scale (Imre et al., 2009a).
455
to the nearly coinciding points of the maximum S lines are about the same. It was shown that the so determined simplest possible (preliminary) density transfer function can be used for the determination of the dry density of a class of sands with basically continuous grading curve (Imre et al., 2009a). 3 3.1 MEASUREMENTS Mixtures used in the research program
Table 2. Fraction 1 2 3 4 5 6 7
Fractions. Grain size d (mm) 0.030.06 0.060.125 0.1250.25 0.250.50 0.501.0 1.02.0 2.04.0
The fractions (see Figs. 2, 3 and Table 2) are numbered consecutively from the finest one. The optimal A 2/3 mixtures measured in the first stage of the research are shown in Table 3. 3.2 Methods
Table 3. Two digit notation for the mixtures (NJ) related to a specified A, N is equal to 2 to 7 (see the fractions in Table 2). Fraction number N 2 Largest fraction J 2 3 4 5 6 7 3 4 5 6 7 4 5 6 7 5 6 7 6 7 7 Mixture notation NJ 22 23 24 25 26 27 33 34 35 36 37 44 45 46 47 55 56 57 66 67 77 Fractions in the mixture 1-2 2-3 3-4 4-5 5-6 6-7 1-2-3 2-3-4 3-4-5 4-5-6 5-6-7 1-2-3-4 2-3-4-5 3-4-5-6 4-5-6-7 1-2-3-4-5 2-3-4-5-6 3-4-5-6-7 1-2-3-4-5-6 2-3-4-5-6-7 1-2-3-4-5-6-7
The retention curve measurement methods of the Soil Science Institute (RISSAC) are the sand box method shown in Figure 4 and the pressure membrane extractor methods, as shown in Figure 5 (Vrallyay, 1973; Rajkai, 1993). In the smaller suction range of ua uw 50 kPa sand boxes are used and, the suction is applied by water
gravel sand Mo
100 80 60
fraction 1 6 5 4 3 2
S (%)
40 fraction 7 20 0
5 6
1.00 0.10 0.01
10.00
d [mm]
Figure 2.
The theoretical (ideal) fractions.

gravel sand
Mo
fraction 1 6 5 4 3 2
S (%)
fraction 7
10.00
1.00
0.10
0.01
d [mm]
Figure 3.
The fractions used in the preliminary tests.
Figure 4.
The RISSAC sand box method.
456
w wr
ws wr
n ua uw ln e a
(15)
where w is water content (wr residual, ws saturated), ua uw is suction, a, n, m are non-linearly dependent model parameters. The merit function to be minimized: wme (ui ) w(ui , p ) L max(w me (ui ))
i L
F ( p)
i 1
(16)
Figure 5. method.
The RISSAC pressure membrane extractor
Table 4. RISSAC method, low suction range. Load (cm) 1 2.5 10 31.5 100 200 501 Semi-permeable membrane Sand Sand fine sand fine sand kaolinit kaolinit kaolinit Suction application gravitational gravitational gravitational gravitational gravitational gravitational vacuum
Table 5. RISSAC method, high suction range. Load (cm) 1000 5000 15000 Method of suction application pressure membrane extractor pressure membrane extractor pressure membrane extractor
where wme is measured, w(ui, p) is the computed water content, p [wr, ws wr a, n, m] is parameter vector, ui, (i 1 ... L) are the applied suction load steps. The solution of the non-linear inverse problem is always approximate partly due to the noise, partly due to the numerical solution techniques. A mathematically precise, automatic, non-linear inverse problem solver was elaborated and used together with some reliability criteria and testing methods (geometric and analytic). These are described in (Imre, 1996 and in Imre et al., 2010) and can be summarized as follows. A sub-minimisation method was used for to the non-linear inverse problem solution, to decrease the number of the parameters by the number of the linearly dependent parameters. It was tested whether the solution was unique and precise enough.
RESULTS AND CONCLUSION
pressure decrease. In the greater suction range (ua uw 50 kPa) pressure membrane extractor is used and, the suction is applied by the axis translation technique. In these methods, the load steps are fixed, as given in Tables 4 and 5. The stage duration is greater than about 6 days. In the preliminary measurements, two sample sizes were used (big: 5 cm tall and, small: 2.5 cm tall). Duplicate measurements were performed on all fractions and mixtures. 4 MODEL FITTING
The Fredlund-Xing water retention curve equation is as follows (1994):
The five parameters of the Fredlund-Xing (1994) water retention curve equation [wr, ws wr, a, n, m] and the standard deviation of these were identified for the measured water retention curves. In this work only the data of the soils tested in the first stage were used, being indicated in Table 3 and by big circles in Figures 6 to 8. On the basis of the results, a tentative, preliminary transfer function was determined for the nonlinearly dependent parameters a, n and m of the Fredlund-Xing model. These are given by some iso-lines as it is shown in Figures 6 to 8. It can be noted that the goodness of the model fit was not very nice around fraction 4 and mixture 77 indicating some limitations of the model in the case of sands (Figs. 9 to 12). Therefore, for example the results in between the iso-lines 2 in Figure 7 are approximate. Similar model fitting experience is reported for sands Agus et al. (2003).
457
2.8
77
Entropy increment, S [-]
1 2 5 7
0.8 measured Van Genuchten Fredlund-Xing
2.4 2.0 1.6 1.2 0.8 0.4 0.0

4
v [-]
0.4 0.0 0 .0 1
N=2
0 .1 0
1 .0 0
1 0.0 0
10 0 .0 0
1 0 0 0 .0 0
[kPa]
0 .0
Figure 6.
2 .0 4 .0 6 .0 Base entropy, S0 [-]
8 .0
Figure 9.
Measured and the fitted data for fraction 4.
Transfer function for parameter n.
1.0 0.8 0.6 measured Frelund-Xing van Genuchten
2.8
N=7 N=6 N=5 N=4 N=3 N=2 1
2 2 1
v [-]
0.4 0.2 0.0 0 .0 1
2.4 2.0 1.6 1.2 0.8 0.4 0.0
0 .1 0
1 .0 0
1 0 .0 0
1 0 0 .0 0
1 0 0 0 .0 0
[kPa]
Figure 10.
1.0 0.8
Measured and the fitted data mixture 55.
measured Fredlund-Xing van Genuchten
0 .0
Figure 7.
v [-]
0.4 0.2 0.0 0 .0 1
2 .0 4 .0 106 6 .0 Base entropy, S0 [-]
8 .0
0.6
Transfer function for parameter m.
0 .1 0
1 .0 0
1 0 .0 0
1 0 0 .0 0
1 0 0 0 .0 0
[kPa]
2.8
N=7
10
Figure 11.
1.0
2.4 2.0 1.6 1.2 0.8 0.4

0.2
0,5 20
0.8 0.6 v [-] 0.4 measured Frelund-Xing van Genuchten
N=2
0.0
0 .0
Figure 8.
2 .0 4 .0 6 .0 Base entropy, S0 [-]
8 .0
0.0 0 .0 1
0 .1 0
1 .0 0 1 0 .0 0 ua-uw [kPa]
1 0 0 .0 0
1 0 0 0 .0 0
Transfer function for parameter a.
Figure 12.
458
The so determined transfer functions can theoretically be used to estimate the water retention curves for well-graded sand mixtures. This outcome will be tested on the basis of the additional data related to the soils tested in the second and third stages of research later on (indicated by small circles in Figures 6 to 8) and, by increasing the precision of the inverse problem solution later on. It is planned that the proposed method is applied to find a preliminary transfer function for the parameters of some other water retention curve models, also in the frame of some further research. ACKNOWLEDGEMENT The support of the National Research Fund Jedlik nyos NKFP B1 2006 08 and the Norwegian research fund HU-0121 was used for this research which is greatly acknowledged. REFERENCES
Agus, S.S., Leong, E.C. & Schanz, T. 2003. Assessment of statistical models for indirect determination of permeability functions from soil-water characteristic curves. Geotechnique, 53(2): 279282. Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater characteristic curve. Canadian Geotechnical Journal, 31: 521532. Imre, E. 1995. Characterization of dispersive and piping soils. Proc. Of XI. ECSMFE, Copenhagen, 2: 4955. Imre, E. 1996. Inverse problem solution with a geometrical method. Proc. of the 2nd Int. Conf. on Inverse Problems in Engineering. Le Croisic, France: 331338. Imre, E., Rajkai, K., Genovese, R., Jommi, C., Lrincz, J., Aradi, L. & Telekes, G. 2003. Soil water-retention curve for fractions and mixtures. Proc. of UNSATASIA, Osaka: 451456. Imre, E., Havrn, K., Lrincz, J., Rajkai, K., Firgi, T. & Telekes, G. 2005. A model to predict the soil water characteristics of sand mixtures. Int. Symp. on Advanced Experimental Unsat. Soil Mech, Trento June 2729: 359368.
Imre, E., Rajkai, K., Firgi, T., Trang, Q.P. & Telekes, G. 2006. Closed-Form Functions For The Soil Water-Retention Curve of Sand Fractions and Sand Mixtures The Fourth Int. Conf. On Unsaturated Soils, Arizona: 24082419. Imre, E., Laufer, I., Trang, Q.P., Lrincz, J., Rajkai, K., Firgi, T. & Telekes, G. 2007. The soil water characteristics of 2-component sand mixtures 2nd International Conference Mechanics of Unsaturated Soils, 2007 Weimar 2: 4559. Imre, E., Laufer, I., Rajkai, K., Scheuermann, A., Firgi, T. & Telekes, G. 2008. The soil functions of sand mixtures. Proc. of the 1st European Conference on Unsaturated Soils, Durham, UK, Proceedings: 299305. Imre, E., Lrincz, J., Trang, Q.P., Fityus, S., Pusztai, J., Telekes, G. & Schanz, T. 2009a. Some dry density transfer function for sands. Invited paper. KSCE Journal of Civil Engineering 13(4): 257272. DOI 10.1007/s12205-009-0257-7. Imre, E., Rajkai, K., Genovese, E. & Fityus, S. 2009b. The SWCC transfer functions of sands. Proc. of 4th Asia Pacific Conference on Unsaturated Soils 2325 November 2009, Newcastle, Australia: 791797. Imre, E., Rzsa, P., Bates, L., Fityus, S. 2010. Evaluation of monotonous and non-monotonous dissipation test results. COGE (Computers and Geotechnics, accepted). Korn, G.A. & Korn, T.M. 1975. Mathematical Handbook for Scientists and Engineers 2nd. Edition, McGraw-Hill Book Company. Lrincz, J. 1986. Grading entropy of soils. Doctoral Thesis, Technical Sciences, TU of Budapest (in Hungarian). Rajkai, K. 1993. A talajok vzgazdlkodsi tulajdonsgainak vizsglati mdszerei Bzs I.(ed): Talaj-s agrokmiai vizsg. INDA4321 Kiad, Bp. 115160. Vrallyay, Gy. 1973. A talajok nedvessgpotencilja s j berendezs annak meghatrozsra az alacsony tenzitartomnyban. Agrokmia s Talajtan 22: 122. van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44: 892898.
459
Prediction of soil-water characteristic curve using neural network

A. Johari
Department of Civil and Environmental Engineering, Shiraz University of Technology, Shiraz, Iran
A.A. Javadi
Computational Geomechanics Group, School of Engineering, Mathematics and Physical Sciences, University of Exeter, Exeter, UK
ABSTRACT: In this paper, a neural network is employed to predict the soil water characteristic curve of unsaturated soils. The network has five input neurons, namely, initial void ratio, initial gravimetric water content, logarithm of suction normalized with respect to the atmospheric pressure, clay fraction, and silt content. The network has five neurons in the hidden layer and the only output neuron is the gravimetric water content corresponding to the assigned input suction. Results from pressure plate tests carried out on clay, silty clay, sandy loam, and loam compiled in SoilVision software, are used for training and testing of the network. After digitization of the data, a computer program coded in Matlab is developed and used for the analysis. The neural network simulations are compared with the experimental results as well as those of a number of models proposed in the literature. The results indicate the superior performance of the proposed method in predicting the soil water characteristic curve. 1 INTRODUCTION such as specific heat, permeability, and thermal conductivity can be related to the soil water characteristic curve (Fredlund and Rahardjo, 1993). In this paper a neural network approach is proposed for estimating the soil water characteristic curve using basic soil properties such as grain size distribution, initial void ratio, initial water content, and logarithm of suction normalized with respect to atmospheric air pressure.
Limitations in describing the mechanical behavior of unsaturated soils based on a single effective stress equation, similar to the one proposed by Bishop and Donald (1961), has led to developing different approaches for modeling the observed behavior of these soils. The engineering behavior of unsaturated soils can be interpreted in terms of two different stress states, namely a net normal stress ( ua), and matric suction (ua uw). Soil Water Characteristic Curve (SWCC), also known as soil water-retention curve, is an important part of any constitutive relationship for unsaturated soils. SWCC for an unsaturated soil defines the relationship between water content and the corresponding suction. SWCC can be considered as a continuous sigmoidal function describing the water storage capacity of a soil as it is subjected to various soil suctions. SWCC includes important information about the amount of water contained in the pores at any soil suction and the pore size distribution corresponding to the stress state in the soil. SWCC and the unsaturated coefficient of permeability functions are required for solving transient water and solute transport problems associated with the vadose zone. Different aspects of unsaturated soil behavior such as shear strength, volume change, diffusivity, and absorption as well as most of soil properties
CURRENT METHODS FOR DETERMINATION OF SWCC
A number of methods are available for predicting the SWCC of a particular soil. These are based on grain size distribution and other soil properties. These methods can be classified into four major groups described below: 1. In the first group, water content at a given suction is correlated to the soil properties such as grain size distribution and porosity. Generally, in this process a regression analysis is required followed by a curve fitting procedure. Among the important contributions to this group are the works of Visser (1969), Gupta and Larson (1979), Rawls and Brakensiek (1982), Hutson and Cass (1987), Williams and Ahuja (1992),
461
Reddi and Poduri (1997), Tinjum et al. (1997), Mbagwu and Mbah (1998). 2. Methods in the second group propose correlations between parameters of an assumed empirical equation for SWCC and the basic soil properties such as grain size distribution and dry density using a regression analysis. Among important contributions are the works of Williams et al. (1983), Cresswell and Paydar (1996), Tomasella and Hodnett (1998). 3. The third group involves physico-empirical modeling of SWCC. In this approach the grain size distribution is converted into a pore size distribution which in turn is related to the distribution of water content and the associated pore pressure. Among the important contributions. 4. Artificial intelligence methods such as neural network, genetic programming, and other machine learning methods have recently formed the fourth group. No significant attempt has been cited on this subject in the literature so far. In the following sections, the fundamentals of the neural network approach are described and its application in determination of SWCC is illustrated.
Figure 1.
A typical ANN (Ghaboussi et al. 1991).
ARTIFICIAL NEURAL NETWORK
Neural Network (NN) is a computer-based modeling technique for computation and knowledge representation inspired by the neural architecture and operation of the human brain. NNs have experienced a considerable resurgence of interest in recent years, though they have been initially developed during the early 1940s. Artificial Neural Network (ANN) is constructed directly from experimental data. This is a fundamentally different approach to modeling of the material behavior, and because of their ability to learn and generalize interactions among many variables, ANNs have the potential to model various aspects of material behavior. The basic architecture of ANN has been covered by Rumelhart and McClelland (1988). NN consists of a large number of highly interconnected processing units. Each processing unit (neuron), acting as an idealized neuron in human brain, receives input from the units to which it is connected, computes an activation level, and transmits that activation to other processing units (Figure 1). A multi-layer perceptron NN has an input layer, an output layer, and a number of hidden layers connected to each other as illustrated in Figure 1. Weights are assigned to the connections between these units. The presence of hidden layers allows the networks to represent and compute more complicated associations between input and output patterns.
A multi-layer feed forward ANN must be trained first and tested afterwards. During training, the weights are adjusted in an iterative process. The propagation of activation takes place in a feedforward manner, from input layer to output layers. Conventionally, error (the difference between the network output and their target values) is backpropagated through the network and the weights are adjusted using a gradient descent rule. With the successful completion of the training, the iterative process reduces the error measure to a minimum and the set of connection weights is captured. Once training is completed, the neural network is ready to be used. When presented with an input pattern, a feed-forward network computation results in an output pattern that is the result of the generalization and synthesis of what it has learned and stored in its connection weights.
DATABASE
Results from pressure plate tests on clay, silty clay, sandy loam, and loam soil reported by different researchers and compiled by SoilVision (2002) were adopted for the analysis. Table 1 indicates the range of basic soil properties adopted for this study. This database consists of the results from 186 pressure plate tests together with their grain size distributions. Final suction values were mostly in the range of 800 to 1700 kPa with few tests having suction values as large as 105 kPa. The results reported on these specimens were digitized to obtain the necessary database. The database thus developed had a total number of 2694 patterns. For normalization, each component of the data set was normalized to lie in the interval [1,1].
462
Table 1. Range of basic soil properties of the specimens (SoilVision 2002).

Actual Gravimetric Water Content
1.0
R2 = 0.94
0.8
Void ratio: 0.4582.846 Specific gravity: 2.282.92 Dry density (kg/m3): 7021811 Water content range (%): 0.1898.27 Initial water content (%): 17.34105.41
Suction range (kPa): 0.2104857.6 Clay (0.005 mm) (%): 4.476.7 Silt (0.005 mm 0.075 mm) (%): 10.387.5 Sand (0.075 mm) (%): 0.155.3
0.6
0.4
0.2
0.0 0.0 0.2 0.4 0.6 0.8 1.0
Predicted Gravimetric Water Content
Figure 3. Measured gravimetric water contents vs those predicted by the ANN model for the training data.
1.0
R2 = 0.91 Actual Gravimetric Water Content

0.8
0.6
0.4
Figure 2. The proposed neural network for the prediction of SWCC.
0.2
APPLICATION OF ANN FOR PREDICTION OF SWCC
0.0 0.0 0.2 0.4 0.6 0.8 1.0
A computer program coded in Matlab was developed for training the network. Five parameters, namely, void ratio, initial water content, logarithm of suction normalized with respect to the atmospheric pressure [log(ua uw)/pa], clay fraction, and silt content were selected as the input neurons. The output neuron yields the gravimetric water content corresponding to the assigned input suction. To find the optimum number of hidden neurons, they were decreased from a maximum of 15 neurons while checking the error measure of the network. This resulted in a total number of five neurons plus a bias neuron for the hidden layer. The optimum neural network structure is shown in Figure 2. 6 6.1 RESULTS AND DISCUSSION Model training and testing
Figure 4. Measured gravimetric water contents vs those predicted by the ANN model for the testing data.
used for model development and the remaining 800 patterns were allocated for testing. To find the optimum network, the sum square of differences between the predicted water content and the actual water content was monitored. Figures 3 and 4 compare the predicted gravimetric water contents with the actual data for training and testing datasets, respectively. These figures show a good correlation between the predictions made using the ANN model and the actual data for both training and testing data. 6.2 Comparison with previous works
From the 186 pressure plate tests used in this study, 131 tests were used for training of the network and the remaining 55 tests were used for testing. After digitization of the test results, from a total number of 2694 patterns generated, 1894 patterns were
As stated above, a number of methods have been presented by different investigators for estimating SWCC. Among these methods, the approach presented by Fredlund et al. (1997) is considered to give a more reasonable estimate of the SWCC (Zapata et al., 2003 and Fredlund et al., 2002). Therefore in this paper, the proposed model is
463
compared to the approach proposed by Fredlund et al. (1997). Fredlund et al. (1997) estimated SWCC from the grain size distribution curve and volume mass properties. Figures 5 and 6 compare the measured gravimetric water contents with those predicted by Fredlund et al. (1997) approach for the training and testing data, respectively. Table 2 presents the errors in the neural network predictions compared with those of the Fredlund et al approach. In this Table, the mean sum square of the error is defined as: MSSE 1 N (A i Pi )2 N i 1
From Table 2 it can be concluded that the proposed ANN approach has a good capability to simulate the SWCC for soils. In order to show the robustness of the proposed method, the simulation results were also compared
Table 2. Comparing the performances of the ANN and Fredlund et al. (1997) approaches. Training data MSSE Neural network Fredlund et al. Testing data MSSE Neural network Fredlund et al. 0.0018 0.0047 R2 0.91 0.89 0.0013 0.0071 R2 0.94 0.85
where: Ai Actual value for data point i. Pi Predicted value for data point i. N Total number of data points in the database.
1.0
0.3
0.25 Gravimetric Water Content

0.8
Experiment NN Prediction Fredlund et al.
0.2
0.6
0.15
0.4
0.1
R2 = 0.85
0.05
0.2
0 0.1
0.0 0.0 0.2 0.4 0.6 0.8 1.0
10
100
1000
10000
Soil Suction (kPa)
Figure 5. Measured gravimetric water contents versus those predicted by the Fredlund et al. approach for the training data.
1.0
Figure 7. Best simulation results among the tests used for training of the ANN model (void ratio: 0.672, initial water content: 25.46%, clay content: 38.75%, silt content: 61.1%).
0.35 0.3
0.8
Gravimetric Water Content
0.25 0.2 0.15 0.1 0.05 0
0.6
0.4
R2 = 0.89
0.2
0.0 0.0 0.2 0.4 0.6 0.8 1.0
0.1
10
100
1000
10000
Soil Suction (kPa)
Figure 6. Measured gravimetric water contents vs those predicted by the Fredlund et al. approach for the testing data.
Figure 8. Average simulation results among the tests used for training of the ANN model (void ratio: 0.78, initial water content: 29.31%, clay content: 28.39%, silt content: 69.77%).
464
0.7 0.6 Gravimetric Water Content 0.5 0.4 0.3 0.2 0.1
0.4
0.35 0.3 0.25 0.2 0.15 0.1 0.05 0
0 0.1 1 10 100 1000 10000 100000 1000000 Soil Suction (kPa)
0.1
10
100
1000
10000
100000
1000000
Soil Suction (kPa)
Figure 9. Worst simulation results among the tests used for training of the ANN model (void ratio: 1.75, initial water content: 64.28%, clay content: 52.72%, silt content: 37.19%).
0.35 0.3
Figure 11. Average simulation results among the tests used for testing of the ANN model (void ratio: 0.95, initial water content: 35.99%, clay content: 37.68%, silt content: 39.41%).
1
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0
0.25 0.2 0.15 0.1 0.05 0 0.1 1 10 100 1000 10000
0.1
10
100
1000
10000
Soil Suction (kPa)
Soil Suction (kPa)
Figure 10. Best simulation results among the tests used for testing of the ANN model (void ratio: 0.77, initial water content: 29.17%, clay content: 19.03%, silt content: 80.30%).
Figure 12. Worst simulation results among the tests used for testing of the ANN model. (void ratio: 2.51, initial water content: 93.0%, clay content: 47.61%, silt content: 51.28%).
on a one-to-one basis. Figures 7 to 9 show SWCCs for 3 specimens used in training of the model. In these figures, the ANN predictions, and those of the method suggested by Fredlund et al. (1997) are compared with the experimental results. From these figures, it may be concluded that the proposed ANN has a good potential for predicting SWCC with reasonable degree of accuracy and the results of the ANN model are considerably more accurate than those of the traditional methods. Similarly, Figures 10 to 12 present the prediction of the ANN model for 3 typical test specimens. From these figures, it may also be concluded that the proposed method is also capable of simulating new test results. Although the results of the testing stage are not as good as those of the training phase, it still outperforms the traditional methods by a large margin. The test results used to measure the performance of the proposed NN model correspond, respectively to the best, average, and worst simulations among the training and testing
data sets. These results indicate the robustness of the ANN approach in learning and predicting the SWCC for soils without making any assumptions or simplifications a priori. 7 CONCLUSION
In this paper a Neural Network model was developed to estimate the SWCC for unsaturated soils. A database containing the result of pressure plate tests carried out on a wide variety of fine grained soils and thus covering a wide range of soil properties was employed to develop the model. Test results were digitized and normalized to obtain the necessary database. During the first phase, the model was trained using the results from 131 pressure plate tests. In the second phase, the model was tested using 55 additional test results that it had not been exposed to during the first phase. The results show that the model has a very good potential
465
in capturing the soil behavior and predicting the SWCC with a high accuracy. The proposed model outperforms the existing traditional methods by a large margin. REFERENCES
Bishop, A.W. and Donald, I.B. The experimental study of partly saturated soil in triaxial apparatus, Proc., 5th Int. Conf. Soil Mech. and Found. Eng., 1, pp 1321 (1961). Cresswell, H.P. and Paydar, Z. Water retention in Australian soil, description and prediction using parametric functions, Aust. J. of Soil Res., 34(2), pp 195212 (1996). Fredlund, D.G. and Rahardjo, H. Soil mechanics for unsaturated soils, Wiley, New York. (1993). Fredlund, M.D. Fredlund, D.G. and Wilson, G.W. Prediction of the soil-water characteristic curve from grain size distribution and volume-mass properties, Proc., 3rd Brazilian Symp. on Unsaturated Soils, Rio de Janeiro, pp 1323 (1997). Fredlund, M.D. Wilson, G.W. and Fredlund, D.G. Use of grain-size distribution for estimation of the soil water characteristic curve, Can. Geotech. J., 39(5), pp 11031117 (2002). Ghaboussi, J. Garrett, Jr. and Wu, X. Knowledge-based modeling of material behavior with neural networks, J. of Eng. Mech., ASCE, 117(1), pp 132153 (1991). Gupta, S.C. and Larson, W.E. Estimating soil water retention characteristics from particle size distribution, organic matter percent, and bulk density, Water Resources Res. J., 15(6), pp 16331635 (1979). Hutson, J.L. and Cass, A. A retentivity function for use in soil water simulation models, Soil Sci J., 38(1), pp 105113 (1987).
Mbagwu, J.S.C. and Mbah, C.N. Estimation water retention and availability in Nigerian soils from their saturation percentage, Communications in Soil Sci. and Plant Analysis, 29(7/8), pp 913922 (1998). Rawls, W.J. and Brakensiek, D.L. Estimating soil water retention from soil properties, J. of the Irrig. and Drain. Div., ASCE, 108(IR2), pp 166171 (1982). Reddi, L.N. and Poduri, R. Use of liquid limit state to generalize water retention properties of fine-grained soils, Geotech., 47(5), pp 10431049 (1997). Rumelhart, D.E. and McClelland, J.L. Explorations in parallel distributed processing, The MIT Press, Boston, Mass. (1988). SoilVision, SoilVision System Ltd., Sask., Saskachewan, Canada (2002). Tinjum, J.M. Benson, C.H. and Blotz, L.R. Soil water characteristic curves for compacted clays, J. of Geotech. and Geoenviron. Eng., 123(11), pp 10601069 (1997). Tomasella, J. and Hodnett, M.G. Estimating soil water retention characteristics from limited data in Brazilian Amazonia, Soil Sci. J., 163(3), pp 190202 (1998). Visser, W.C. An empirical expression for the desorption curve, Proc. of UNESCO IASH Symp. on Water in the Unsaturated Zone, Netherlands, pp 329335 (1969). Williams, J. Prebble, R.E. Williams, W.T. and Hignett, C.T. The influence of texture, structure and clay mineralogy on the soil moisture characteristic, Aust. J. of Soil Res., 21, pp 1532 (1983). Williams, R.D. and Ahuja, L.R. Estimating soil water characteristic using measured physical properties and limited data, Proc. of the Int. Workshop on Indirect Methods for Estimating the Hydraulic Properties of Unsaturated Soils, California, pp 405416 (1992). Zapata, C.E. Houston, W.N. and Walsh, K.D. Soilwater characteristic curve variability, Advances in Unsaturated Geotechnics, Geotech. Special Pub., 99, pp 84124 (2003).
466
Hysteresis between wetting and drying diffusivity parameters

D. Mabirizi & R. Bulut
School of Civil and Environmental Engineering, Oklahoma State University, Stillwater, USA
ABSTRACT: Determination of the unsaturated soil moisture diffusivity parameter is very important to variety of engineering applications, such as design and analysis of pavements, highway embankments, slopes, and shallow foundations. The unsaturated diffusivity coefficient controls transient moisture flow conditions within a soil in response to suctions or fluxes imposed at the boundaries of the soil mass. This research paper presents laboratory test results for wetting and drying unsaturated diffusion coefficients, investigates the hysteresis between the wetting and drying processes on the measured coefficients, and compares the laboratory measured diffusivities with the empirically determined field diffusivity values. Shelby tube soil specimens were obtained from compacted highway embankments from two different sites in Oklahoma. Soil specimens were exposed to drying and wetting cycles, and the resulting wetting and drying coefficients were evaluated. The wetting diffusion coefficients were generally higher than the drying diffusion coefficients by factors between one and two. 1 INTRODUCTION Twelve Shelby tube soil specimens, obtained from two different sites in Oklahoma, were tested to determine the wetting and drying diffusivity parameters. The drying and wetting testing method adopted in this study is based on the methodology proposed by Mitchell (1979) and subsequent improvements by Lytton et al. (2004) and Bulut et al. (2005) for the drying diffusion coefficient measurements. However, there are no cases in the literature to the knowledge of the authors where both the drying and wetting diffusion coefficients were measured on the same soil samples. Therefore, this study describes improvements in the current testing equipment and method developed at Oklahoma State University for determining both the drying and wetting diffusivity parameters. An evaluation of the effect of hysteresis on the single diffusivity parameters with the drying and wetting of soils was performed. The laboratory measured values were also compared with field values. 2 BRIEF THEORETICAL BACKGROUND
Determination of realistic estimates of the moisture diffusion properties is critical for predicting moisture flow in unsaturated soils. The unsaturated moisture diffusion controls transient moisture flow conditions within a soil mass in response to suctions or fluxes imposed at the boundaries of the mass. The flow of moisture through unsaturated soils is governed by the total suction gradient within the soil profile; with moisture traveling from regions of low total suctions to regions of high total suctions (Mitchell 1979; Fredlund & Rahardjo 1993). The description of the water flow mechanism within unsaturated soils is necessary for a number of issues including design and analysis of pavements, highway embankments, slopes, and shallow foundations. Mitchell (1979) proposed a very practical basis for characterizing unsaturated soil behavior. Mitchell (1979) showed that the rate at which water will move into the soil both vertically and horizontally can be determined using thermocouple psychrometers inserted into cylindrical soil specimens. Thermocouple psychrometers measure total suction changes over time as moisture evaporates from or infiltrates into the soil. Fitting a theoretical solution to the suction data permits an estimate of the moisture diffusion parameter. Laboratory methods based on Mitchell (1979) approach have the advantage of well-defined boundary conditions and permit moisture flow description using a single diffusivity parameter with a relatively high degree of confidence.
Darcys law describing one-dimensional flow of moisture through unsaturated soils is defined as: q k h dh dx (1)
where q soil water flux; k(h) unsaturated water coefficient of permeability; h total head (total suction); and x distance. Equation 1 is nonlinear due to dependence of permeability on suction. In unsaturated soils, permeability is a function of
467
total suction. Laliberte and Corey (1967) define the relationship between permeability and suction as: k h ko ho h
n
(2)
where ko saturated reference permeability; ho reference total suction; and n material constant. In Equation 2, total suction must be within the range for which the Laliberte and Corey (1967) permeability relationship is valid. Equations 1 and 2 lead to a nonlinear moisture diffusion equation, when continuity equation for water flow is applied. However, for the special case n 1, the solution reduces to a linear equation when total suction is expressed on a logarithmic scale, u log|h|. In this case the diffusion process is governed by a linear diffusion equation with the rate of diffusion controlled by a single diffusivity coefficient, : u t 2u (3)
Drying Test Specimen length, L 1 Shelby tube specimen Initial suction x (uo) Sealed sides and end
Figure 1.
Boundary conditions for equation 4.
where u total suction on logarithmic scale; del operator; and t time. Equation 3 defines the distribution of suction throughout the soil body as a function of space and time. Mitchell (1979) considered a cylindrical soil specimen sealed along its curved surface and one end while the other end is left open to permit flow of moisture in or out of the specimen as shown in Figure 1 and 2. Mitchell (1979) derived two linear expressions from Equation 3; first using separation of variables to derive the expression to estimate the drying diffusion coefficient (Equation 4) and second using Laplace transforms to derive the expression to estimate the wetting diffusion coefficient (Equation 5): u x,t ua
n 1 n sin zn cos zn z 2 t z x exp n 2 cos n
Thermocouple psychrometer measurements, u(x, t)
Sealed sides and end Specimen length, L Shelby tube specimen Initial suction (uo) Water level x
2 1
Wetting Test
z
2 uo ua sin zn
Thin layer Wetting from open end Soaking suction, u s of cloth

(4)
Figure 2. Boundary conditions for equation 5.
L
L
where ua atmospheric suction in the laboratory; uo initial equilibrium suction in the soil; diffusion coefficient; t time; L sample length; x coordinate; zn solution of cot zn zn/heL; and he evaporation coefficient. An evaporation coefficient he 0.54 cm1 was used in test data interpretation, based on previous work by Mitchell (1979). u x,t us 4 us uo
n 1 2 n 12 2t 2 n 1 x cos exp (5) 2 2L 4L
2n 1
1n
work by Mitchell (1979). This level of suction is very low and representative of the suction energy in the soaking water. Water movement from the soil matrix through the cracks and different permeable soil layers can be very complex and extremely difficult to model. However, total suction as a function of space and time can be used to realistically characterize moisture flow through unsaturated soil mass by a single diffusivity parameter. This approach provides a very practical basis for simple, economical, and relatively rapid laboratory measurements of unsaturated soil moisture diffusivity characteristics. 3 LABORATORY MEASUREMENTS
where us soaking (minimum) suction. Soaking suction of 1.75 log kPa was used in the diffusion coefficient measurements based on the previous
Thermocouple psychrometers with stainless steel screen shields were used to measure total suction
468
Thermocouple psychrometer measurements, u(x, t)
Atmospheric suction, ua Evaporation from open end
in the soil. Before diffusion tests are performed, the psychrometers were calibrated using salt solutions having known water potentials by immersing them in different concentrations of sodium chloride solutions. A plot of the millivolt output from the psychrometer versus water potential gives the calibration curve for the psychrometer. A datalogger was employed to record the water potentials on a computer. The drying and wetting testing method adopted in this study is based on the methodology proposed by Mitchell (1979) and subsequent improvements by Lytton et al. (2004) and Bulut et al. (2005) for the drying diffusion coefficient measurements. The wetting diffusion coefficient measurements have been incorporated to provide a unified testing procedure for measuring both the drying and wetting diffusion parameters on the same soil specimen as follows: 1. Select a soil specimen of about 20 cm long and trim the ends to provide a planar surface. Record the actual length (L) of the soil specimen. 2. Choose which end of the specimen will be exposed to open boundary and mark psychrometer positions (x) on the lateral side of the specimen. Mark the first psychrometer position about 5 cm from the exposed end. Provide a 3 to 5 cm interval between psychrometers. The distance from the open end to the first psychrometer may be changed depending on the soil type, soil moisture condition, and/or method of making psychrometer holes in the specimens. 3. Use a drill-bit to make holes for psychrometers, keeping the depth of the hole half the diameter of the soil specimen. The diameter of the holes should be large enough for psychrometers to fit. 4. Insert calibrated psychrometers into the holes and seal the holes tightly to prevent loss or gain of moisture. Record the psychrometer numbers and their distances from the closed end. 5. Seal the whole specimen (except the end that will be exposed to boundary suction level) with plastic wrap and aluminum foil to prevent loss or gain of moisture. 6. Place the specimen in one of the water bath tubes (Figure 3 by first exposing the open end to the atmosphere of known suction (if a drying test will be performed). Maintain the water bath at a temperature 25 0.1C throughout the testing period. 7. Connect the psychrometers to the datalogger and collect total suction values with time. 8. At the end of the drying process (if a drying test is being performed first), turn the exposed end of the same soil sample face upside-down and make contact with the liquid of known suction (Figure 4). 9. Continue with collecting total suction values with time. 10. Repeat steps 1 to 9 for each soil specimen.
Cylindrical PVC tubes to hold
Drying test in progress
Temperature control unit and circulator Thermocouple psychrometer lead wires Piezometric tubing system for wetting test
Water bath
Wetting test in progress
Figure 3. Drying and wetting diffusion test setup.
To datalogger and computer Psychrometer lead wires PVC tube sample holder Upward moisture movement
Water supply
Shelby tube soil specimen 2 1
Control valve
Water level
Thin layer of cloth Porous stone

Figure 4. ting test. Schematic drawing of soil specimen in wet-
With the testing protocol mentioned above, cycles of drying-wetting and wetting-drying can be performed to study moisture flow process by measuring the corresponding diffusivity parameters. Temperature control is extremely important in total suction measurements (Bulut et al. 2005). The water bath developed at Oklahoma State University is an important tool in maintaining the testing temperature of the specimens at relatively constant levels. The water bath can be used to perform drying and wetting coefficient measurements for up to ten samples at the same time. The testing room is maintained at 25 0.1C and a dehumidifier was used to control the relative humidity in the room where necessary. A typical diffusion test takes about seven to ten days to complete a drying and wetting
469
10000 Total Suction (kPa)
R = 0.9952 1000 R = 0.9978

2
100
10 1000
Drying-Theoretical curve Drying-Measured data Wetting-Theoretical curve Wetting-Measured data 10000 Time (minutes) 100000
Figure 5. Theoretical versus measured total suction values with time.
cycle, depending on the soil texture and the initial suction level of the sample. Typical suction values with time on a soil specimen exposed to one drying and wetting cycle using the above mentioned testing procedure is depicted in Figure 5. The filter paper method was also used to validate the initial total suction levels of the soil specimens. The testing procedure in Bulut et al. (2001) was adopted for the laboratory filter paper measurements. Schleicher & Schuell No. 589 filter papers were used in the study. The use of the filter paper method is especially warranted for low suction levels at which the reliability of thermocouple psychrometers are questionable. This method gives relatively consistent measurements at low suctions compared to thermocouple psychrometer readings. The atmospheric suction in the laboratory was determined by measuring the relative humidity in the air using a digital thermo-hygrometer. The atmospheric suction was then calculated using Kelvins equation (Fredlund & Rahardjo 1993) given by: h RT V ln RH (6)
open end were used for data interpretation. The testing sequence started with the drying process to obtain the drying diffusivity coefficient and then the wetting process to obtain the wetting diffusivity coefficient. The drying (dry) and wetting (wet) diffusion coefficients for each sample were determined on the same soil specimen. The results are given in Table 1 and Table 2 for Site I and Site II, respectively. The drying diffusion measurements range from 1.18 103 to 4.16 103 cm2/min while the wetting diffusion measurements range from 1.63 103 to 5.89 103 cm2/min. The diffusivity measurements indicate that wet values are generally higher than dry values by factors ranging from 0.6 to 1.7. These differences are probably due to the formation of crack networks within the soil column, which lead to moisture migration paths that are much shorter than those that would exist if the soil mass behaves as an intact medium. Several hairline cracks were observed on the open end of the soil column during the drying process and those cracks were closing up at the end of the wetting process. Also, root-holes in the soil column caused by the vegetative influence create water migration paths thus contribute to the differences in the diffusion values. Site I (Table 1) specimens generally have higher dry and wet values compared to Site II (Table 2). The laboratory values of both wetting and drying diffusivity parameters are substantially lower than the values predicted by the empirical equation given by Jayatilaka and Lytton (1997):
field 0.0029 0.000162 S 0.0122 h
(7)
where h total suction; R universal gas constant; T absolute temperature in Kelvin; V molecular volume of water; and RH relative humidity. 4 DISCUSSION OF TEST RESULTS
where S slope of suction versus water content curve and h volume change coefficient. Empirical methods for estimating S and h from basic soil index properties are given by Jayatilaka and Lytton (1997) and Covar and Lytton (2001), respectively. Based on the given soil index properties listed in Table 1 and Table 2, the field values of diffusivity parameters are within range from 2.38 101 cm2/min to 2.81 101 cm2/min. The amount of the soil finer than 2 size was assumed to be 40% for each soil in calculating the field values using Equation 7. 5 CONCLUSIONS
Twelve Shelby tube soil specimens, obtained from two different sites in Oklahoma, were tested to determine the wetting and drying diffusion coefficients. The soils from both sites were obtained near the ground surface at shallower depths. The soil specimens had initial soil suction values ranging from 1.73 to 2.64 log kPa, as determined by the filter paper method. The atmospheric suction in the testing room was relatively constant ranging from 5.20 to 5.32 log kPa. Two psychrometers were used for each diffusion test and the suction values from psychrometer closest to the
Testing equipment developed for the drying and wetting diffusion coefficient measurements in the laboratory provides a strong tool to study the hysteresis between the drying and wetting processes in a soil profile. The equipment can be used for running multiple tests at the same time under controlled temperature environment. The wetting diffusivity parameters are generally higher than the drying diffusivity values.
470
Table 1. Soil moisture drying and wetting diffusivity coefficients for site I. Atterberg limits Boring no. 1 2 3 4 5 6 7
1 4
Depth (cm) 3.0542.67 0.0044.20 0.0042.67 0.0038.25 0.0035.99 0.000.38 1.5239.62
LL (%) 53.2 48.8 50.3 51.3 49.6 50.2 51.4
PL (%) 27.6 25.3 25.7 27.1 26.8 24.9 27.2
% fine #200 (%) 78.0 79.8 76.2 77.2 79.3 77.8 76.0
Drying diffusion test uo dry 103 (cm2/min) 3.4211 1.8421 3.7368 4.1579 1.1842 2.9474 2.2632
3
Wetting diffusion test uo wet 103 (cm2/min) 5.8947 2.4211 5.0526 5.7474 1.6316 2.1053 2.7368
3
ua
R2
R2
5.30 5.23 5.28 5.30 5.24 5.25 5.20
2.46 2.60 2.64 3.35 2.55 1.73 2.02
0.9835 0.9936 0.9960 0.9989 0.9682 0.9937 0.9943
3.49 3.53 3.45 3.52 3.42 3.46 3.37
0.9802 0.9822 0.9878 0.9913 0.9889 0.9894 0.9988
ua Atmospheric suction (log kPa); 2uo Initial suction (log kPa); 3R2 Goodness of Fit values between measured and predicted (theoretical) total suctions; 4Testing sequence: first performed drying test and then wetting test.
Table 2. Soil moisture drying and wetting diffusivity coefficients for site II. Atterberg limits Boring no. 1 2 3 4 5
1 4
Depth (cm) 44.279.86 0.0039.62 0.0039.62 2.3541.22 0.0032.86
LL (%) 34.9 48.4 44.8 36.5 42.8
PL (%) 19.3 24.9 22.5 18.6 22.5
% fine #200 (%) 70.5 70.0 76.6 71.2 69.8
Drying diffusion test uo dry 103 (cm2/min) 1.3684 1.7368 3.4737 2.2632 1.2105
3
Wetting diffusion test uo wet 103 (cm2/min) 1.8158 1.9474 2.0789 3.3684 1.6316
3
ua
R2
R2
5.27 5.23 5.28 5.32 5.25
2.48 2.43 2.05 1.95 2.42
0.9978 0.9788 0.9998 0.9980 0.9980
3.52 3.45 3.54 3.47 3.39
0.9952 0.9955 0.9992 0.9937 0.9937
ua Atmospheric suction (log kPa); 2uo Initial suction (log kPa); 3R2 Goodness of Fit values between measured and predicted (theoretical) total suctions; 4Testing sequence: first performed drying test and then wetting test. Bulut, R., Aubeny, C.P. & Lytton, R.L. 2005. Unsaturated soil diffusivity measurements. Advanced Experimental Unsaturated Soil Mechanics; Proc. intern. symp. Trento, June 2729: 281286. Covar, A.P. & Lytton, R.L. 2001. Estimating soil swelling behavior using soil classification properties. In Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE Geotechnical Special Publication No. 115, Houston, Texas: 4463. Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils. John Wiley and Sons, Inc., New York. Jayatilaka, R. & Lytton, R.L. 1997. Prediction of expansive clay roughness in pavements with vertical moisture barriers. Research Report No. FHWA/TX-98/197-28F, Texas Transportation Institute. Laliberte, G.E. & Corey, A.T. 1967. Hydraulic properties of disturbed and undisturbed soils. Permeability and Capillarity of Soil, ASTM Special Technical Publication No. 417: 5671. Lytton, R.L., Aubeny, C.P. & Bulut, R. 2004. Design procedure for expansive soils. Research Report No. FHWA/ TX-05/0-4518-2, Texas Transportation Institute. Mitchell, P.W. 1979. The structural analysis of footings on expansive soils. Research Report No. 1, Kenneth W.G. Smith and Associates Pty. Ltd., Newton, South Australia.
This approach provides a very simple framework for experimental measurement of diffusion properties on an economical and routine basis. ACKNOWLEDGEMENTS The authors acknowledge the financial support provided for this research study by the Oklahoma Transportation Center. The authors would also like express their appreciation to Dr. Vincent (Butch) Reidenbach and Mr. Christopher Clark at the Oklahoma Department of Transportation for providing soil samples for this research study. REFERENCES
Bulut, R., Lytton, R.L. & Wray, W.K. 2001. Soil suction measurements by filter paper. In Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE Geotechnical Special Publication No. 115, Houston, Texas: 243261. Bulut, R. & Wray, W.K. 2005. Free energy of water suction in filter papers. ASTM Geotechnical Testing Journal 28: 355364.
471
A simple method to consider water density changes in the calculation of the degree of saturation of swelling clays
D. Marcial
Universidad Central de Venezuela, Caracas, Venezuela
ABSTRACT: When dealing with numerical computations in geotechnical engineering, the use of adequate values of concerned parameters is an aspect of mayor interest. Water transfers play an important role in the behavior of engineering clay barriers for deep underground waste disposal facilities. When absorbed by clays, water properties may considerably differ from the free water state. Water density w is a main parameter in the calculation of volumetric water content w and thus the degree of saturation Sr, a main state parameter in unsaturated soils mechanics. Many authors have pointed out an important increase of water density in swelling clays (see Winterkorn 1936, Martin 1962, Villar 2000, Marcial 2003, among others). This increase of w may lead to considerable errors if one assumes that w keeps equal to 1 Mg/m3 in such materials. This work presents a simple method to consider possible increases of w when calculating Sr values. A we curve is indirectly obtained from saturated high plasticity clay samples, and then some hypotheses based on experimental works (Prost 1975, Sposito & Prost 1982, Tessier 1984, Marcial 2003) permit to estimate w changes in the unsaturated state. The we curve is obtained for a MX80 clay in saturated conditions and values ranging from 1,02 to 1,57 Mg/m3 are obtained when the void ratio decreases from 0,957 to 0,358. The proposed method is applied in a real case to illustrate the procedure. The water retention curve in constant volume condition is represented in terms of Sr for w 1 Mg/m3 and considering w changes. In the first case Sr values up to 118% are obtained for a suction of 100 kPa; in the second one, Sr reduces to 86% for the same suction value. The presented method is a simple tool to deal with the increase of w in swelling clays and keep the degree of saturation within the defined range of 0 to 100%. 1 1.1 INTRODUCTION Experimental evidence with Sr values w value necessary to fit the condition Sr 1. This was done with 150 clay specimens, first compacted then saturated at constant volume. Results where grouped for dry density increments of 0.05 Mg/m3 and presented as in Figure 1. The figure clearly shows an increase of w with d. 1.2 Experimental evidence with absorbed water Figure 2 summarizes infrared absorption spectra obtained by Prost (1975) with homoionized
1.24 1.2
The use of highly compacted swelling clays in the actual concept of deep underground waste disposal facilities have led to a wide international laboratory testing campaign, encouraged by the corresponding international agencies. When calculating the degree of saturation in tested clay specimens, values higher than 100% are often observed (see Komine & Ogata 1994, Yahia-Aissa 1999, Villar 2000). Villar (2000) has conducted an important number of hydromechanical tests to characterize a high swelling clay at constant volume conditions (swelling pressure, permeability, water retention properties, among others). The author has systematically found Sr values much higher than 100% and thus dedicated special attention to this matter. After considering the possible sources of error, and based on experimental results (Martin 1962, Sposito & Prost 1982, Pusch et al. 1990, among others) Villar concluded that the main effect to consider in the increase of Sr values is an important increase of water density w when absorbed in high swelling clays. Villar proposed to calculate the
w (Mg/m3)
1.16 1.12 1.08 1.04 1.3 1.4 1.5 1.6 1.7 1.8 1.9
d (Mg/m3)
Figure 1. Increase of w with d for compacted and then saturated Cabo de Gata Bentonite (from Villar 2000).
473
GETTING THE we CURVE
2.1 Experimental setup and procedure In order to obtain a global measure of water density changes with dry density in saturated clay specimens, the procedure suggested by Villar (2000) was implemented and a 3 points we curve obtained. The main experimental set-up consisted of a rigid stainless steel infiltration cell, permitting to hold a 50 mm diameter and 5 mm height specimen, and a volume/pressure control system to inject deaired water to the specimen. A schematic cross section of the infiltration cell is presented in Figure 3. The cell is composed of a central ring to place the specimen, two rigid plates with porous discs and fixing screws to keep constant volume conditions. Rigid plates have a central inlet/ outlet drainage point for water exchanges and a high pressure seal is necessary to avoid leakage. The soil specimen may be prepared compacted or consolidated at a desired density, then placed in the central ring of the infiltration cell. In this case soil specimens were compacted in a mould and then pushed to the central ring. Target dry densities in this work are 1.4, 1.7 and 2.0 Mg/m3. Once the specimen placed in the ring, the cell is assembled as shown in Figure 3. Then the cell is immerged in a deaired distilled water bath, under vacuum condition (p 103 bar), to eliminate most of the air trapped in the compacted specimen. The cell is retired from the bath when no air bubbles are observed to emerge during at least 24 h. After that, the cell is connected to the injection system (Fig. 4) to be infiltrated with at least two pore volumes of distilled deaired water to ensure saturation (see Peirce & Witter 1986). The pore volume for the target d values of 1.4, 1.7 and 2.0 Mg/m3 are respectively 4633, 3521 and 2409 mm3. Because of very the low permeability values of swelling clays, high injection pressure are used to infiltrate the specimens (e.g. 1 MPa). However, long time periods are often necessaries to infiltrate high density specimens with two pore volumes of distilled water. In this case, 17, 57 and 106 days were necessaries to infiltrate the specimens with dry density of 1.4, 1.7 and 2.0 Mg/m3 respectively.
Figure 2. Infrared absorption spectra for homoionized Bentonite at different water content (Modified from Prost, 1975).
bentonite for different exchangeable cations and water contents. Note that for higher water contents all homoionized bentonites spectra present the same shape. A first peak appears at low frequency (32803420 cm1) and a second one at higher frequency (3630 cm1). However, as water content reduces, the shape of the spectra differs considerably depending on the type of exchangeable cation. At lower water contents, the hydration energy of the exchangeable cation influence de organization of adsorbed water layers strongly bonded to the clay surface and the hydrated cations located at the Sterns Layer. Sposito and Prost (1982) have shown that in swelling claywater systems at low water content the wetting/drying mechanisms are controlled by adsorption forces. These forces, exerted by the clay surface and exchangeable cations, strongly modify the structure and properties of adsorbed water. Denser water layers keep strongly bonded to hydrophilic sites and water exchanges first affect the water layers located at greater distances from hydrophilic sites. These water layers present lower density and weaker bonds with the adsorption complex of the clay. The appropriate understanding of water distribution when absorbed in clays, and the change of water properties with the distance to the clay surface (see Ichikawa et al. 1999) is necessary to deal with water density changes in unsaturated clayed soils.
Figure 3. Cross section of the stainless steel constant volume infiltration cell.
474
1962, Pusch 1982, Kenney et al. 1992, Dixon et al. 1996, Marcial 2003, among others). Some index properties of this clay are presented in Table 1. The specimens were compacted from the clay powder equilibrated at a suction of 103 MPa, corresponding to a water content of 8,24%. Then, the procedure indicated in 2.1. was applied to each specimen and the data reported in Table 2 was obtained. The we curve is plotted in Figure 5, note that a global density of absorbed water
Table 1. Properties Mineralogy CEC, meq/g Liquid limit, % Plastic limit, % Specific gravity Gs Activity Specific surface S, m2 Table 2. curve. 82% Na/Ca Montmorillonite 69.6 520 42 2.65 5.80 800 Index properties of MX80 bentonite.
Figure 4. Schematic view of the experimental set-up used to saturate compacted clay specimens.
A burette is connected to the outlet point of the infiltration cell. To avoid evaporation from the burette, the open side is connected to a zero suction ambience as seen in Figure 4. When infiltration ends, injection pressure is reduced to zero and the monitoring of the burette continues until stabilization is observed. This indicates that excess of pore pressure is completely dissipated. Then, the cell is disassembled and central ring is carefully whipped with a soft cloth and both sides of the specimen are trimmed to eliminate the excess of soil due to swelling. This procedure guarantees that soil specimen occupies only the volume of the ring. Then, the ring is weighed with the soil specimen. It is important to eliminate the excess of water from the surface of the cell to avoid a source of error in measuring the saturated weight of the specimen. After that, the specimen is placed to dry in the oven at 150C to obtain the dry weight Ws of the specimen and thus the weight of absorbed water Ww. After Tessier (1984), it is necessary to dry smectite clays at 150C to eliminate water molecules strongly bonded to exchangeable cations and the mineral surface. Marcial (2003) found that the loose of weight when drying from 105 to 150C is more important in Ca2 smectites than in Na smectites. Thus, to have a consistent mass reference system it is important to oven dry the clay specimens at 150C. 2.2 Material and results
Data of tested MX80 specimens for we Target dry density (Mg/m3)
Data Saturated weight W, g Dry weight Ws, g Volume V, cm3 Specific gravity Gs Dry density d, Mg/m3 Volume of voids Vv, cm3 Water weight Ww, g Water density w, Mg/m3 Void ratio e
1.6
1.4 18.39 13.43 9.916 2.65 1.354 4.849 4.96 1.02 0.957
1.7 20.90 16.44 9.719 2.65 1.691 3.515 4.46 1.27 0.567
2.0 23.46 19.35 9.916 2.65 1.952 2.612 4.11 1.57 0.358
1.4
w(Mg/m3)
1.2 1 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
The clay used in this work is the MX80 bentonite, a reference swelling clay studied by many authors (Winterkorn 1936, Bolt 1956, Aylmore & Quirk
Figure 5. we curve obtained with MX80 bentonite, compacted then saturated at constant volume.
475
increases significantly from 1.02 to 1.57 Mg/m3 as void ratio decreases from 0.957 to 0.358. 3 3.1 USING THE we CURVE TO GET Sr Proposed procedure
As indicated in 1.1, 1.2 and as obtained in 2.2, the density of water absorbed in bentonites increases as void ratio decreases. Also, densest water layers are strongly bonded to the adsorption complex of the clay, and water exchanges first occur in water layers of lower density, at higher distances from the mineral surface. One finds in the literature (Aylmore & Quirk 1962, Tessier 1984) that microporosity of bentonites keeps saturated for suction values up to 100 MPa. Thus, one can assume that in the unsaturated state, w increases as Sr decreases in the same manner that w increases when e decreases in the saturated state. Considering that in the unsaturated state water is fixed in the microporosity, the proposed procedure consist in obtaining w in the unsaturated state with the we curve obtained in the saturated state. To do so, the void ratio e Vv/Vs is substituted by the water ratio w Vw/Vs. One has to obtain indirectly the w value because one measures Ww instead of Vw. To do so, an iterative procedure between w and the we curve is necessary. This procedure is presented in Figure 6. The data necessary to obtain the degree of saturation Sr is the void ratio of the specimen e0, the water content w, the density of solids s, and the we curve. The first step is to read the w value corresponding to e0 in the we curve. With this w value, w and s calculate the water ratio w Vw/Vs. Then, read a new w value with e w in the we curve. With this w value, calculate again w and compare it with the precedent w value. If the difference between wi and wi1 is higher than an appropriated error (e.g. 0.001), read a new w value with e wi1 calculate w again, and so on. When the difference between wi and wi1 is lower than the error, stop the iteration and calculate Sr with the last w value. 3.2 Example with a water retention curve
Figure 6.
Iterative procedure to obtain Sr.
The proposed procedure is illustrated with a real example. Figure 7 shows the water retention curve WRC obtained at constant volume conditions for MX80
bentonite for a dry density of 1.7 Mg/m3. The WRC is presented in terms of Sr and suction s. Two curves are represented, the first one was obtained with w 1 Mg/m3, and the second one was obtained considering w changes with the procedure proposed in 3.1. In the first case, Sr values are higher than 100% for s lower than 11 MPa and Sr reaches 118% when s 0.1 MPa. However, when one considers w changes, Sr is kept within the defined range of 0 to 100%, as expected. If one calculates Sr for each w value obtained during the iteration process, the changes of Sr with the number of iterations N can be plotted. The convergence of Sr with N is represented in Figure 8 for s values lower than 11 MPa. Note that for N 0 Sr corresponds to the curve obtained with w 1 Mg/m3, and after a few iterations Sr converges to the solution.
476
Figure 7. Water retention curve obtained with MX80 bentonite at constant volume.
Figure 9. Changes of w and w with suction in compacted MX80 bentonite at constant volume (d 1.7 Mg/m3).
in the obtained we curve (Fig. 5) is not enough to obtain Sr in the same manner. In these cases, w was extrapolated until w reduces to 0.3 and then considered constant and equal to 1.71 Mg/m3. The retained w and w values are plotted as a function of suction in Figure 9. 4 CONCLUSIONS
Figure 8. tions N.
Convergence of Sr with the number of itera-
In this case, Sr is adjusted from 118, 115, 112 and 104% to 86, 82, 79 and 68% respectively. This suggests that using w 1 Mg/m3 an error of 37, 40, 43 and 53% is introduced for s values of 0.1, 1.0, 4.2 and 9 MPa respectively. For these s values, the average water density w was estimated in 1.37, 1.40, 1.43 and 1.53 Mg/m3 respectively. This is consistent with the formulated principle that densest water layers keep adsorbed in the clayed soil and lower density water layers are first exchanged. Concerning higher suctions, obtained w values are so reduced that the range of w values defined
Macroscopic and microscopic evidence reported by different authors has permitted to propose a procedure to deal with the increase of water density in high swelling clays and keep the degree of saturation in the defined range of 0 to 100%. The we curve, obtained in saturated state, clearly shows that w values much higher than 1 Mg/m3 can be found when e decreases to low values. In the case of numerical applications with bentonite clays, a special attention must be addressed to the degree of saturation Sr. A Sr value of 100% obtained with w 1 Mg/m3 does not necessarily indicate a full saturation condition and important errors can be introduced. A simple method to deal with this problem has been proposed in this work. ACKNOWLEDGMENTS Author thanks Universidad Central de Venezuela and Ecole Nationale des Ponts et Chauses for their financial support to this work. A special acknowledgment is addressed to Professor Pierre Delage who has firmly encouraged the research on this interesting topic.
477
REFERENCES
Aylmore L.A.G. & Quirk J.P. 1962. The structural status of clay systems. Proc. 9th Nat. Conf. on Clays and Clay Minerals.: 104130. Oxford: Pergamon. Bolt G.H. 1956. Physico-chemical analysis of the compressibility of pure clays, Geotechnique 6: 8693. Dixon D.A., Gray M.N. & Graham J. 1996. Swelling and hydraulic properties of bentonites from Japan, Canada and the USA. Environmental Geotechnics 1: 4348. Ichikawa Y., Kawamura K., Nakano M., Kitayama K. & Kawamura H. 1999. Unified molecular dynamics and homogenization analysis for bentonite behavior: current results and future possibilities. Engineering Geology. 54: 2131. Kenney T.C., van Veen W.A., Swallow M.A. & Sungalia M.A. 1992. Hydraulic conductivity of compacted bentonite-sand mixtures. Canadian Geotechnical Journal 29: 364374. Komine H. & Ogata N. 1994. Experimental study on swelling characteristics of compacted bentonite. Canadian Geotechnical Journal 31: 478490. Marcial D. 2003. Comportement hydromcanique et microstructurale des matriaux de barrire ouvrage. Thse de doctorat, Ecole Nationale des ponts et Chausses, CERMES. Paris, France. Martin R.T. 1962. Adsorbed water on clays: a review. Proc. 9th Nat. Conf. on Clays and Clay Minerals.: 5570. Oxford: Pergamon. Peirce J. & Witter K. 1986. Termination criteria for clay permeability testing. ASCE Journal of Geotechnical Engineering 112: 841854.
Prost R. 1975. Etude de lHydratation des Argiles: Interactions eau-minral et mcanisme de la rtention de leau. Thse de Doctorat, Universit Paris VI, INRA. Paris, France. Pusch R. 1982. Mineral-water interactions and their influence on the physical behavior of highly compacted Na bentonite. Canadian Geotechnical Journal.19: 381387. Pusch R., Karnland, O. & Hmark H. 1990. A general microstructural model for qualitative and quantitative studies on smectite clays. Thecnical report 9043; SKB. Stockholm, Sweden. Sposito G. & Prost R. 1982. Structure of water adsorbed on smectites. Chemical Reviews. 82: 552573, USA. Tessier D. 1984. Etude exprimentale de lorganisation des matriaux argile. Hydratation, gonflement et structuration au cours de la dessiccation et de la rehumectation. Thse de Doctorat, Universit Paris VI, INRA. Paris, France. Villar M.V. 2000. Caracterizacin termo-hidro-mecnica de una bentonita de Cabo de Gata. Tesis de doctorado, Universidad Complutense de Madrid. Madrid, Espaa. Winterkorn H.F. 1936. Studies on the surface behaviour of bentonites and clays. Soil Science 41: 2532. Yahia-Aissa M. 1999. Comportement hydromcanique dune argile gonflante fortement compacte. Thse de doctorat, Ecole Nationale des ponts et Chausses, CERMES. Paris, France.
478
Hydromechanical behaviour of a natural sand using different suction controlled techniques

Hossein Nowamooz & Maria Ioana Arsenie
Laboratoire de Gnie de la Conception, INSA de Strasbourg, Strasbourg Cedex, France Laboratoire Environnement Gomcanique & Ouvrages, Nancy-Universit, Vanduvre-ls-Nancy Cedex, France
Cyrille Chazallon
Laboratoire de Gnie de la Conception, INSA de Strasbourg, Strasbourg Cedex, France
Farimah Masrouri
Laboratoire Environnement Gomcanique & Ouvrages, Nancy-Universit, Vanduvre-ls-Nancy Cedex, France
ABSTRACT: The article presents an experimental study on compacted natural sand of Missillac in France using two different suction methods: Filter paper to measure the soil matrix suction; Tensiometer to measure the matrix suction range between 0 and 100 kPa. All the applied techniques permit to obtain the SWRC of the compacted samples. These experimental results, beside their modelling applications, provide a good database to compare the accuracy of each applied method in a suction range very close to the saturated state. Generally, we can find a good correspondence between these curves. Additionally, the Soil Water Retention Curve (SWRC) is used to add the suction parameter in a non linear elasticity model (respectively shakedown elastoplastic model) for the finite element modelling of the resilient behaviour (respectively long term behaviour) of road pavement under traffic loading. Unfavorable moisture conditions are one of the most common mechanisms of deterioration of low traffic pavements, with unbound granular bases. 1 INTRODUCTION In this context, this paper consists in a study of the effect of water content on the resilient behaviour of compacted granular materials in unsaturated state. These materials are used as subbase layer of pavement structures of low traffic roads. This research was made on an unbound aggregate subjected to cyclic loading and its purpose is to improve pavement design of the noticed roads. Additionally, the Uzan-Witczak model (Uzan, 1985) was adapted to replicate the effects of variations in suction that occur in partially saturated soils and aggregates. The parameters of this model have been optimized for the studied material and it is able to predict resilient deformations for different water contents and stress paths. 2 EXPERIMENTAL PROGRAM
In the pavement mechanic field, the researches are widely carried out in the saturated state, while the variation of the water content in the granular material layers has a significant influence on their resistance and on their mechanical behaviour. A complete description of material behaviour necessitates an effective stress approach with the pore pressures (or pore suctions) being separately controlled or monitored, from the applied pressures. Because most road materials are coarse grained and partially saturated and/or above the ground water table, it usually proves impossibleand certainly it is impractical in most situationsto monitor the pore suctions during each transient pulse. For this reason, almost all testing programs determine parameter values for resilient and incrementally-developed plastic strain models in terms of total, not effective, stresses. Instead, test procedures typically seek to control the moisture or suction conditions so that data is collected at representative conditions.
This section will present all the physical, hydraulic and mechanical tests performed on the Missillac sand. Missillac sand is a clayey sand, used as
479
subgrade soil on the L.C.P.C. (Laboratoire Central des Ponts et Chausses) accelerated pavement testing facility, for full scale pavement tests. Seven water contents have been studied with Repeated Load Triaxial Tests (RLTT). 2.1 Studied material
Mongiovi 2001, Tarantino and Tombolato 2005, Rahardjo et al. 2005). For the matrix suction measurements with filter paper method, the soil samples were compacted into two halves of 1 cm. A filter paper was inserted in between two larger in diameter protective filter papers. The filter paper sandwich was put between the soil halves. This test method controls the variables for measurement of the water content of filter paper that is in direct contact with soil. The tensiometer suction measurement technique permits to measure the matrix suction range between 0 and 100 kPa. The porous ceramic cups soaked in water were carefully put in contact with the samples. The measurements take a few hours after the installation to reach an equilibrium state. Figure 2 compares the experimental results obtained by tensiometer method and the filter paper method. To modify the original formula of Brooks and Corey (1964) for bimodal SWRCs, the original function is replaced by a three-part equation as follows:
ws M w wrM (ws wrM ) bM m w (w w ) bm rm rM rm
The underlying soil of the pavement structures are typically the mixture of the granular materials and water without being treated with hydraulic addictive. This type of material is generally used for the construction of the pavement assuring a low to average traffic defined less than 500 vehicles per day. Our studied material is the sand of Missillac. The particle size distribution of this material was shown in Figure 1. It is continuous (d60 1.10 mm; d30 0.48 mm; d10 0.01 mm) where the maximum particle size is less than 31.5 mm. The compaction curve of Missillac sand is also shown in Figure 1. The optimum of the normal proctor compaction is wopt 9% and d 2.06 Mg/m3. 2.2 Soil Water Retention Curve (SWRC)
To find the variation of suction with water content of the soil, laboratory tests for measurement of suction were made using two methods: the filter paper method (ASTM, 1995) and the tensiometer method (Konrad and Ayad 1997, Tarantino and
d bM bm bm
100 90 80 70 Fraction [%] 60 50 40 30 20 10 0 0.001 0.01 0.1 Diameter (mm) 1 10
(1) where ws, wrM, wrm, bM and bm are saturated water content, residual water content for macro-pores (saturated volumetric water content for micropores), residual water content for micro-pores, bubbling pressure of macro pores and bubbling
14 12 10 8 6 4 Filter paper suction measurement Tensiometer suction measurement Model prediction
2.10
2.05
Dry density [g/cm3]
2.00
1.95
1.90
1.85
Water content [%]

5 6 7 8 9 Water Content [%] 10 11 12 13
2
1.80
10
Suction [kPa]
100
1000
Figure 1. Particle size distribution curve. b) Normal proctor compaction curve of the Missillac sand.
Figure 2. Experimental results obtained with the tensiometer and the filter paper method.
480
Table 1. Parameters of SWRC of Missillac sand. Parameter bM (MPa) bm (MPa) M m wrm wrM ws Value 4.91 39.25 1.57 1.40 5.25 (%) 10.2 (%) 12.3 (%)
where 1 and 3 are the major and minor principal stresses. then, to study the resilient behaviour, series of short loadings (100 cycles) have been applied, with different stress paths (q/p), and different maximum stress levels (pmax, qmax). Six stress paths were studied (q/p 0; 0.5; 1; 2; 2.5; 3) (Figure 6) and three or four stress levels were applied for each stress path. For each load sequence, the last cycle was used to determine the resilient behaviour. To study the reversible behavior of material, the sample was subjected to a conditioning of 104 loading cycles by the repeated load triaxial testing, at a frequency of 1 Hz. To obtain stabilized deformations (the increase in the axial plastic deformation was lower than 108 per cycle), the stress path which has been followed is (p, q) (40 kPa, 80 kPa) from the initial state of stresses (p0, q0) (10 kPa, 5 kPa). A series of 100 cyclic loadings was applied to various ways (q/p) and for each level of (pmax, qmax). Six stress paths were studied (q/p 0; 0.5; 1; 2; 2.5; 3) and three or four stress levels were applied for each way. The last cycle is used to determine the resilient modulus (secant modulus). 3 MODELING
pressure for micro-pores, respectively. To find a proper curve for bimodal SWRC of the double porosity of the sand employed in this study bM, bm and ws were directly determined from experimental data and wrM, wrm, m and M were determined by fitting equation 1 to the data points (Table 1). 2.3 Triaxial tests In laboratory testing procedures it is well known that the size of the sample may have a very important influence on the results. In order to have a continuum condition in the sample, it is necessary to satisfy some conditions with regard to its microstructure and sample size. For the Missillac sand tests, the sample diameter was 76.2 millimetres with a 147.2 millimetres height. These tests have been carried out at the L.C.P.C. The material is prepared in a Proctor mould at a given water content and compacted with the normal Proctor test procedure. During the RLT test, the cylindrical specimen is tested by applying confining and vertical stresses. The Variable Confining Pressure (VCP) method (Gomes Correia, 1999) has been used here, where both the axial and the confining pressures are cycled. The frequency which has been used is 1 Hz. To study the resilient behaviour of the material, the following procedure has been used: the sample was first subjected to a conditioning, which consists in applying 104 loading cycles, at a frequency of 1 Hz, to stabilize the permanent deformations (at the end of this conditioning, the increase in the axial plastic deformation was lower than 108 per cycle). The cyclic stresses applied during the conditioning were (p, q) (40 kPa, 80 kPa), from an initial state of stresses (p0, q0) (10 kPa, 5 kPa), with p the mean normal stress and q the deviatoric stress. For a triaxial test, p and q are defined by: p 1 2 3 3 and q 1 3 (2)
When subjected to repeated mechanical loadings, two types of deformations are exhibited, linear or non-linear elastic (or resilient) and plastic deformations (Dawson, 2008). However more elaborate models take into account the irreversible behaviour, e.g. the Chazallon-Hornych model (Chazallon et al., 2006), the Suiker model (Suiker and de Borst, 2003) and the Mayoraz elastovisco-plastic (Mayoraz, 2002) model. But routine pavement design is mostly based on an elastic calculation, using a resilient modulus and Poissons ratio for each layer. Consequently, the Uzan model is a good option in pavement modeling to ensure a valid stress, strain, and deflection evaluation in pavements. The resilient modulus for most unbound pavement materials and soils is stress-dependent but the Poissons ratio is not, or at least to a much smaller extent. Biarez (1961) described the stressdependent stress-strain behaviour of granular materials subjected to repeated loading. Independently, similar work was performed in the United States (Hicks and Monismith 1971). Both results presented the k- model, which is written with dimensionless coefficients like: 3 p M r k1 k2 k1 pa pa
k2
(3)
481
where Mr is the resilient modulus, p is the mean stress, pa is the reference pressure (pa 100 kPa) and k1, k2 are coefficients from a regression analyses usually based on repeated load triaxial test results. This model has been very popular for describing non-linear resilient response of unbound granular materials. It assumes a constant Poissons ratio and that the resilient modulus is independent of the deviatoric stress. To address this latter limitation the Uzan-Witczak model has become widely promoted, especially, in recent years, by authors in North America, e.g. (Pan et al. 2006). It takes the form: p Mr k1 . pa pa
k2
a) 100
Mr [MPa]
w=6% w=7% w=8% w=9,6% w=10% w=11% w=12,3% w=6% model w=7% model w=8% model w=9,6% model w=10% model w=11% model w=12,3% model
10 100 1000
10
b)
100
p' [kPa]
q . 1 pa
k3
w=6% w=7% w=8% w=9,6% w=10% w=11% w=12,3% w=6% model w=7% model w=8% model w=9,6% model w=10% model w=11%model w=12,3% model
10
(4)
where Mr is the resilient modulus, p is the mean stress, q is the deviatoric stress, pa is the reference stress (pa 100 kPa) and k1, k2, k3: are the model parameters. The calibration of the Uzan model, expressed in terms of effective stress, has been performed. The suction has been taken into account and the following expression has been proposed for the normal stress component of the stress tensor:
M r [M Pa]
10
p [kPa]
100
i ( i ua ) . (ua uw ) 0 1, i 1, 2, or 3
p Mr k1 . pa pa
k2
and
Figure 3. Evolution of resilient modulus Mr with a) the effective stress and b) the total stress for q/p 0.5.
q . 1 p a
k3
a)
(5)
1000
w=6% w=7% w=8% w=9,6% w=10% w=11%
M r [M Pa]
To obtain the effective stress values, two hypotheses were made: Since the water contents of the sample were not modified during the cyclic triaxial tests, the suction values were supposed to be equal to their initial values presented in Figure 2. The parameter was taken equal to the initial degree of saturation of the samples. The model was applied separately for the stress paths q/p 0.5 and 3. The optimized parameters k1, k2and k3 for the different water contents are shown in Table 2. It should be mentioned that three experimental values of Mr for each water
Table 2. Results of the parameter optimisation on the parameters of the Uzan model. Water contents (w (%)) 6; 7; 8; 9,6; 10; 11; 12,3 Stress paths (q/p) 0.5; 3 Model parameters k1 1098.92 k2 0.09 k3 0.52
100
w=12,3% w=6% model w=7% model w=8% model w=9,6% model w=10% model w=11% model
10 10 100 1000
w=12,3% model
b)
p' [kPa]
1000
w=6% w=7% w=8% w=9,6% w=10% w=11% w=12,3% w=6% model w=7% model w=8% model w=9,6% model w=10% model w=11% model w=12,3% model
10 100
M r [M Pa]
100
10
p [kPa]
Figure 4. Evolution of resilient modulus Mr with a) the effective stress and b) the total stress for q/p 3.
482
content of the different stress paths were used for the modeling estimations. These values have been determined at the minimum, medium and maximum mean effective stresses. The correlations obtained with the model are satisfying comparing with the experimental results (Figures 3-a, and 4-a) where all the points of the cycle are used. The experimental results were estimated with the total stress concept, as well in Figures 3-b and 4-b. It can be noted that the modeling results based on the total stress produces the same form for all the water contents of the different stress paths. 4 CONCLUSION
The article presents an experimental study on compacted natural Missillac sand using two different suction methods (filter paper and tensiometer) to measure the matrix suction range between 0 and 100 kPa. These techniques permitted to obtain the SWRC of the compacted samples. These experimental results beside their modeling applications provide a good database to compare the accuracy of each applied method in a suction range very close to the saturated state. Generally, we can find a good correspondence between these curves. Additionally, the Soil Water Retention Curve (SWRC) is used to add the suction parameter in non linear elasticity model and to measure the effective stress for the different cyclic triaxial tests. The capability of the existing Uzan model with the estimated effective stresses was verified for simulating the resilient behaviour. The effective stress concept permits a better modeling estimation compared to the total stress estimation. ACKNOWLEDGMENT The authors want to acknowledge Dr. Pierre Hornych from L.C.P.C. for providing the RLT tests results. REFERENCES
ASTM D 5298-94, 1995. Standard test method for measurement of soil potential (suction) using filter paper, 4.09, 154159.
Biarez, J., 1961, Contribution ltude des proprits mcanique des sols et des matriaux pulvrulents, PhD Thesis, Faculty of Sciences, University of Grenoble, France. Brooks, R.H. and Corey, A.T., 1964, Hydraulic properties of Porous Medium, Hydrology paper, No. 3. Civ. Eng. Dep., Colorado State Univ., Fort Colins, Colo. Chazallon C., Hornych P., Mouhoubi S., 2006, An elastoplastic model for the long term behaviour modelling of unbound granular materials in flexible pavements, International Journal of Geomechanics, ASCE, 6 (4), p 279289. Dawson, A., 2008, Water in road structures, movement, drainage and effects , Springer, p 436. Gomes Correia, A., 1999, Proceedings, Unbound Granular MaterialsLaboratory testing, In-situ testing and modelling. A.A. Balkema, Rotterdam. Hicks, R. and Monismith C., 1971, Factors Influencing the Resilient Response of Granular Materials, Highway Research Record 345, Highway Research Board, Washington, D.C. Konrad, J.M., and Ayad, R., 1997, Desiccation of a sensitive clay: field experimental observations, Canadian Geotechnical Journal, Vol. 34, pp. 929942. Mayoraz F., 2002, Comportement mcanique des milieux granulaires sous sollicitations cycliques: Application aux fondations des chausses souples, Ph. D. thesis, Ecole Polytechnique Fdrale de Lausanne, Suisse. Pan, T., Tutumluer, E. & Anochie-Boateng, J., 2006, Aggregate Morphology Affecting Resilient Behavior of Unbound Granular Materials, Transportation Research Record 1952, Transportation Research Board, Washington, DC, pp. 1220. Rahardjo, H., Lee, T.T., Leong, E.C., and Rezaur, R.B., 2005, Response of a residual soil slope to rainfall, Canadian Geotechnical Journal, Vol. 42, pp. 340351. Suiker, A.S.J., and de Borst R., 2003, A numerical model for cyclic deterioration of railways tracks, International Journal for Numerical Methods in Engineering, 57, pp. 441470. Tarantino, A., and Mongiovi, L., 2001, Experimental procedures and cavitation mechanisms in tensiometer measurementsGeotechnical and Geological Engineering, Vol. 19, pp. 189210. Tarantino, A., and Tombolato, S., 2005, Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay, Gotechnique, Vol. 55, No. 4, pp. 307317. Uzan, J., 1985, Characterization of Granular Materials, TRR 1022, TRB, Washington, D.C.
483
Investigations on water retention behaviour of deformable soils

S. Salager, A. Ferrari, M. Nuth & L. Laloui
Swiss Federal Institute of Technology EPFL, Lausanne, Switzerland
ABSTRACT: The trend in modern unsaturated soil mechanics is to enhance the physical understanding of the soil water retention properties. It is agreed that the retention characteristics depend on the state of compaction and are related to the deformability of the media. Therefore, any appropriate experimental set up and behavioural model shall take into consideration the evolution of the density state in the course of drying or wetting processes. The results from an exhaustive experimental characterization of the retention behaviour of a granular soil is synthesized. The information concerning degree of saturation, suction and void ratio is assessed to quantify in particular the effect of the passed and present states of mechanical stress and strain on the shape of the curves. The void ratio is noticed to exert a clear influence on the air entry value. In the plane water content versus suction, the experimental results highlight the fact that from a certain value of suction, the retention curves corresponding to different densities of the same soil converge into a single curve. The observed features of behaviour are conceptualized into a new modelling framework expressing the evolution of the degree of saturation as a function of suction. The proposed retention model makes use of the theory of elasto-plasticity. The calibration of the model requires experimental retention data for two initial void ratios. The prediction of tests for further ranges of void ratios proves to be accurate which supports the adequacy of formulated concepts. 1 INTRODUCTION degree of saturation as a function of suction and deformations. The proposed elasto-plastic retention model takes into account the effect of the initial compaction and of the density variations due to suction in the evolution of the degree of saturation. Finally, to prove the model abilities to reproduce the observed behaviour, the experimental results are used for the model validation. 2 EXPERIMENTAL DETERMINATION OF SOIL WATER RETENTION CURVES
Understanding of the water retention features is an essential aspect for modelling the behaviour of unsaturated soils. It has been evidenced that the capillary characteristics of a deformable soil depends on the state of compaction (Miller et al. 2002, Sugii et al. 2002, Tarantino & Tombolato 2005, Salager 2007). Moreover, the soil deformation is strongly responsive to the suction variations and it becomes necessary, for the construction of the water retention constitutive model, to take into account the evolution of the density state along drying or wetting processes (Delage et al. 1998, Romero 1999, Fleureau et al. 2002). Consequently, the challenge is firstly to identify definite trends in the retention behaviour, focusing on the suction effects on both the volumes of water and voids, and secondly the influence of density on these trends. From these conclusions, the other stake is to formulate an adequate modelling framework for the water retention behaviour of deformable soils. An experimental programme is presented for this purpose in this paper concerning a granular soil (Salager et al. 2007). The analysis of these results leads to understand the soil water retention features and the hydro-mechanical couplings involved in drying processes. The observed features of behaviour are conceptualized into a new modelling framework accounting for the evolution of the
The testing programme performed by Salager (2007) on a clayey silty sand is used here to investigate the water retention behaviour of a granular soil. The soil is classified as SC-CL according to the USCS. The liquid and plastic limits are respectively 25% and 14.5%. Sand, silt and clay fractions are respectively 72%, 18%, and 10%. The optimum water content is 14.5% at a dry density d equal to 1860 kg.m3. The specific gravity of the soil is Gs 2.65. The experimental campaign involved five series of samples compacted at five different densities (Table 1). For the loosest state, the samples were compacted at a dry density of 1350 kg.m3 corresponding to a void ratio e of 0.96. Below this dry density, the presence of macro-pores does not ensure anymore a good homogeneity of the
485
Table 1.
Moulding and initial states for soil samples. Moulding states Initial states
Void ratio e (-)
1.2
e =1.009 e =0.857
0 0
e =0.682 e =0.551
0 0
e =0.436
0
Series 1 2 3 4 5
e(/) 0.96 0.77 0.61 0.47 0.36
d (kg.m3) 1350 1500 1650 1800 1950
w(%) 18 18 18 14 12
e(/) 1.01 0.86 0.68 0.55 0.44
d (kg.m3) 1319 1427 1575 1708 1845
w(%) 38.1 32.3 25.7 21.0 16.5
0.8
0.6
0.4
0.2 0.01
Suction s (kPa)
100
104
106
40 35
e =0.96 3 e 0=0.76 6 e =0.60 6
0 0 0 0
Figure 2. Variation of void ratio under the suction loading for the 5 densities of the granular soil.
Water content w (%)
30 25 20 15 10 5 0 0.01 1 100
e =0.47 2 e =0.35 9
Suction s (kPa)
104
10 6
Figure 1. Water retention curves for the 5 densities of the granular soil.
samples. The samples of the most compacted series had a dry density of 1950 kg.m3. In order to obtain the retention curves of the soil in a wide range of suctions, two experimental techniques were used. Concerning the lower suction values, from 1 to 1000 kPa, the axis translation technique appeared to be the most suitable and it was used in a pressure plate apparatus. To investigate the retention behaviour in the lower saturation states, suction was imposed in desiccators by means of the vapour equilibrium technique. This technique allowed reaching suction values in the range 4 400 MPa. For each series, an average of 16 to 18 suction values were imposed and the main drying paths determined. At equilibrium, the water content w and void ratio of the samples were measured. Each measure reported hereafter is an average over three samples. Figure 1 presents the results obtained for the five sample series in the plane water content versus suction. These points lead to define the water retention curves of the clayey silty sand compacted at five different initial void ratios e0. Water content variations are limited for suction values lower than some kPa; significant variations have been observed for suction values between some kPa and about 3000 kPa; these variations correspond to the high desaturation linear part. For suction values higher than
200 kPa, the curves tend to converge in a unique shape, independently of the initial void ratio. Void ratio variations associated to the suction increase are depicted in Figure 2. For each initial void ratio, the curve can be split into three parts. For low suction values (lower than some kPa), no evolution of the density is noted with respect to the suction. The capillary forces between grains are to low to induce granular rearrangement. For the intermediate suction values (between some kPa and about 3000 kPa) the suction loading leads to decrease strongly the void ratio. These variations are in the range from 0.2 for the series corresponding to the highest initial void ratios, to 0.06 for the series corresponding to the lowest initial void ratios. For the high suction values, in this type of soil, the water is mainly present in an adsorption form and the capillary effects are lower. 3 WATER RETENTION BEHAVIOURAL FEATURES
From the presented experimental results, it can be noticed that the parameter of the water retention curve which is the most influenced by the density is the air entry value. Figures 1 shows that the air entry value increases with a decrease of the initial void ratio. This result corroborates the trend drawn by several authors (Vanapalli et al. 1999, Kawai et al. 2000, Zhou & Yu 2004). As well known, a suction increase applied to an unsaturated soil induces a migration of a part of the liquid phase. When the air entry value is reached and exceeded, the drainage process becomes quicker; this transition in the water retention behaviour reflects a limit level of capillary tension for the soil. This limit varies oppositely to the size of the pores, and thus to the void ratio. That is the reason why one can admit that the air entry value decreases when void ratio increases. To take into account this coupling between water retention behaviour and density, Nuth & Laloui
486
(2008) presented a model in which the air entry value seH is updated as a function of the skeleton volume information. The proposed explicit expression is the following linear approximation: seH se H . v (1)
where se is the reference air entry value, v the volumetric strain and is a material parameter. The function of the Equation 1 presents two advantages: firstly it is easy to calibrate because only one parameter is needed besides the reference air entry value and secondly, this function showed its predictive power for a wide range of materials (Nuth 2008). Where se is the reference air entry value, v the volumetric strain and h is a material parameter. Equation 1 presents two advantages: firstly it is easy to calibrate because only one parameter is needed besides the reference air entry value and secondly, this function showed its predictive power for a wide range of materials (Nuth 2008). Another important feature noted from the experimental results presented previously is the fact that in the plane water content versus suction, the water retention curves corresponding to different initial densities converge from a certain suction value to become a unique curve for high suctions. In fact, the void ratio influence is no more visible beyond a suction in the range of 100 to 200 kPa. During a mechanical loading, here at the time of the static compaction in the process of the sample fabrication, the porous network of the soil is modified. The larger pores are easily collapsed while the smaller pores need more important stress to be reduced. However, there is a part of the pores, the thinnest, which could only be reduced applying enough stress to break some grains (Fredlund & Pham 2007). Consequently, the compaction regards the low suctions part of the water retention curve while for the high suctions this curve can be considered as unique. In term of modelling, this convergence point in the plane water content versus suction (the couple (sconv ,wconv)) is unique and independent from the density. This convergence point can be translated in the plane degree of saturation versus suction by the following relation: Sr wconvGs e (2)
is formulated. The proposed concept is based on previous works from Nuth & Laloui (2008), i.e. the model ACMEG-s intended for modelling the stressstrain and retention behaviour of unsaturated soils. Although the reference model ACMEG-s comprehends the capillary hysteresis, it is chosen here to not to discuss this feature, in order to better put into light the mere effects of void ratio on the shape of the soil water retention curve. However, the changes of degree of saturation, which is the hydraulic strain variable, keep on being modelled as either reversible or dissipative, like in the reference model. The increment of degree of saturation dSr is thus decomposed into an elastic part dSer and a plastic part dSpr: dSr dSre dSrp (3)
In the first place the decomposition Equation 3 was justified essentially by the need of describing the scanning lines, which are the unloading-reloading lines during drying-wetting cycles (Viaene et al. 1994). Here the study is limited to the drying paths exclusively; still, Equation 3 will help to improve the quantitative response of the model in the neighbourhood of the air entry value. For the matter of understanding, it is assumed provisionally that the void ratio remains perfectly constant during a drying path. In the plane degree of saturation versus suction (Fig. 3a), the model for soil water retention curve evidences 4 main domains of behaviour, which are also noticeable in plane water content versus matric suction (Fig. 3b). Starting with fully saturated conditions, the degree of saturation remains equal to one for suctions that are lower than the air entry value seH (domain (a)). Beyond this limit value of suction, air can break into the pore space and the saturation level decreases first moderately (domain (b)), and then more rapidly as suction continues to increase (domain (c)). In domain (b), the increment of degree of saturation is entirely elastic and equal to: dSre ds K H s /se H (4)
KH being the elastic modulus. The drying yield suction sD is defined as the limit between domains (b) and (c) in Figure 3a. It is the parameter of the yield surface defined as: f s sD (5)
where Gs is the specific density. 4 MODELLING FRAMEWORK FOR RETENTION CURVE
In domain (c), the threshold sD evolves with the history of suction, from a given initial value sD0, and pilots irreversible changes in the degree of saturation s ln D H Srp sD 0 (6)
In the light of the features of the retention behaviour presented above, a new modelling framework
487
has been validated by Nuth & Laloui (2008) on a number of fine-grained materials. It is consistent with the previous observations evidencing an increase in the air entry value for denser states in silty sand and clay. If the change in void ratio has a direct effect on the air entry value sHe, it also affects indirectly the complete shape of the retention curve. It is indeed assumed that the size of the elastic domain in plane (Sr, s) remains constant, that is to say that whatever the void ratio, the width of domain (b) in Fig. 6a is unchanged, see paths AB and AB. This implies the following condition on the initial yield suction sD0: sD 0 sDI seH se (8)
Figure 3. Conceptual representations of the proposed retention model.
H is the coefficient of compressibility for the plastic part of degree of saturation Spr. As explained later, the coefficient of plastic compressibility defined in Equation 6 is fully determined by the values of the model parameters and needs not to be calibrated. The model response in domain (d) (Fig. 3a) needs to be addressed from the viewpoint of water content versus suction plane (Fig. 3b). As mentioned previously, a convergence point is identified at the suction sconv. Beyond this value, all retention curves in plane (w, lns) converge into a unique curve, assumed here to be a straight line up to the maximum suction smax, which is a material parameter. The shape of the retention curves in terms of degree of saturation are thus deduced from planes (w, lns), (e, lns) and Equation 2. The effect of void ratio is introduced via the dependency law of the updated air entry value sHe presented in Equation 1 which can be also written: seH se H e e0 1 e0 (7)
with e0 being the reference void ratio, se the air entry value at this reference void ratio, and H a material parameter. The relationship Equation 7
where sDI is the initial yield suction corresponding to the void ratio of reference. Supposing at first that no deformation occurs during the drying process, the modelled retention curves of two samples at different initial void ratios e0 and e1, can be compared (solid thick lines and dotted lines, respectively, in Figure 3). Provided that e1 is lower than e0, the corresponding air entry suction for sample 1 (point A, Figure 3a) is greater than the one of the reference sample (point A). The drying yield suction corresponding to e1 (point B) is also greater than the reference one measured at e0 (point B). Beyond the drying yield suction, for constant void ratios, the response to suction in terms of degree of saturation is a straight line until reaching the suction of convergence (points C and C in plane). Consequently, the coefficient H depends on the void ratio. In domain (d), the degree of saturation is calculated on the basis of the water content and the void ratio. The main concern of the paper is the understanding of the retention behaviour of deformable materials, that is to say soils that evidence changes in void ratio not only due to mechanical loading but also during drying and wetting processes. In those types of soil the assumption of a void ratio that remains equal to its initial value while suction changes does not hold. Due to Equation 7, any change in void ratio will provoke an update of the air entry value, which in turn will affect the drying yield suction (Eq. 8). As a consequence the whole retention curve in the plane (Sr, s) is shifted right, while the suction of convergence sconv is not affected. The solid gray line in Figure 6a shows that the degree of saturation will in consequence evolve non linearly with lns, which is necessary to join points A and C. The deformable soil water retention curve can thus be drawn by making the point (Sr, s) jump from one retention curve drawn at a constant void ratio to the other. In the case of Figure 3, this array of
488
curves is bounded by the two curves obtained at constant void ratios e0 and e1. The quality of the prediction for the degree of saturation will thus depend on the accuracy of the prediction of the evolution of the void ratio with suction. The proposed water retention model needs consequently to be coupled with a stressstrain model for unsaturated soils. An appropriate constitutive mechanical framework shall be elasto-plastic and accounts for the suction effects on the stress-strain response (e.g. Nuth & Laloui 2007, Wheeler et al. 2003, Khalili & Loret 2002). As it is not the scope of the paper to describe a stress-strain model for unsaturated soils, it is proposed to decouple the mechanical model from the present conceptual framework. Hence, the void ratio is hereafter uniquely considered as an input for the model of retention. In order to avoid justifying and discussing the parameter determination of any given stress-strain model, the experimental data void ratio versus suction will be taken in the following as the required input. 5 SIMULATIONS
Degree of saturation Sr (-)
10
(a )
Test 5
Air entry value s (kPa)
Exp. Model
Test 4
Test 3
Test 2
Test 1
0 0.3
0.5
0.7
0.9
1.1
Void ratio e (-)

0. 7 (b) 0. 6
Degree of saturation Sr (-)
0. 5 0. 4 0. 3
Test 5
0. 2 0. 1 0
Test 3 Test 1 Test 4 Test 2
0.4
0. 8
conv
Void ratio at convergence e
(-)
1.2 (c) 1 0.8 0.6 0.4 0.2 0 0.1

Exp. Test 1 Model Test 1 Exp. Test 2 Model Test 2 Exp. Test 3 Model Test 3 Exp. Test 4 Model Test 4 Exp. Test 5 Model Test 5
The experimental results presented above are used to assess the capabilities of the proposed modelling framework. The aim is to validate the concepts and to formulate recommendations for further advanced modelling of the retention curve. The numerical results are issued from (i) the calibration of the reference retention curve at the given void ratio for each data set and (ii) the prediction of the water retention curves for lower void ratios. The material parameters for the modelled soil are gathered in Table 2. They are calibrated on the basis of the retention curve determined for a reference void ratio of e equal to 0.7 with the following procedure: reference air entry value se is first determined the coefficient of coupling h is fitted with the help of a second air entry value the reversible slope is KH fitted the drying yield suction sDI is calibrated the point of convergence sconv, wconv is determined the maximum suction smax is then adjusted The calibration of the term of coupling h is plotted in Figure 4a. The increase in air entry value with void ratio is fairly reproduced by the
Table 2. se 300 kPa Material parameters used for the simulations. KH 25 h 15 h 1e5 sconv 1278 kPa smax 400 kPa
10
Suction s (kPa)
1000
10
0.4 (d) 0.35

E xp. Test 1 Model Test 1 E xp. Test 2 Model Test 2 E xp. Test 3 Model Test 3 E xp. Test 4 Model Test 4 E xp. Test 5 Model Test 5
Water content w (-)
0.3 0.25 0.2 0.15 0.1 0.05

w
con v
0 0.1
scon v
10
1000
10 5
Suction s (kPa)
Figure 4. Simulation of the retention behaviour of a clayey silty sand, experimental data from Salager (2007) (a) Calibration of parameter H (b) evolution of degree of saturation versus void ratio at the suction of convergence (c) modelled retention behaviour: degree of saturation versus suction (d) modelled retention behaviour: water content versus suction.
489
model, indicating that the linear approximation is acceptable. Provided that the air entry suctions cannot be negative, the function plotted in Figure 4a reaches a plateau (minimum value) for void ratios greater than 0.7. The plane (Sr, econv) is also of interest (Fig. 4b), as it evidences the non linearity of the relationship between the degree of saturation and the void ratio at the suction of convergence. The modelled degree of saturation (Fig. 4c) is more particularly accurate for lower void ratios (e.g. tests 3, 4 and 5). The predicted water contents are satisfactory both in the zone of low suction and beyond the suction of convergence (Fig. 4c). It may be noticed that the calibration of parameter sconv has an important effect on the quality of the prediction in plane (w, lns). In the present case, the maximum suction smax could be calibrated either in the plane (w, lns) or (Sr, lns), as all the retention curves remarkably reach a quasi null water content and degree of saturation for a suction of 400 MPa. Provided that the air entry value is void ratio dependent, the variable e must be known anytime. As it is not the scope of the paper to present a constitutive model for the mechanical behaviour, the experimental void ratios have been directly taken into consideration for the calculations (Fig. 2). The proposed model thus constitutes a noteworthy alternative to conventional non-linear elastic model. It makes use of a number of physical parameters with straightforward determination, and can readily feature the capillary hysteresis without any supplementary parameter. The modelled response relies anytime on information on the void ratio, the quality of which will have a direct effect on the prediction of the retention behaviour. 6 CONCLUSION
degree of saturation as a function of suction and deformations. The proposed retention model makes use of the theory of elasto-plasticity. The calibration of the model requires experimental retention curve for two initial void ratios. Finally, in order to analyze in details the water retention behaviour and to prove the model abilities, the experimental results were used for the model validation. REFERENCES
Delage, P., Howat, M.D. & Cui, Y.J. 1998. The relationship between suction and swelling properties in a heavily compacted unsaturated clay. Engineering Geology, 50:3148. Fleureau, J.M., Verbrugge, J.-C., Huergo, P.J., Gomes, A., Correia & Kheirbek-Saoud, S. 2002. Aspects of the behaviour of compacted clayey soils on drying and wetting paths. Canadian Geotechnical Journal 39:13411357. Fredlund, D.G. & Pham, H.Q. 2007. Independent roles of the stress state variables on volume mass constitutive relations. 2nd international conference on mechanics of unsaturated soils: 12231256. Springer,. Kawa, K., Karube, D. & Kato, S. 2000. The model of water retention curve considering effect of void ratio. Unsaturated soil for Asia: 329334. Rotterdam: Balkema. Nuth, M. & Laloui, L. 2008. Advances in modelling hysteretic water retention curve in deformable soils. Computers and Geotechnics 35(6):835844. Miller, C.J., Yesiller, N., Yaldo, K. & Merayyan. S. 2002. Impact of soil type and compaction conditions on soil water characteristic. Journal of Geotechnical and Geoenvironmental Engineering 128:733742. Romero, E. 1999. Characterisation and thermo-hydromechanical behaviour of unsaturated boom clay: experimental study. PhD thesis, Universitat Politechnica de Catalunya. Salager, S. 2007. Etude de la retention deau et de la consolidation des sols dans un cadre thermohydromcanique. PhD thesis, Universit Montpellier 2. Salager, S., El Youssoufi, M.S. & Saix, C. 2007 Experimental study of the water retention curve as a function of void ratio. Int. Conf. Geo-Denver, Denver. Sugii, T., Yamada, K. & Kondou, T. 2002. Relationship between soil-water characteristic curve and void ratio. 3rd International Conference on Unsaturated Soils: 209214. Swets and Zeitlinger. Tarantino, A. & Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique 55(4):307317. Vanapalli, S.K., Fredlund, D.G & Pufahl, D.E. 1999. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Gotechnique 44:892898. Viaene, P., Vereecken, H., Diels, J. & Feyen. J. 1994. A statistical analysis of six hysteresis models for the moisture retention characteristic. Soil Sci. 157:345355. Zhou, J. & Yu, J.L. 2004. Influences affecting the soil-water characteristic curve. Journal of Zhejiang University Sciences 6A(8):797804.
This paper reports a theoretical and experimental investigation of water retention phenomena in soils involving deformation effects. The results from an exhaustive experimental campaign on retention behaviour of soil was synthesized. The results were presented through the variation of the degree of saturation, the void ratio, and the water content with respect to the suction in order to get the complete information on the effect of suction loading on volumes of water and voids. The void ratio was noticed to exert a clear influence on the air entry value and the overall shape of the retention curves. In the plane water content versus suction, from a certain suction value the retention curves corresponding to different densities of the same soil converge to become unique. The conclusions done on the experimental data were conceptualized into a new modelling framework accounting for the evolution of the
490
Study on the soil-water characteristic curve including stresses of an unsaturated soil

Chun-Ni Shen
School of Civil Engineering and Architecture, Chongqing University of Science and Technology, Chongqing, China
Xiang-Wei Fang & Zheng-Han Chen Zheng-Bin Zhou
Department of Architectural and Civil Engineering, Logistic Engineering University, Chongqing, China Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Hong Kong
ABSTRACT: The Soil-Water Characteristic Curve (SWCC) is one of the major research subjects in unsaturated soil mechanics. Traditionally, only degree of saturation (or water content) and sucion were considered in SWCC equations. A series of triaxial drained shear tests by controlling constant net mean stress and suction were conducted to study the influence of deviatoric stress on the SWCC of an unsaturated soil. A unified SWCC equation is proposed including not only water content and suction, but also net mean stress and deviatoric stress. Keywords: unsaturated soil; sucton; net mean stress; deviatoric stress; the Soil-Water Characteristic Curve (SWCC) 1 INTRODUCTION
The soil-water characteristic curve (SWCC) describes the relation between matric suction (s ua uw) and degree of saturation Sr (or the water content), for a given soil (Fredlund & Rahardjo 1993; Leong & Rahardjo 2003). Because of its broad practical applications (Fredlund & Xing 1994; Fredlund et al. 1996; Vanapalli 1996; Rahardjo & Leong 1997), the SWCC has long been one of the major research subjects in soil science and unsaturated soil mechanics. However, few study was conducted to the influence of deviatoric stress on the SWCC. Therefore, the objective of this paper is to study the influence of deviatoric stress on the SWCC by unsaturated triaxial drained shear tests with constant net mean stress and suction. A unified SWCC equation is proposed to consider the effects of deviatoric stress on the SWCC. 2 2.1 RESEARCH METHOD Definitions of stress and volume-mass variables
p
1 2 3 ua 3
(1) (2) (3)
q 1 3 s ua uw
where 1, 2 and 3 are the principal stresses; ua and uw are the pore-water pressure and the pore-air pressure, respectively; p, q and s are the net mean stress, deviatoric stress and the matric suction, respectively. Volume change associated with the soil structure and the water phase is denoted using the variables, v and w, respectively.
v
V V0 Vw V0
(4) (5)
w
The following notations are used to denote the stress state variables for an unsaturated soil
where V, Vw, and V0 are the total volume change, water volume change, and initial volume of the specimen, respectively. The specific volume v, (v 1 e, e is void ratio) and gravimetric water content w can be determined by the volumetric variables v and w, respectively, as follows
491
v (1 e0 )(1 v ) v0 (1 v ) w w0 1 e0 w ds
(6) (7)
where v0, e0, w0 and ds are the initial specific volume, initial void ratio, initial gravimetric water content and the specific gravity of soil particles, respectively. 2.2 Laboratory testing equipment The modified triaxial apparatus used in this study was developed at the Logistical Engineering University in Chongqing, China (Chen, 1999). The stress control technique was adopted which the axial load was applied by a pneumatic loading machine. 2.3 Laboratory test procedure The soil used in the study was a remolded loess from Qiaotou electric power plant, Qinghai Province, China. The specimens were prepared by static compaction with a stainless steel mold. Each specimen was compacted in five layers. The height of each layer was controlled by using stainless steel rings attached to the piston of the mold. The diameter and height of each specimen were 39.1 mm and 80 mm respectively. A total of six unsaturated triaxial tests were performed under drained conditions to study the relationship between gravimetric water content w and deviatoric stress q during triaxial tests. The initial dry density of the first three tests (SI) and the last three tests (SII) were 1.5 g/cm3 and 1.68 g/cm3, respectively. The initial conditions of the compacted specimens are summarized in Table 1. In the tests, suction and net mean stress were kept constant with increase in deviatoric stress. The controlled constant net mean stress was 150 kPa for all six tests. Both series I (SI) and series II (SII) include three tests each. For each series of test constant suction was kept at 75 kPa, 100 kPa and 200 kPa. For each test, the deviatoric stress was increased until failure. Axis translaion technique was used by controling constant air pressure ua and constant pore water pressure uw (zero). Equalization during each increment of deviatoric stress or suction was assumed to have been
Table 1. Test series SI SII
attained when the rates of volume change for both the soil structure and the water phase were less than 0.012 cm3 over an interval of 2 h. Equalization was typically attained after about 23 days. Each test took about 15 days to complete. At the end of each test, the specimen was divided into three parts. The water content was measured for each part and the water contents were found to be approximately equal. From the water content measured at the end of the test and the initial water content values (i.e., w0 16%, see Table 1), the amount of water that was drained out of the specimens could be calculated. The water volumes measured during the test were corrected according to the elapsed time into the test, since there were slight differences between the two sets of measurements. Some test data are shown in Table 2. The difference between the measured water volume changes and the corrected water volume changes is insignificant. The corrected water volume changes are adopted in this paper. 3 3.1 ANALYSE OF WATER CONSTENT Change of water constent during consolidation
During consolidation, the net mean stress was kept at constant with zero deviatoric stress. Huang et al (2000) developed the SWCC considering effects of both p and s and proposed an equation as follows s pat w w0 ap b ln pat (8)
where a and b are constants; pat is atmospheric pressure. The constants a and b are found by curve-fitting of equation (8). For specimen with dry density of 1.68 g/cm3 (SI), a and b are 1.47e5 kPa1 and 0.0329, respectively. While for specimen with dry density of 1.50 g/cm3 (SII), a and b are 1.87e5 kPa1 and 0.0353, respectively. Figure 1 shows the fitted curves from equation (8). Experimental data is also included in this figure. It seems that the equation represents the relationship among water content, suction and net mean stress very well. Eq (8) implies an assumpt that suctions during consoldation were greater than the the initial suction(S0), so that, water can be drained out of specimen of unsaturated soil.
The initial conditions of the compacted specimens. Dry density d/103 kgm3 1.68 1.50 Specific gravity ds 2.72 2.72 Water content w0/% 16 16 Void ratio e0 0.62 0.81 Degree of saturation Sr /% 70.2 53.7
492
Table 2.
Amount of water drained from the test specimens and calculated from the initial and final water contents. Measured value/cm3 Calculated value/cm3 Shear 1.10 1.18 1.35 1.28 1.31 1.44 Total 4.29 5.34 7.61 4.26 5.18 7.55 Consolidation 3.16 4.11 6.24 3.00 3.83 6.08 Shear 1.13 1.23 1.37 1.26 1.35 1.47 Difference/cm3 0.11 0.2 0.11 0.06 0.17 0.15 Relative error/% 2.56 3.75 1.45 1.41 3.28 1.99
Test series SI SI-1 SI-2 SI-3 SII-1 SII-2 SII-3
s/kPa 75 100 200 75 100 200
Total 4.18 5.14 7.50 4.32 5.01 7.40
Consolidation 3.08 3.96 6.15 3.04 3.70 5.96
SII
15 Dry density=1.68g /cm 3(SI) Dry density=1.50g /cm 3(SII) 14 w =w 0-1.47e-5p -0.0329ln[(s +p at)/p at] w /% 13
w /%
16
s=75kPa s=100kPa s=200kPa
14
12
12
w =w 0-1.87e-5p -0.0353ln[(s +p at)/p at]

10
11 0.4 0.6 0.8 ln [(s+p at )/p at ] 1 1.2
100
200 q /kPa
300
400
Figure 1. Change of water content curves during consolidation.
(a) d 1.68 g / cm 3 (SI)

16 s=75kPa s=100kPa s=200kPa
14 w /% 12 10 0 100 200 q /kPa 300 400
3.2
Change of water content during shear
Figure 2 shows the change of water content during shear. The w q relationship can be approximated by a straight line and the slope ((s)) at different suctions is shown in Table 3. The changes of water content of specimen with dry density at d 1.50 g/cm3 were greater than those at d 1.68 g/cm3. This demonstrates that the dry density affects the rate of changes of water content during shear. This is because the structure of soils with smaller density is easy to be damaged. For the specimen with same density but different suctions, (s) are very close. The (s) for the specimen with same density is denoted by by the average value. The volumetric water content increment can be writen as dw dp ds dq Kwpt Hwt Kwqt (9)
(b) d 1.50 g / cm 3 (SII)
Figure 2. w q curves during shear.
where the subscript t indicates tangent; Kwpt, Hwt and Kwqt are tangent bulk modulus of water related to net mean stress, suction and deviatoric stress, respectively. Kwpt and Hwt can be determined based on the method proposed by Chen et al (1999); Kwpt is constant; Hwt ln10 s Pat /w(p), where w(p) is constant.
Following the procedure of determining tangent bulk modulus of water related to deviatoric stress, Kwpt will be addressed. For specimen with dry density of 1.68 g/cm3, the average value of (s) is 2.09e5 kPa1. While for specimen with dry density of 1.50 g/cm3, the average value of (s) is 3.07e5 kPa1. From equation (7), it can be obtained 1 e0 d w dw dq d s dq (10)
Equation (10) reflects the relation between the change of the gravimetric water content and volumetric water content. When water is drained out
493
Table 3. Test series
The values of (s). SI 75 1.98 2.09 100 2.09 200 2.19 SII 75 3.02 3.07 100 3.17 200 3.02
s/kPa (s)/105 kPa1 AV/105 kPa1
of the specimens, dw 0 and w decreases. When water is absorbed into the specimens, dw 0 and w increases. At the critical state, both gravimetric and volumetric water content remain constant. Substituting dw/dq (s), Kwqt dq/dw into equation (10) yields Kwqt 1 e0 d s ( s ) (11)
CONCLUSIONS
Kwqt can be obtained by substituting e0, ds and (s) into the above equation(11). For specimens with dry density of 1.68 g/cm3 (SI) and 1.50 g/cm3 (SII), Kwqt are 28.5 MPa and 21.7 MPa, respectively. Hence, Kwqt is related to dry density of specimens. Substituting Kwqt, Kwpt and Hwt into equation (9), w can be obtained by integration as follows
1. During the triaxial drained shear tests with constant net mean stress and suction, the influence of deviatoric stress on the SWCC of unsaturated soil is obvious. 2. The proposed unified SWCC equation includes not only water content and suction, but also net mean stress and deviatoric stress. It represents the relationship among water content, suction, net mean stress and deviatoric stress well. In this paper, the influences of deviatoric stress on SWCC of unsaturated soil are studied during drying. The influences during wetting should be studied further in the future. ACKNOWLEDGEMENTS The financial support received from the Natural Science Foundation Project of ChongQing through grant CSTC, 2009BB6174 is gratefully acknowledged. REFERENCES
Chen, Z.H., Fredlund, D.G. & Gan, J.K.M. 1999. Overall volume change, water volume change, and yield associated with an unsaturated compacted loess. Canadian Geotechnical Journal 36(5), 321329. Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils. John Wiley and Inc. New York. Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater characteristic curve. Canadian Geotechnical Journal 31(4), 521532. Fredlund, D.G., Xing, A. & Fredlund, M.D. 1996. Relationship of the unsaturated soil shear strength to the soil-water characteristic curve. Canadian Geotechnical Journal 33(3), 440448. Huang, H., Chen, Z.H. & Li, G. 2000. Study on yield locus of unsaturated soils on p-s plane and soil-water characteristic curve. Rock and Soil Mechanics 21(4), 316321. Leong, E.C. & Rahardjo, H. 1997. Review of soil-water characteristic curve equations. Journal of Geotechnical and Geoenvironmental Engineering 123(12), 11061117. Rahardjo, H. & Leong, E.C. 1997. Soil-water characteristic curves and flux boundary problems. Geotechnical Special Publication, 88112.
w
p K wpt
w ( p ) s pat q ln ln10 pat Kwqt
(12)
Substituting equation (12) into equation (7) yields w w0 1 e0 ds p ( p ) s pat ln w K ln10 pat wpt (13)
q K wqt
The unified form of SWCC can also be written as w w0 ap b ln( in which a 1 e0 (1 e0 )w ( p ) 1 e0 ,c ; ,b d s Kwqt d s ln10 d s Kwpt s pat ) cq pat (14)
where a, b and c are constants; pat is atmospheric pressure. The constant c is found by curve-fitting of equation (14). For specimen with dry density of 1.68 g/cm3, c is 2.09e5 kPa1. While for specimen with dry density of 1.50 g/cm3, c is 3.07e5 kPa1. It should be pointed out that, the specimen in this study is contractive. The unified form of SWCC can be obtained by integration once. While for the specimens which is contractive initially and then dilative, eq (9) should be integrated in different stages.
494
Unsaturated hydraulic characteristics of soil with significant oversize particles

S. Somasundaram & T. Shenthan C. Benson
Advanced Earth Sciences, Inc., Irvine, California, USA University of Wisconsin, Madison, Wisconsin, USA
S. Nannapaneni
Los Angeles County Sanitation Districts, Whittier, California, USA
ABSTRACT: Development of Soil Water Characteristic Curves (SWCC) and unsaturated hydraulic conductivity functions for materials containing a significant fraction of coarse granular material, including gravels, cobbles and small boulders, is constrained by the limitations of laboratory equipment and test specimen sizes. This paper presents an approach for obtaining the unsaturated hydraulic parameters for such materials by testing the matrix materials that primarily govern the hydraulic behavior, and developing and applying oversize correction functions to account for the influence of the larger particles and matrix density/porosity. Laboratory SWCC and saturated hydraulic conductivity (Ks) tests on minus #4, minus 25-mm and minus 75-mm fractions of desert alluvium, mine ore, and mine overburden materials were used to develop the oversize correction functions for Ks, and the van Genuchten/Mualem parameters , n, s and r. 1 INTRODUCTION area of southern California, were evaluated and used to develop appropriate oversize correction functions. Based on combined field/laboratory gradations on large bulk samples (weighing up to 2000 kg each for the field gradation), the materials were classified as sands with gravel/gravels with sand with varying amounts of cobbles and fines. The alluvial fan materials were characterized as well graded materials falling within a narrow gradation range with average 3% cobbles, 36% gravel, 53% sand and 8% fines. The grain size distribution of the mine overburden soils varied over a wide range; on average they consisted of 5% boulders, 19% cobbles, 38% gravel, 30% sand and 8% fines. The crushed mine ore materials obtained from a heap leach pad complex consisted of 38% gravel, 44% sand and 18% fines, on average. The objective of the study was to evaluate the feasibility of using these materials for the final cover system subject to four levels of potential material processing or screening: minus 25-mm (1-inch), minus 75-mm (3-inch), minus 300-mm (12-inch) and unprocessed. The feasibility evaluation included modeling the percolation control performance of final covers constructed with the processed and unprocessed materials utilizing a
Materials containing significant oversize particles (gravels and cobbles) up to 300 mm in size are used for evapotranspirative final covers on landfills and mine waste piles in desert climates. Development of Soil Water Characteristic Curves (SWCC) and unsaturated hydraulic conductivity functions required to model the hydraulic performance of such materials is constrained by the limitations of laboratory equipment and test specimen sizes. Although the unsaturated hydraulic characteristics of these materials are primarily governed by the matrix materials, they are heavily influenced by the extent and distribution of the oversize particles. The typical approach to characterizing such materials is to test the matrix materials in the laboratory and then apply correction functions to account for the influence of oversize particles. However, previous work (e.g. Gardner, 1986, Bouwer & Rice, 1984, and Mehuys et al, 1975) on rock and gravel corrections are limited and not adequately validated. Three material sources, consisting of mine overburden soils, crushed ore and alluvium with significant oversize content, that are being considered for use as evapotranspirative final cover soils for a municipal solid waste landfill in an arid
495
numerical unsaturated flow model (UNSAT-H). Material input parameters required for the modeling included the saturated hydraulic conductivity Ks and the unsaturated hydraulic characteristics represented by the van Genuchten (1980)/Mualem (1976) parameters , n, s and r. 2 LABORATORY TESTING APPROACH
The key constraints associated with adequately characterizing these materials were: Material Variability: The gradation of the potential cover soils is highly variable; a procedure had to be developed to obtain test samples that are repeatable, representative, and reflect the variability. Oversize Materials: Typically for laboratory hydraulic testing, the specimen diameter has to be at least 4 to 6 times the maximum particle size. The largest diameter specimen that could be conveniently tested, even with specially fabricated equipment, was 300 mm (12 inches). Thus the largest particle size in the laboratory specimen had to be limited to 75 mm (3 inches). Considering the above constraints, laboratory test samples from each of the three sources were created by scalping the plus 75-mm (3-inch) fraction, separating the various particle sizes by dry sieving, and reconstituting (batching) by mixing the individual particle size ranges in predetermined proportions to yield a material that matched the average gradation curve of the parent material. This process yielded consistent representative samples of the parent material suitable for testing. The material variability was represented by reconstituting samples with upper bound (fine range) and lower bound (coarse range) gradations corresponding to the Mean 1 Standard Deviation (SD) and Mean1SD grain size distribution curves. To quantify the impacts of oversize material on the hydraulic characteristics (Ks and SWCCs), the reconstituted samples were screened to yield three progressively increasing particle size ranges: minus #4 fraction (tested in a 50-mm mold), minus 25-mm fraction (tested in a 75-mm mold) and minus 75-mm fraction (tested in a 300-mm mold), each remolded to 85% relative compaction (with rock correction applied in accordance with ASTM D4718). In addition, the minus 25-mm fraction was also tested at 90% relative compaction. Results of the test program on progressively coarser materials were utilized to develop the correction functions, which were then used to extrapolate the results to materials containing larger size particles that could not be tested with conventional laboratory equipment.
The SWCC (relationship between the soil suction and water content) was obtained using a combination of the hanging column, pressure plate, water activity meter, and relative humidity box methods for progressively increasing ranges of soil suction (ASTM D6836). Saturated hydraulic conductivity was obtained using a flexible wall permeameter (ASTM D5084) at an effective confining pressure of 14 KPa (2 psi). Typical results from the test program for all three materials are presented in Table 1. The SWCCs for one of the materials (crushed ore) are plotted in Figure 1 to illustrate the effect of oversize content and relative compaction on the shape of the SWCCs. 3 CORRECTION FUNCTIONS
Correction functions were developed for Ks, volumetric moisture content , and the van Genuchten parameters and n. To develop these corrections, the matrix material had to be defined. The matrix material is the predominant size fraction in the mix, and the oversize particles are randomly distributed as inclusions that are sparse enough so that they do not form clusters or have excessive particle to particle contact. Under such conditions the matrix materials dominate the hydraulic characteristics. Conventionally the minus #4 fraction (sands and fines) is defined as the matrix. However, for the materials considered in this study, the oversize content is relatively large and the proportion of the minus #4 fraction (35% to 62% on average) was considered too low. The minus 25-mm fraction (60% to 91% on average) was selected instead. With the minus 25-mm fraction as the matrix material, the correction functions for finer materials (the minus #4 fraction) become reverse corrections.
Figure 1. Effect of oversize particles and relative compaction on SWCCs (Crushed Ore).
496
Table 1.
Results of unsaturated hydraulic characteristics tests. 85% Relative compaction 90% Relative compaction Minus 75-mm 6.2E-04 0.320 0.000 0.039 1.458 4.2E-04 0.280 0.000 0.016 1.813 3.5E-04 0.286 0.000 0.051 1.425 Minus 25-mm 2.5E-03 0.274 0.000 0.093 1.177 4.8E-05 0.324 0.000 0.041 1.209 1.2E-04 0.252 0.000 0.070 1.168
Material Alluvium
Parameter Ks (cm/s) s (vol/vol) r (vol/vol) (cm1) n Ks (cm/s) s (vol/vol) r (vol/vol) (cm1) n Ks (cm/s) s (vol/vol) r (vol/vol) (cm1) n
Minus No. 4 3.2E-03 0.321 0.000 0.075 1.365 4.7E-05 0.382 0.000 0.028 1.411 2.8E-05 0.377 0.000 0.024 1.431
Minus 25-mm 1.9E-03 0.327 0.000 0.114 1.425 1.6E-04 0.313 0.000 0.053 1.409 1.3E-04 0.304 0.000 0.183 1.399
Mine Over-burden
Crushed Ore
Notes: Ks Saturated hydraulic conductivity; s Volumetric water content at saturation; r Residual volumetric water content; and n van Genuchten parameters.
3.1
Correction for Ks
The conventional relationship between the saturated hydraulic conductivity Ksm of the matrix and the saturated hydraulic conductivity Ksb of the bulk material (matrix and oversize mixture) is based on the assumption that Ks is directly proportional to the void ratio and can be expressed as (Bouwer & Rice, 1984): Ksb Ksm eb em (1)
Based on the test data the following correction function is proposed for Ks: Ksb Ccb Ksm Ccm
0.5
eb 1 em . D60 b . D60 m em 1 eb
3
(2)
where eb and em are the void ratio of the bulk material and the matrix material, respectively. A more rational correction function is proposed herein, based on the following assumptions: Ks is directly proportional to e /(1 e), where e is the void ratio. This is consistent with the Kozeny-Carman expression (Lambe & Whitman, 1969) for coarse grained soils. Ks is proportional to the coefficient of curvature (Cc) of the grain size distribution curve given by the ratio (D30)2/(D10 D60) where D60, D30 and D10 are the particle sizes corresponding to 60 percent, 30 percent and 10 percent finer on the grain size distribution curve. Cc is a measure of the breadth of the particle gradation. Ks is directly proportional to the particle size (represented by D60)
3
Figure 2 shows the measured and predicted values of Ks based on the correction function postulated above and the Bouwer and Rice (1984) equation. The proposed correction function provides better predictions, particularly for materials with Ks values less than 1.0 103 cm/sec. 3.2 Correction function for oversize inclusions The relationship between the volumetric water content b of the bulk material and volumetric water content m of the matrix material can be expressed as (Bouwer & Rice, 1984): b m (1 Vr) (3)
where Vr is the volume fraction of the oversize material in the bulk material. This is based on the assumption that there is no moisture stored in the oversize particles. Laboratory absorption tests on the oversize materials showed percentage absorption ranging from 1.41 to 1.97. The assumption of no moisture storage in the oversize would only result in up to 5% error in Vr and negligible error in the estimate of b. Application of this
497
of and n can be attributed to, and correlated with the change in the matrix/bulk density of the material. It should be noted that a minus 1-inch sample remolded to 85% relative compaction and a minus 3-inch sample remolded to 85% relative compaction do not have the same matrix density even if the maximum dry density value is corrected appropriately for oversize content. When the bulk material is compacted to corrected maximum dry density, (db)max, the matrix density is given by:
d ,matrix d , matrix max 1 fr .Gsr . w . db max Gsr . w fr . db

max
(4)
where Gsr is the bulk specific gravity of oversize; fr is the oversize fraction by weight; and w is unit weight of water. When the bulk material is compacted to a relative compaction of RC, the corresponding matrix density is given by:
d ,matrix
1 fr .Gsr . w . RC . db max Gsr . w fr . RC . db max
(5)
The matrix density may be represented in terms of the matrix porosity matrix (inversely proportional to matrix density). The matrix porosity may be expressed as:
Figure 2. Comparison of predicted and measured hydraulic conductivity values.
matrix 1
1
d , matrix Gs, matrix . w 1 fr . RC . db
correction results in an overall shift of the matrix SWCC (Fig. 1) to the left. As a result of the change in b, the van Genuchten parameter s changes by the same proportion, i.e. (s)bulk (1 Vr) . (s)matrix. The van Genuchten parameter r being close to or equal to zero, remains unchanged by the oversize correction. Fitting the oversize corrected SWCC to the van Genuchten equation resulted in negligible changes to the other two parameters and n. 3.3 Matrix and bulk density corrections
Gsr max (6) . . . . Gsr w fr RC db max Gs, matrix
Comparing the test data on the minus 25-mm and minus 75-mm fraction of the same material compacted at the same relative compaction (Table 1, Fig. 1), it is apparent that the oversize correction (shifting of the SWCC to the left) described in Section 3.2 would account for the change in s but not the values of and n (i.e. changes to the shape of the SWCC). The changes to the values
where Gs,matrix is the specific gravity of matrix material. The influence of matrix density or matrix porosity on and n was evaluated by plotting the measured parameters for the minus 1-inch sample (compacted to 85 and 90 percent relative compaction) and the minus 3-inch sample (compacted to 85 percent relative compaction), against the calculated matrix porosity (matrix) for each of the three materials. The matrix porosity decreases with increasing relative compaction and increases with oversize content for a given relative compaction. Figure 3 shows the normalized values and n (normalized with respect to the baseline values 0 and n0, representing the matrix material remolded to 85 percent relative compaction) plotted against the normalized porosity of the matrix. The parameter n (which represents the slope of the mid portion of the SWCC) increases with increasing matrix porosity (decreasing matrix density) for all
498
The plot shows a trend, but significant scatter. Based on this data, a second order correlation is proposed below as the correction function for : 4.0 s 4.0 s 1.0 0 s0 s0 4 PREDICTED HYDRAULIC CHARACTERISTICS WITH OVERSIZE CORRECTIONS
2
(8)
Figure 3. porosity.
Normalized and n versus normalized matrix
Figure 4.
Normalized versus normalized porosity.
three materials. The parameter (which is approximately equal to the inverse of the air entry suction) shows no clear trend. Based on Figure 3, the following expression is proposed for correcting the n-parameter as a function of the matrix porosity:
s,matrix n 1.6 0.6 s 0,matrix n0
(7)
Table 1 shows that for all three materials the parameter decreases with increasing relative compaction as well as with increasing oversize content (for a given relative compaction). In both cases (increasing relative compaction and increasing oversize content) there is a reduction in bulk porosity. This observation suggests a correlation between and the bulk porosity of the material. Figure 4 shows normalized (normalized with respect to the baseline value 0) plotted against the normalized value of the bulk density s/s0.
The corrections developed above were applied for each of the cover material sources to predict the unsaturated hydraulic characteristics of cover materials subject various levels of processing. The predictions are illustrated with respect to one of the material sources (mine overburden materials) below. The mine overburden soils were evaluated for three potential levels of field processing (screening) prior to use as the final cover material: minus 25-mm (baseline case); minus 75-mm and minus 300-mm. The corresponding average particle size distributions of the processed materials are summarized in Table 2. Table 3 summarizes the predicted unsaturated hydraulic characteristics for the mine overburden soils processed to a minus 75-mm and minus 300-mm size, and compacted to 85% and 90% relative compaction. The predictions were based on the baseline parameters determined in the laboratory on the minus 25-mm fraction remolded to 85% relative compaction. The baseline parameters were corrected in accordance with the oversize and matrix/bulk porosity correction functions developed in Section 3.0. The corresponding SWCCs are illustrated in Figure 5. Both Table 3 and Figure 5, also show the laboratory results for the minus 75-mm fraction (remolded to 85% relative compaction) for comparison with the predicted values. The laboratory determined parameters are shown in plain type whereas the predicted parameters are shown in italics in Table 3. Comparison of the predicted and laboratory determined parameters for the minus 75-mm fraction compacted to 85% relative compaction (Table 3 and Fig. 4) indicates a good match
Table 2. Mine overburden material gradation after processing. Cobbles (%) Minus 25-mm Minus 75-mm Minus 300-mm 0.0 0.0 12.9 Gravel (%) 42.2 57.7 50.2 Sand (%) 43.0 31.4 28.2 Fines (%) 14.8 10.9 8.7
499
Table 3.
Predicted hydraulic parameters with oversize correctionmine overburden example. 85% Relative compaction 90% Relative compaction Minus 75-mm 3.5E-04 0.278 0.000 0.032 1.647 Minus 300-mm 6.0E-04 0.264 0.000 0.025 1.755 Minus 75-mm 1.9E-04 0.236 0.000 0.014 1.294 Minus 300-mm 3.1E-04 0.220 0.000 0.009 1.380
Parameter Ks (cm/s) s (vol/vol) r (vol/vol) (cm1) n
Minus 25-mm 1.6E-04 0.313 0.000 0.053 1.409
Minus 75-mm 4.2E-04 0.280 0.000 0.016 1.813
Note: Values in Italics were predicted using the correction functions proposed in this paper.
crushed mine ore). The proposed correction function for Ks incorporates the void ratio, D60 and the coefficient of curvature of the gradation curve for the bulk material. For the SWCCs, the conventional Bouwer & Rice (1984) oversize correction for does not adequately account for changes in the shape of the SWCC. Correction functions for and n, based on the bulk and matrix porosity, respectively, are proposed. The proposed correction functions are able to incorporate the effects of oversize content as well as remolding density.
Figure 5. Predicted SWCCs of mine overburden with oversize correction.
REFERENCES
American Society of Testing and Materials (ASTM), 2009. Annual Book of ASTM Standards, Soil and Rock, Volumes 4.08 and 4.09, ASTM International, West Conshohocken, Pennsylvania. Bouwer, H. & Rice, R.C. 1984. Hydraulic Properties of Stony Vadose Zones. Ground Water 22(6). Gardner, W.H. 1986. Water Content. In A. Klute (ed.), Methods of Soils Analysis, Part 1, Am. Soc. Agron., Madison, Wisconsin: 493544. Lambe, T.W. & Whitman, R.V. 1969. Soil Mechanics. New York:Wiley. Mehuys, G.R., Stolzy, L.H., Letey, J. & Weeks, L.V. 1975. Effect of Stones on the Hydraulic Conductivity of Relatively Dry Desert Soils. Soil Science Society of America Proceedings 39: 3742. Mualem, U. 1976. A New Model for Predicting the Hydraulic Conductivity of Unsaturated Porous Media. Water Resources Research 12: 513522. van Genuchten, M.T. 1980. A Closed Form Equation for Predicting Hydraulic Conductivity of Unsaturated Soils. Soil Science Society of America 44: 892898.
considering that the correction functions are not-material specific and were developed (averaged) on the basis of tests on 3 different material types. The predictions also demonstrate the ability of the correction functions to incorporate the oversize content as well as variations in remolding and matrix densities. The predicted parameters suggest that Ks increases slightly with increasing oversize content and decreases with increasing relative compaction. The parameter s (bulk porosity) decreases with increasing oversize content and increasing relative compaction. The parameter (inverse of air entry suction) decreases with increasing oversize content and relative compaction. The parameter n increases with increasing oversize content and decreases with increasing relative compaction. 5 CONCLUSIONS
Oversize correction functions for saturated and unsaturated hydraulic parameters were developed and validated for three different types of sand/gravel materials (desert alluvium, mine overburden and
500
Unsaturated hydraulic properties of fine-grained soil from the blanket layer of old railway lines in France
V.N. Trinh, A.-M. Tang, Y.-J. Cui, J.C. Dupla & J. Canou
Ecole des Ponts ParisTech, France
N. Calon, A. Robinet & L. Lambert

Direction of EngineeringSNCF, France
O. Schoen
Railway network of FranceRFF, France
ABSTRACT: In the structure of old railway lines in France, a blanket layer was created mainly by interpenetration of subgrade soil into the ballast layer under train action since the construction of the lines. The hydraulic proprieties of this layer are essential in evaluating the stability of the old railway platform after heavy rainfalls upon train circulations. In the present work, the blanket layer soil taken from an old railway platform situated in Snissiat (near Lyon, France) were characterised in laboratory. The Atterberg limits, methylene blue value were determined on the fine-grained portion, whereas analysis of the grain size distribution covered the whole grain elements of different dimensions for both the blanket layer soil and the subgrade soil. The blanket layer soil is constituted of large grain elements and of fine grains from subgrade. Estimating that mainly the fine-grained fraction governs the permeability of the blanket layer, the unsaturated hydraulic properties of the blanket layer soil passed through 2 mm sieve were studied. The water retention curve was measured by the filter paper method. Unsaturated hydraulic conductivity was determined by the instantaneous profile method using an infiltration column of 50 mm in diameter and 200 mm high. Further developments planned for the analysis of the properties of the blanket layer soil are discussed. 1 INTRODUCTION Siswosoebrotho et al., 2005; Ekblad & Isacsson, 2007). Estimating that mainly the fines smaller than 2 mm (25% of the blanket layer soil) govern the permeability of the blanket layer, in the present work, the unsaturated hydraulic properties of the blanket layer soil sieved through 2 mm sieve were studied. In order to determine the hydraulic conductivity of unsaturated soils, three methods are usually used in laboratory (Delage & Cui, 2000): the Gardners method (Gardner, 1956) determines the hydraulic conductivity in transitory conditions. It consists in measuring the time evolution of the water volume that moves out of the sample under a gas pressure increment. The Coreys method (Corey, 1957) determines the hydraulic conductivity in stationary conditions. A constant suction is applied to the whole sample by means of axis translation technique. The hydraulic conductivity is determined from the water volume evolution measured under a gradient of pressure applied to each fluid. The instantaneous profile method (Daniel, 1982) consists in measuring the variation of the soil suction profile as a function of time during the infiltration.
In the structure of the old railway platform in France, a blanket layer, equivalent to the subballast layer on new lines (after UIC 719R, 2003), was created mainly by the interpenetration of subgrade soil into the ballast layer under train circulation action. Obviously, the hydraulic properties of this layer are essential in evaluating the stability of the old railway platform after heavy rainfalls upon train circulations. The grain size distribution analysed in laboratory shows that the blanket layer soil corresponds to a heterogeneous material that is very compact and constituted of large grains (up to 60 mm) from the ballast layer and of fine grains from the subgrade layer. The fine grains lower than 0.08 mm represent 16% of the blanket layer (Trinh, 2008) against 48% in the sub-ballast (GNT 0/31.5 mm) of the new lines (IN 3278, 2006; ST N 590B, 1995). It is well known that the properties of fine grains and its content govern the hydraulic characteristics of heterogeneous soils (Ct & Konrad, 2003;
501
In this study, the instantaneous profile method was used. The variation of the suction profile during the infiltration was measured within an infiltration column of 50 mm in diameter and 200 mm high. The suctions were measured by five highcapacity tensiometers (Cui et al., 2008) located along the infiltration column. The water retention curve determined by the filter paper method allows the determination of the water content profile from the suction profile. These data allows determination of the relationship between the unsaturated hydraulic conductivity and soil suction. 2 GEOTECHNICAL INVESTIGATIONS ON THE SOIL STUDIED
2.40 2.30
Dry density d (Mg/m3)
2.20 2.10 2.00 1.90 1.80 1.70
BL sieved at 20 mm BL sieved at 31.5 mm BL Sr = 80% Sr = 100%
2.1 Blanket layer soil The blanket layer soil taken from an old railway platform situated in Snissiat (near Lyon, France) was characterised in laboratory. The Atterberg limits, methylene blue value and mineralogical composition were determined on the fine-grained portion, whereas analysis of the grain size distribution covered the whole grain elements of different dimensions for both the blanket layer soil and the subgrade soil. The results show that the blanket layer soil corresponds to a heterogeneous material that is very compact. Figure 1 presents the particle size distribution curve of Blanket Layer soil (BL) and Subgrade Soil (SG). The curve of blanket layer soil shows that the percentage of soil lower than 2 mm was 25%. This portion is composed of sand, degraded ballast, polluted materials (product of wear, slags ) and fine grains from the subgrade soil. The percentage
100 90 80
Water content w (%)
12
16
20
Figure 2.
Modified Proctor compaction curves.
Percentage passing (%)
of clay (d 0.002 mm) in the blanket layer soil is 5%. On the other hand, the clay fraction in the subgrade soil is as high as 50%. The compaction properties of the blanket layer soil under modified Proctor energy was also studied using the soil sieved at 20 mm. The required corrections were made to take into account the effect of large grain elements. It was observed that the maximum dry density obtained in laboratory (2.28 Mg/m3, Figure 2) was slightly lower than that measured in the field (2.39 Mg/m3). Note that the dry density in the field was measured using the membrane densitometer method. The difference between the maximal density obtained in laboratory and in-situ value shows the difficulties associated with the preparation of a representative soil samples in laboratory by manual compaction. For this reason, the sample of soil sieved through 2 mm compacted at the in situ density was used to estimate the hydraulic conductivity of the blanket layer. 2.2 Blanket layer soil sieved at 2 mm
70 60 50 40 30 20 10 0 0.0001 0.001 0.01 0.1
BL-1 BL-2 BL-3 BL-4 SG-1 SG-2
10
100
Particle size (mm)
Figure 1. Particle size distribution curve of blanket layer soil and subgrade.
The Atterberg limits and the methylene blue value were determined on the fine-grained portion lower than 0.1 mm for the blanket layer and subgrade. The plasticity indexes are 14.8% for the blanket layer and 24.1% for the subgrade. The methylene blue values of blanket layer soil and subgrade are 3.58 g/100 g and 6.23 g/100 g respectively. Moreover, the mineralogy of these fines was analysed using X-ray diffractometry. The results show that the particles of the blanket layer soil smaller than 0.002 mm have a mineralogical composition similar to that of the subgrade soil. The carbonate content in the
502
Table 1 . Geotechnical properties. IP Soils BL sieved at 2 mm Subgrade % 10.1 24.1 VBS g 1.97 6.23 s Mg/m3 2.67 2.74 2 m % 20 50 80 m % 40 98
Suction (Log h(cm))
5 6
20
Degree of saturation Sr (%)

40 60 80
100
120 10
10
10
fines portion (d 0.1 mm) was 16% for blanket layer soil and 0% for subgrade soil (Trinh, 2008). These data show that the blanket layer soil sieved at 2 mm contains the subgrade soil (clay), sands, degraded ballast and polluted materials (hydrocarbon, product wear, etc.). In Table 1, the geotechnical properties of the blanket layer soil sieved through 2 mm and that of the subgrade layer are shown. 3 3.1 DETERMINATION OF HYDRAULIC PROPERTIES Water retention curve
10
Experimental data Van Genuchten (1980) model

0 0.1 0.2 0.3
10
Volumetric water content (cm3/cm3)
0.4
10
Figure 3. Water retention curve of blanket layer soil sieved at 2 mm.
The water retention curve was determined by the filter paper method. Ten samples were prepared at different water contents and at the equivalent in-situ dry density. In order to calculate the equivalent in-situ dry density, the grains larger than 2 mm were considered as solid particles. The density of this layer determined is 2.68 Mg/m3 (Trinh, 2008). The dry density of blanket layer soil sieved at 2 mm can be then calculated as follows:
d 2
to avoid any evaporation. After 7 days, the water content of the filter papers was determined using a balance of 0.1 mg accuracy. Finally, the soil suction was determined according to the calibration curve of ASTM D5298 (2003) following the wetting path: h 5.327 0.0779w f if w f 45.3% h 2.412 0.0135w f if w f 45.3% (2)
M M ss s V2 V Vss (1 m )dV (1 m )d ss 1.81(Mg / m3 ) (1) m m V V ss d ss d Ms2
where h is the soil suction (in logkPa); wf is the water content of filter paper. The results obtained are plotted in Figure 3 where volumetric water content (and degree of saturation) is plotted versus soil suction. The experimental data were also fitted with the function proposed by Van Genuchten (1980).
1 r ( s r ) 1 ( hw )1
where d2 is dry density of the soil smaller than 2 mm; d is the in-situ dry density of blanket layer soil (2.39 Mg/m3); ss is the solid density of particle greater than 2 mm; Ms2 and V2 are respectively mass and volume of soil smaller than 2 mm; Ms and V are respectively solid mass and volume of blanket layer soil; Mss and Vss are respectively mass and volume of solid particle greater than 2 mm; m is the percentage of solid particles greater than 2 mm (m 0.75). Ten pairs of sample of 50 mm in diameter and 20 mm high were statically compacted until the in situ dry density (d2 1.81 Mg/m3) with ten different water contents. Three Whatman N42 filter papers were embedded in each pair. The soil specimens were then carefully wrapped by plastic film and wax
where is volumetric water content; s is saturated volumetric water content; r is residual volumetric water content; hw is suction in cm; and are fitting parameters. Figure 1 shows the experimental results and the fitting curve using equation 3 with s 0.32; r 0.012; 0.23 and 0.0007. It can be noted that the soil reaches the saturation state at a suction lower than 40 kPa. 3.2 Determination of hydraulic conductivity Unsaturated hydraulic conductivity was determined by the instantaneous profile method using an infiltration column of 50 mm in diameter and 200 mm high. The soil was statically compacted in
503
Suction (kPa)
(3)
where V is the water volume infiltrated between two instants t and t t, A is the section area of the column, is the volumetric water content, H is the total height of column and zi is the current height considered. The unsaturated hydraulic conductivity k was calculated from the ratio of the water flux to the hydraulic gradient following the Darcys law as indicated in equation 6. k q 1 A 0.5(it it t ) (6)
4
Figure 4. Infiltration column for determining the unsaturated hydraulic conductivity.
the column in five layers of 40 mm at the in-situ dry density (d2 1.81 Mg/m3). Suction measurements were performed by five high-capacity tensiometers (Mantho, 2005; Munoz et al., 2008) placed at 40, 80, 120, 160 and 200 mm from the base of the column. Two metallic discs of 40 mm thickness covered the two ends of the cylinder. Two taps in the bottom and one tap in the upper end allow the connexion to water source or air source (Figure 4). The high capacity tensiometers were saturated and calibrated in a saturation cell filled with demineralized and de-aired water. A positive pressure of 4 MPa was applied for 24 hours by means of a pressure-volume controller. The calibration curves of the tensiometer were determined in saturation cell with positive pressure. The suction value (negative pressure) was extrapolated from these calibration curves. The electrical signals were automatically recorded by a computer. The suction measured in the infiltration column was used to determine the water content isochrones using the water retention curve. The hydraulic gradient was calculated from the suction isochrones as follows: i z (4)
Figure 5 shows the suction measured by the tensiometers in the column. After installation of the tensiometers T1, T3, T4, T5, and T7, it was observed that the initial suction of the soil was equal to 2045 kPa. This low value is related to the relatively high initial water content of the soil at compaction (12%, note that the water content at saturated state is 18%). In order to increase the soil suction, a pneumatic pump was used to dry the soil column from its bottom. This drying process induced a quick increase of soil suction measured by the tensiometers. The highest value of suction was observed by tensiometer T1 (z 40 mm from the bottom). This tensiometer reached cavitation at a suction of 480 kPa (at t 14 h). It was then replaced by tensiometer T9. For t 24 72 h, all the inlets of the column were closed for the soil suction homogenisation. The suction value at equilibrium was about 290310 kPa. From t 72 h, the soil column was saturated by injecting water from the bottom under a pressure of 10 kPa. The results show that the suction decreases progressively in the column. The suction measured by tensiometer
0
-100
Suction (kPa)
-200 T1 - z = 40 mm T9 - z = 40 mm T3 - z = 80 mm T4 - z = 120 mm T7 - z = 160 mm T5 - z = 200 mm
where i is hydraulic gradient, is water potential or suction, z is elevation. The water flux q between time t and t t was calculated following equation 5.
H H A t t dz t dz zi V zi q t t
-300
-400
-500 0 24 48
Temps (hours)
72
96
120
144
(5)
Figure 5. Suction measured during drying, equalization and wetting phases.
504
200 t=0h t=2h t=4h t=8h t = 12 h t = 20 h t = 32 h t = 48 h
1.0x10
-9
160
120
Hydraulic conductivity (m/s)
1.0x10
-10
Height (mm)
1.0x10
-11
80
ksat kunsat - z = 60 mm kunsat - z = 80 mm kunsat - z = 100 mm kunsat - z = 120 mm Best fit
1.0x10
-12
40
1.0x10
-13
0 0 50 100
Suction (kPa)
150
200
250
300
350
1.0x10-14
Figure 6.
200
Suction profile at different times. Figure 8. suction.

t=0h t=2h t=4h t=8h t = 12 h t = 20 h t = 32 h t = 48 h
10
100
1000
Suction (kPa)
Hydraulic conductivity as a function of
160
10
Height (mm)
120
80
Relative permeability k/k sat
10-1
Unsaturated Saturated
40
10
-2
0 0.24 0.26 0.28 0.3 0.32
10
-3
Volumet ric water content
Figure 7.
Water content profile at different times.
10-4
T9 (close to the wetting end) reached 0 after a few hours while the suction measured by tensiometer T5 (situated on the other end) reached 0 after 48 h. Figure 6 shows the suction profile measured during the wetting phase at 0, 2, 5, 10, 20, 32, 53 hours. The isochrones of water content (Figure 7) were determined from suction isochrones using the water content curve in Figure 3 and equation 3. Unsaturated hydraulic conductivity was calculated from equations 4, 5 and 6 based on the isochrones of suction and water content measured (Figures 6 and 7). The results are shown in Figure 8. The saturated hydraulic conductivity was measured at the end of the test when the soil column was totally saturated. The value obtained is ksat 9.54 1010 m/s. According to Munoz et al. (2008) and Tang et al. (2009), a linear relationship is usually observed between the logarithm of hydraulic conductivity and the logarithm of suction. In the present work, the following relationship is established:
10
-5
70
75
80
85
90
95
100
Figure 9. Relative permeability as a function of degree of saturation.
Log( k ) 1.29 Log( ) 9.72
(7)
Figure 9 shows the variation of the relative permeability as a function of degree of saturation. The unsaturated permeability increases slightly when the degree of saturation varies in the range of 7595% but it increases quickly when the soil approaches the saturated state. The value of the saturated hydraulic conductivity determined for the blanket layer soil sieved at 2 mm is in the same range for clayey soils. Actually, as the dry density of the compacted sample was high (d 1.81 Mg/m3), the porosity of the soil was quite low (0.32) in field condition.
505
Ct and Konrad (2002) shows that the hydraulic characteristics of granular base-courses depends on the fines content smaller than 80 m (%F), the porosity of soil (n) and the specific surface area of the fines fraction (Ssf). They developed empirical relationships to determine the hydraulic parameters: saturated hydraulic conductivity ks (m/s) Log( ksSsf ) 9.94 n f 12.64 (8)
1.0x10
-9
1.0x10-10
ksat- measurement kunsat - measurement Best fit-measurement k- Ct and Konrad's Model
1.0x10-11
1.0x10-12
where nf is the porosity of the fines fraction (nf n/nc); nc is the porosity of the coarsegrained skeleton (nc n (1 n)F) air entry value a (kPa) Log( a ) 3.92 5.19n f (9)
1.0x10-13
1.0x10-14
10
100
1000
pore-size distribution index ( Log()/ Log())

.65 0.385 0.021n0 f Ssf
Suction (kPa)
Figure 10. Comparison between hydraulic conductivity measured and that from the Ct and Konrads model.
(10) result presented in Figure 10 can be obtained. It can be observed that the Model used by Ct and Konrad is not suitable for the blanket layer soil passing through 2 mm sieve. In order to better understand the hydraulic properties of the blanket layer, a large infiltration column, 300 mm in diameter and 600 mm high is being developed allowing testing the whole portion of the blanket layer soil. The results obtained will be correlated with that on the portion passing through 2 mm sieve. 5 CONCLUSION
The model of Brooks and Corey (1964) was used by Ct and Konrad (2002) to assess the unsaturated hydraulic conductivity: k ks k ks a
a a
(11)
where 2 3 In order to check the model used by Ct and Konrad on the blanket layer soil, the hydraulic parameters of this soil sieved through 2 mm were calculated from equations (8), (9), (10) and compared with the measured value (Table 2). It is observed that the air entry values are similar; the measured saturated hydraulic conductivity is one order of magnitude greater than the models one. The pore-size distribution index predicted by the model is quite different from the measured one. Actually, the specific surface area of fines of the studied soil (Ssf 60.9 m2/g) is much larger than the soil tested by Ct and Konrad (2002, Ssf 2.3, 7.9 and 11.4 m2/g). If measured from the water retention curve (Figure 3) is used to calculate the unsaturated hydraulic conductivity following equation 11, the
Table 2. Hydraulic parameters. Ct and Konrads model 42 0.47 8.22 1011 Measurement 40 0.28 9.59 1010
Parameters a (kPa) ks (m/s)
Estimating that mainly the fine grains smaller than 2 mm govern the permeability of the blanket layer, the hydraulic properties of the blanket layer soil passing through 2 mm sieve were studied. The water retention curve was determined by filter paper method. Van Genuchtens model (1980) was used to fit the experimental results showing an air entry value of 40 kPa. The hydraulic conductivity of this soil was determined by the instantaneous profile method using the infiltration column of 50 mm in diameter and 200 mm high. Five high capacity tensiometers were used to measure the suction. The saturated hydraulic conductivity obtained is ksat 9.54 1010 m/s. For extrapolating these results to the properties of the blanket layer soil, a porosity model by Ct and Konrad (2003) was used to calculate the hydraulic properties. Comparison between the measures and the calculations shows that the model is not suitable for the description of the hydraulic behaviour of the soil studied.
506
In order to verify the correlation between the hydraulic conductivity of fine grains smaller than 2 mm and these of blanket layer soil, a large infiltration column is being developed. Moreover, in order to study the soil response under real field conditions, suction and temperature sensors were installed in the blanket layer on an old railway line in northern France which is being exploited. Furthermore, a meteorological station was also installed at the experimental site in order to correlate the soil response to the climatic conditions. REFERENCES
ASTM D5298 (2003) Standard Test Method for Measurement of Soil Potential (Suction) Using Filter Paper. Brooks, R.-H. & Corey, A.-T. (1964) Hydraulic properties of porous media. Hydrology Paper 3, Colorado State University, Fort Colilins, Colo. Corey, A.T. (1957) Unsaturated conductivity and diffusivity measurements by a constant flux method. Proc. Soil Sci. Soc. Am. 21(1), 710. Ct, J. & Konrad, J.M. (2003) Assessment of the hydraulic characteristics of unsaturated base-course materials: a practical method for pavement engineers. Can. Geotech. J. 40(1), 121136. Cui, Y., Tang, A., Mantho, A. & De Laure, E. (2008) Monitoring Field Soil Suction Using a Miniature Tensiometer. Geotechnical Testing Journal 31(1), 95100. Daniel, D.E. (1982) Measurement of Hydraulic Conductivity of Unsaturated Soils with Thermocouple Psychrometers. Soil Sci. Soc. Am. J. 46(6), 11251129. Delage, P. & Cui, Y.J. (2000) Leau dans les sols non saturs. Technique de lIngnieur C301.
Ekblad, J. & Isacsson, U. (2007) Time-domain reflectometry measurements and soil-water characteristic curves of coarse granular materials used in road pavements. Can. Geotech. J. 44(7), 858872. Gardner, W.R. (1956) Calculation of capillary conductivity from pressure plate out-flow data. Soil Science Society of America Proceedings 20, 317320. IN 3278 (2006) Rfrentiel technique pour la ralisation des LGVpartie Gnie Civil. Rfrentiel Infrastructure SNCF. ISSMFE (1973) Internatinal Society of Soil Mechanics and Foundation Engineering. Mantho, A.T. (2005) Echange sol-atmosphre application la scheresse. Thse de lENPC. Munoz, J.J., De Gennaro, V. & De Laure, E. (2008) Experimental determination of unsaturated hydraulic conductivity in compacted silt. Unsaturated Soils: Advances in Geo-EngineeringToll et al. (eds). Taylor & Francis Group, London, 123127. Siswosoebrotho, B.I., Widodo, P. & Augusta, E. (2005) The influence of fines contents and plasticity on the strength and permeability of aggregate for base course material. Proceeding of the Eastern Asia Society for Transportation Studies. Vol. 5, pp. 845856. ST N 590B (1995) Spcification technique pour la fourniture des granulats utiliss pour la ralisation et lentretien des voies ferres. Tang, A.M., Ta, A.N., Cui, Y.J., Thiriat, J. (2009). Development of a Large Scale Infiltration Tank for Determination of the Hydraulic Properties of Expansive Clays. Geotechnical Testing Journal 32(5), 385396. Trinh, V.N. (2008) Caractrisation des matriaux constitutifs de plate-forme ferroviaire ancienne. Rapport de rechercheSNCF. UIC 719R (2003) Earthworks and track bed for railway lines. International Union of Railways. Van-Genuchten, M.T. (1980) A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, 892898.
507
Soil-Water Characteristic Curve (SWCC) and volumetric deformation law for a plastic clay under high suction
M. Vzquez, P. Durand & J.L. Justo
Department of Continuum Mechanics, University of Seville, Spain
ABSTRACT: The Soil-Water Characteristic Curve (SWCC) for red clay of high plasticity has been determined with suction control techniques such as membrane pressure, and vapour equilibrium with acid and saline solutions. The behaviour of the sample for high suctions ( 1 GPa), applied using solutions with high acidic concentration has been observed. During the process, the sample dimensions have been measured, and the law of free volumetric deformation in wetting and drying paths has been determined. 1 INTRODUCTION
Table 1. Geotechnical properties. 95.6 27.6 1570 19.6 94.8 53 24.2
The relationship between water content and suction is of vital importance in the study of partially saturated soils. The soil-water characteristic curve represents precisely this relationship and gives information about the capacity of soil to absorb or expel water during the wetting and drying paths, and about the hysteretic behaviour existing between both paths. To obtain this curve at the laboratory, it is necessary to subject the sample to several suctions, and to obtain later the moisture content or saturation degree. Suction () can be applied using several techniques, among them the vapour equilibrium in vacuum desiccators. This method subjects the sample to a relative humidity controlled by the equilibrium vapour pressure with saline or acid solutions. High suctions applied to the sample imply high concentrations. With high acid concentrations there is a high level of acid gases in the sample environment. In this paper the behaviour of samples in contact with these gases, especially sulphuric acid that can change physic and mechanics properties of clay, has been studied.
(%) 0.08 mm w (%) d (kg/m3) (kN/m3) 0 (kPa) wL IP
100
Percent finer (%)
90 80 70 60 50 40 30 20 10 0
0.001
0.01
0.1
10
100
Grain Size(mm)
Figure 1. Particle size distribution of red clay.
was 2.25% and swelling pressure 19 kPa. The pH of the sample is slightly basic (pH 7.5), not detecting the presence of sulphates. 3 EXPERIMENTAL TESTS
MATERIAL TESTED
The material used in this study is Red Clay of High Plasticity (CH) whose geotechnical properties are presented in Table 1. Figure 1 presents the particle size distribution of this red clay. The initial matrix suction of the sample was measured using a minitensiometer UMS T5. Free swelling
3.1 Experimental techniques In order to obtain the soil-water characteristic curve, two suction control techniques have been used: the vapour equilibrium method with saline or acid solutions and the pressure membrane method.
509
Figure 2.
Dimensional control of samples. Figure 3. Soil-water characteristic curve. and Xing (1994) Wetting 1,306 1.28 1.10 200,000
In the zone of low suctions ( 500 kPa) the pressure membrane method has been used (Houston et al. 2006, Puppala et al. 2006). For the range of suctions between 1000 kPa and 150 MPa, the vapour equilibrium technique with saline solutions (NaCl, CaCl2) in vacuum desiccator and sealed containers (Barrera 2002, vila 2004, Lee et al. 2008) has been applied. In the zone of high suctions ( 150 MPa) acid solutions have been introduced in the desiccator (Pintado 2002, Khattab et al. 2006, Chao et al. 2008). Free volumetric deformation has been controlled as suction changes. To do so, sample dimensions have been checked, at the end of each suction stage, measuring diameter and thickness (Fig. 2). The equilibrium time for each suction stage ranges from ten to fifteen days. The time to measure the weight and dimensions has been quickened and gloves have been used to avoid direct contact between hands and sample. 3.2 Soil-Water Characteristic Curve (SWCC)
Table 2. Fredlund parameters.
Drying a m n r (kPa) 9,340 0.70 4.09 199,991
The soil-water characteristic curve has been obtained for a range of suctions between 50 kPa and more than 1 GPa. The results are presented in Figure 3. The wetting and drying paths have been approximated using the curves proposed by Fredlund and Xing (1994). The corresponding parameters are collected in Table 2. For suctions larger than 1 GPa, the vapour equilibrium technique has been used, with sulphuric acid solution at 98% concentration, in a vacuum desiccator. The solution used generates a suction of 1,665 MPa for a temperature of 25C (Pintado, 2002). The sample has been subjected to high levels of suction to observe the behaviour of the sample exposed to the atmosphere in equilibrium with high-concentrated sulphuric acid solutions.
The sulphuric acid sends to the atmosphere volatile gases that affect the samples changing their water retention properties. The number of references that describe the effect of highconcentrated sulphuric acid solutions and use vapour equilibrium techniques is quite small. Once the samples have been subjected to the higher suctions, a wetting process has been undertaken (Fig. 4). During this process, the samples that had been in contact with high-concentrated acid solutions, behave quite differently to those in contact with saline solutions. The samples subject to high suctions display a high degree of alteration. Many cracks appear both, in the surface and in the inside of the samples, especially in its central parts (Fig. 5). The slope of the wetting curve associated to the samples subject to highly concentrated acids solutions is steeper than the slope for samples under saline solutions. This means that the samples under acid solutions in drying paths absorb more water in a subsequent wetting than the samples subject to atmospheres in equilibrium with saline solutions. Figure 4 shows for the red clay, the drying and wetting paths associated to SWCC. Samples subject during drying to saline or acid solution present different behaviour during wetting. In the former case wetting is initiated at point A. In the latter
510
Figure 4.
SWCC. Drying and wetting paths.
Figure 6. suction.
Volumetric deformation as a function of
The measurements were carried out in different directions, and quickly to prevent changes in the water content of samples during drying and wetting processes. Figure 6 shows the evolution of volumetric deformation versus suction for wetting and drying paths. Volumetric deformations follow an exponential law of type: C e A v , where the coefficients A and C are obtained by the method of least squares (Pintado, 2002). The following relationships v are obtained: Wetting path: For 2.105 kPa: 4.9754 e 1.3617 v For 2.105 kPa: 4 47083.2062 e Drying path:
Figure 5. Disintegration of sample under high suction (point A in Figures 4 and 6).
0.1283 v
(1) (2)
For 2.105 kPa: 141.2173 e 0.8595 v
(3)
For 2.105 kPa: 469582.8476 e0.0491 v (4) There is a change in the behaviour of the law v (see Figure 6), when it passes through point associated to the residual suction ( 2.105 kPa, see Table 2). From the corresponding to the residual suction, and as suction increases, the volume change passes from contraction to expansion. This change is due to the internal rupture of the clay structure, produced by grain failure. The effect is multiplied by the presence of gases that come from highlyconcentrated acid solutions. The sample starts to disintegrate at point A (Figures 4 and 6). 5 CONCLUSIONS
case wetting is initiated at a suction of 1,665 MPa; the sample begins to disintegrate at point A (Figures 4 and 6), due to high suction and chemical changes that produce the failure of grains; point B (Figures 4 and 6) corresponds to the total rupture of the sample that cannot be handled anymore, and the experiment must stop. The pH measured after rupture of the samples is 6.5. The exposure to the acid gases changes the pH from basic (7.5) to acid (6.5). 4 VOLUMETRIC DEFORMATION WITHOUT CONFINEMENT
After each suction stage, the sample diameter and thickness were measured for monitoring the volumetric deformation without confinement (Figure 2).
Two suction control techniques have been used for obtaining the soil-water characteristic curve in plastic clay: pressure membrane and vapour
511
equilibrium with saline and acid solutions. This curve represents the drying-wetting paths and the hysteric behaviour. The samples have been subject to suction values exceeding 1 GPa. Stages of high suction with acid solutions (SO4H2 at 98%) have caused some change in the pH of the sample, passing from a slightly basic to a slightly acid. The sulphate content of the sample has not varied. The disintegration of the sample subject to high suction ( 1 GPa), by means of the vapour equilibrium technique with acid solutions, causes a change in behaviour in the wetting path, drastically increasing the slope of this path in comparison with the natural behaviour of the samples unexposed to acid. The volumetric deformation of the tested samples has been measured during drying and wetting processes. The relation v is represented by a law of the type: C e A v . For low suctions, the slope of the wetting path is greater than the slope of the drying path. The slope of the relationship v suffers a drastic change for values of suction exceeding the residual. At these high suctions there are no great differences between the drying and wetting paths. REFERENCES
vila, G.E. 2004. Estudio de la retraccin y el agrietamiento de las arcillas. Aplicacin a la arcilla de Bogot. PhD Thesis. UPC. Spain.
Barrera, M. 2002. Estudio experimental del comportamiento hidro-mecnico de suelos colapsables. PhD Thesis. UPC. Spain. Chao, K.C., Nelson, J.D., Overton, D.D. & Cumbers, J.M. 2008. Soil-water retention curves for remolded expansive soils. Unsaturated Soils:Advances in Geo-Engineering. ISBN 978-0-415-47692-8: 243248. Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater charasteristic curve. Canadian Geotechnical Journal. v31, n4:521:524. Houston, W.N., Dye, H.B., Zapata, C.E., Perera, Y.Y. & Harraz, A. 2006. Determination of SWCC using onepoint suction measurement and standard curves. Unsaturated Soils 2006. pp 14821493. Khattab, S.A.A. & Al-Taie, L.Kh.I. 2006. Determination of SWCC using one-point suction measurement and standard curves. Unsaturated Soils 2006. pp 14821493. Lee, T.T., Delage, P., Cui, Y.J. & Tang, A.M. 2008. Water retention properties of Boom clay: A comparison between different experimental techniques. Unsaturated Soils:Advances in Geo-Engineering. ISBN 978-0-415-47692-8: 229234. Pintado, X. 2002. Caracterizacin del comportamiento termo-hidro-mecnico de arcillas expansivas. PhD Thesis. UPC. Spain. Puppala, A.J., Punthutaecha, K. & Vanapalli S.K. 2006. Soil-water characteristic curves of stabilized expansive soils. Journal of Geotechnical and Geoenvironmental engineering. June 2006. pp 736751.
512
Atmospheric drying and laboratory wetting of stockpiled product coal

D.J. Williams
The University of Queensland, Brisbane, Australia
ABSTRACT: Following beneficiation, product coals are subjected to mechanical dewatering, typically by centrifuging and vacuum filtration, to meet the specified moisture content for sale. The re-combined composite product coal size fractions are then stockpiled prior to use or transport to the port for export. On the stockpile, the product coal undergoes drainage under gravity, cycles of atmospheric drying and wetting, and wetting up due to the spraying of water on the stockpile to suppress dust. In this paper, the results of laboratory characterisation testing, atmospheric drying, and laboratory simulated wetting of stockpiled product coal are presented and discussed, leading to an improved understanding of the ex-plant drying and wetting processes involved. This improved understanding can be used to optimise the design and management of product coal stockpiles to ensure the required moisture state. 1 INTRODUCTION up at the crest of the pile and the coarser particles preferentially ravelling to the base of the stockpile. Depending on the ex-plant total moisture content, which is set to achieve the contract value, the product coal may undergo further dewatering on the stockpile as a result of gravity drainage, and wind and solar action. Re-wetting of stockpiled product coal may also occur as a result of heavy rainfall, watering to suppress dust, or the direction of surface drainage towards the stockpile. Atmospheric drying and laboratory simulated wetting of stockpiled composite product coal, were carried out to improve understanding of the ex-plant drying and wetting processes involved. Further details of the testing presented in this paper are given in Williams and Whitton (2008). 2 2.1 CHARACTERISATION OF PRODUCT COAL Particle size distribution
The mining, handling, beneficiation, stockpiling, reclaiming (particularly if this involves dozing), and transportation of product coal results in the production of fines, which have an affinity for moisture. Improvements in fine coal recovery, and the mining of coal that is prone to breakdown, add to fines and moisture retention. The moisture content at which product coal is utilised or exported has implications for its handleability, stockpiling, shipment and final use. Excessive moisture makes the coal difficult to unload from reclaimers and ships, there is no financial return from supplying excess water, and excess moisture in the coal requires energy to remove it. Product coal is therefore required to be delivered for sale below a specified total moisture content (mass of water/total mass, expressed as a percentage), typically 10 to 11% (corresponding to about 35% saturated at an average stockpile dry density of 0.9 t/m3). This requires mechanical dewatering of the different product coal size fractions produced by wet beneficiation, prior to their combination in the composite product coal. For each 1% by which the composite product coal exceeds the specified contract value, a financial penalty is imposed equivalent to the energy value of the extra moisture, up to a rejection total moisture content, typically about 15%, at which rejection of the product may occur or severe financial penalties are imposed. The dewatered composite product coal is stacked on a stockpile prior to transportation and use, which causes the product coal to ravel and segregate, with the finer particles being held
The composite product coal used in the field atmospheric drying trial and laboratory simulated rainfall tests was sourced from a Hunter Valley operation in New South Wales, Australia. The particle size distribution curves of the initial sample and the two drums supplied for laboratory testing were obtained by wet sieving according to AS1289 (2000), and are shown in Figure 1. All particle size distribution curves are similar, but with the drum samples containing more fines (10% silt and clay size compared with none for the initial sample).
513
100 90 80 % PASSING 70 60 50 40 30 20 10 0 0.01 0.1 1 PARTICLE SIZE (mm) 10 100

Initial sample Drum 1 Drum 2
TOTAL MOISTURE CONTENT (%)
SILT
SAND
GRAVEL
COBBLES
35 30 25 20 15 10 5 0 0.1 1 10 100 1000 MATRIC SUCTION (kPa)
Drying data Wetting data Fitted drying curve
10000
100000
Figure 1. Particle size distribution curve for composite product coal.

1.5
Figure 3. Soil water characteristic curve data and fitted drying curve for composite product coal, in terms of total moisture content.
MATRIC SUCTION (kPa) 100 1000 10000
0.1 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 1.0E-15 1.0E-16 1.0E-17 1.0E-18 UNSATURATED PERMEABILITY (m/s)
10
100000 1000000
1.4 DRY DENSITY (t/m3)
Zero Air Voids Line for acalculated specific gravity of 1.417 (assumed ash content of 8.7%)
Base and crest fines ~ 700
Interior ~ 4,000
Face ~ 20,000
1.3
~ 2x 10 -12 ~ 1x 10 -13 ~ 6 x 10-15
1.2
1.1
1 0 2 4 6 8 10 12 14 16 18 20 GRAVIMETRIC MOISTURE CONTENT (%)
Figure 4. Calculated hydraulic conductivity function for composite product coal, in terms of matric suction.
Figure 2. Standard compaction curve for composite product coal (MDD Maximum Dry Density, OMC Optimum Moisture Content).
2.2
Dry density
The dry density of loosely-dumped composite product coal is typically of the order of 0.9 t/m3. To place this in context, a laboratory Standard compaction test was carried out according to AS1289 (2000) on the composite product coal tested, the results of which are shown in Figure 2. Laboratory Standard compaction testing gave a Maximum Dry Density (MDD) of 1.165 t/m3 at an Optimum Moisture Content (OMC) of 13.5% gravimetric ( mass of water/mass of solids, expressed as a %). Hence, the loosely-placed composite product coal is at a dry density equivalent to 77% of its MDD and, at 10.5% gravimetric moisture content, is typically about 3% dry of its OMC for Standard compaction. 2.3 Soil water characteristic curves
drying curve fitted using the method of Fredlund et al. (1994) and the SoilVision software. The SWCCs are plotted as matric suction versus total moisture content ( mass of water/[mass of water mass of solids], expressed as a %; the usual form used in mining applications), following conversion from the measured volumetric water content ( volume of water/total volume, expressed as a decimal). There is seen to be little hysteresis between drying and wetting due to the coarse-grained nature of the composite product coal, and the fitted drying SWCC is used in the subsequent analyses. 2.4 Hydraulic conductivity function
Since composite product coal exists in an unsaturated state, Soil Water Characteristic Curves (SWCCs) are an essential element in their characterisation. Drying and wetting SWCC data for the composite product coal tested were obtained using a Tempe cell with a 100 kPa air-entry disc, and are shown in Figure 3 as drying and wetting data points, and a
The hydraulic conductivity function of the composite product coal tested was calculated using the method of Fredlund et al. (1997), and scaled using the saturated hydraulic conductivity of the material obtained from a falling head permeability test conducted on a 150 mm diameter loose specimen (AS1289, 2000). The saturated hydraulic conductivity obtained was 1.6 105 m/s. Figure 4 shows the calculated hydraulic conductivity function in its usual form of matric suction versus hydraulic conductivity to a log10-scale. Figure 5 shows the relationship in the form of total moisture content versus hydraulic conductivity to a log10-scale, which is more useful for the purposes of this paper.
514
0 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 1.0E-15 1.0E-16 1.0E-17 1.0E-18 UNSATURATED PERMEABILITY (m/s)
Face
5
Interior
TOTAL MOISTURE CONTENT (%) 10 15 20 25

Base and crest fines
30
35
~ 2x 10-12 ~ 1x 10-13 ~ 6 x 10-15
Figure 5. Calculated hydraulic conductivity function for composite product coal, in terms of total moisture content.
Figure 6. 11.5 m high, 38 sloping composite product coal field stockpile.

12 11 10 9 8 HEIGHT (m) 7 6 5 4 3 2 1 0 0
6.3 12.3 12.9 8.5 11.1 11.9 10.0 12.1 9.5 7.4 10.5 12.6 8.2 7.7 8.2 7.7 7.7 7.8 7.8 6.3 7.5 6.0 5.5 3.7 7.0 6.5 6.7 6.2 6.8 5.7 5.3 6.1 5.6 4.6 3.2 5.5 12.1 12.1 12.3 11.2 10.5 9.2 5.0
5% total moisture content contour 7.5%total moisture content contour 10% total moisture content contour
7.6 7.2 7.8 3.5
3 3.1
ATMOSPHERIC DRYING TRIAL Description of trial
4.7 7.7
7.7
7.5
6.9
Figure 6 shows the 11.5 m high, 38 sloping composite product coal field stockpile prior to atmospheric drying and sampling. The stockpile was constructed in the shape of a cone by delivering composite product coal from a stacker fixed in position (the stacker is mounted on rails and is normally run along these to form a continuous stockpile with angle of repose side slopes). The ex-plant total moisture content of the composite product coal is about 9.5%. Following 8 days of drying under hot (maximum daytime temperatures of the order of 35C), dry weather, the stockpile was deconstructed by excavator to allow hand sampling for total moisture content determinations by oven-drying at 60C (the oven temperature was limited to prevent combustion of the coal) and particle size distribution analysis by sieving. 3.2 Total moisture contents after drying
10
11
12
13
14
15
DISTANCE FROM CENTRELINE (m)
Figure 7. Contours of total moisture content after 8 days of drying of composite product coal field stockpile.
SILT SAND GRAVEL COBBLES
100 90 80
% PASSING
Composite Base 1.2 m height 3.0 m height 4.3 m height 6.7 m height 8.5 m height 10.4 m height 11.5 m crest Composite Crest
70 60 50 40 30 20 10 0 0.01
Base
0.1
1 PARTICLE SIZE (mm)
10
100
Figure 8. Particle size distributions on face of composite product coal field stockpile.
Figure 7 shows the measured total moisture contents and the inferred 5%, 7.5% and 10% contours. The high total moisture contents towards the crest are due to the hanging-up of moisture-retaining fines, while those at the base are due to gravity drainage. The lowest total moisture contents are along the face of the stockpile, due to solar drying and the predominance of coarse-grained particles on the surface. The total moisture content contour values plotted in Figure 7 of 5% towards the face, 7.5% within much of the interior of the stockpile, and 10% towards the crest and base, have been plotted in Figure 5 to estimates of the hydraulic conductivities of 6 1015 m/s, 1 1013 m/s and 2 1012 m/s, respectively. Plotting these values in Figure 4 gives estimated matric suctions of about 20,000 kPa, 4,000 kPa and 700 kPa, respectively.
Hence, the stockpiled composite product coal is very effectively dried, and rendered almost impermeable, by the combined effects of gravity drainage, and solar and wind action over 8 days of hot, dry conditions. 3.3 Particle size distribution curves within stockpile
The samples taken from the dried product coal field stockpile were also subjected to particle size distribution analysis by wet sieving, the results of which are shown in Figures 8 and 9, together with the particle size distribution curve of the composite product coal (initial sample from Figure 1). Figure 8 shows that the particle size distribution curves of the materials on the face straddle the all-in (average) particle size distribution curve. The material at the crest of the stockpile is the finest
515
SILT
SAND
GRAVEL
COBBLES
100 90 80 % PASSING 70 60 50 40 30 20 10 0 0.01 0.1 1 PARTICLE SIZE (mm) 10 100

Composite 0.2 m height 0.9 m height 1.8 m height 2.7 m height 3.5 m height 4.4 m height 5.3 m height 6.2 m height 7.1 m height 8.0 m height 8.9 m height 9.7 m height 10.6 m height 11.5 m crest Towards base
Figure 9. Particle size distributions along centreline of composite product coal field stockpile.
10 TOTAL MOISTURE CONTENT (%) 9 8 7 6 5 4 3 2 1 0 0 1 2 3 4 5 D10(mm) 6 7 8 9 10
Figure 11. Slumping due to wetting up of laboratory product coal stockpile.
Figure 10. Total moisture content after 8 days of drying of composite product coal stockpile versus size through which 10% of material passes.
and the material at the base of the stockpile is the coarsest, as would be expected due to particle ravelling down an angle of repose slope under gravity. Figure 9 shows that the particle size distribution curves along the centreline of the field stockpile approximate the all-in (average) particle size distribution curve, apart from the near base, where they are much finer-grained. This may be due to some of the base material being old composite product coal that has been broken down on recurring handling. 3.4 Relationship between moisture and particle size Figure 10 shows that there is some relationship between the total moisture content and the particle size distribution of the material in the stockpile. While most of the data plotted in Figure 10 are grouped around a small range of D10 values, there appears to be a trend of decreasing total moisture content with increasing D10 value, as would be expected. 4 LABORATORY RAINFALL SIMULATIONS
placed loosely in the tank to form an embankment with a height of 0.75 m and an angle of repose outer slope (initially at about 35). The top of the embankment had a nominal 5% slope towards this outer slope to promote any runoff. Simulated rainfall was applied via a series of fine sprays located above the tank to provide a uniform intensity over the area of the model. The (monitored) pressure from the sprays was maintained constant to apply a uniform rate of simulated rainfall of 24 mm/hour. Runoff and base seepage from the model were separately directed to tipping buckets to measure their rates. Figure 11 shows the simulated composite product coal stockpile in the laboratory tank at the end of the first test, when the loose-placed material had undergone some obvious slumping as a result of wetting up. Two simulations were performed: (i) rainfall at 24 mm/hour for 5 hours (equivalent to 120 mm of rainfall), immediately followed by sampling for total moisture content and, (ii) rainfall at 24 mm/hour for 4 hours (equivalent to 96 mm of rainfall), followed by drain down, then sampling. 4.2 Monitored flows The two simulated rainfall tests gave similar runoff and base seepage responses during the application of the simulated rainfall, since both were run for long enough to establish steady-state base seepage. Focusing on the first simulation, the monitored flow rates, and cumulative input, runoff, base seepage and net storage volumes, are shown in Figures 12 and 13, respectively. It is apparent from Figures 12 and 13 that surface runoff is minimal; hence almost all of the simulated rainfall infiltrates the embankment. For the first 50 min of the test, the infiltration goes entirely into storage within the embankment, after which base seepage commences, increasing with time until it matches the infiltration rate after about 100 min. Over the same time interval, there is a diminishing contribution to net storage, with some net drain down after that.
4.1 Description of simulations In the absence of natural rainfall at the field test site, the same composite product coal was subjected to simulated rainfall in a laboratory tank measuring 2.4 m long by 750 mm high by 300 mm wide, with a Perspex front. The composite product coal was
516
0.5 0.45 0.4 FLOW RATE (l/min) 0.35 0.3 0.25 0.2 0.15 0.1 0.05 0 -0.05 0 25 50
Input Base Seepage 30 per. Mov. Avg. (Base Seepage)
Runoff Net Storage 30 per. Mov. Avg. (Net Storage)
Table 1. Average total moisture contents of composite product coal from field stockpile drying and laboratory rainfall simulations. Total moisture content (%) Test Field 8 days of drying Lab. (i) 5 hrs rain, immediately Lab. (ii) 4 hrs rain, drain down Initial 9.5 5.1 5.4 Final (range) 6.8 (0.6 to 12.9) 11.8 (8.5 to 16.2) 9.5 (5.4 to 13.3)
75
100
125
150
175
200
225
250
275
300
TIME (min)
Figure 12. Input, runoff, base seepage and net storage during first rainfall simulation applied to laboratory product coal stockpile.
100 CUMULATIVE VOLUMES (litres) 90 80 70 60 50 40 30 20 10 0 0 25 50 75 100 125 150 175 200 225 250 275 300 TIME (min)
Input Runoff Base Seepage Net Storage
Table 2. Average degrees of saturation of composite product coal from field stockpile drying and laboratory rainfall simulations. Degree of saturation (%) Test Field 8 days of drying Lab. (i) 5 hrs rain, immediately Lab. (ii) 4 hrs rain, drain down Initial 35.0 10.6 11.2 Final (range) 15.1 (1.3 to 30.7) 27.8 (19.3 to 40.0) 21.8 (11.8 to 31.7)
Figure 13. Cumulative input, runoff, base seepage and net storage volumes during first rainfall simulation applied to laboratory product coal stockpile.
SUMMARY OF FIELD AND LABORATORY FINDINGS
The total moisture contents of the composite product coal before and after field atmospheric drying and laboratory wetting are given in Table 1, which highlights the dramatic effects that climatic conditions and drain down can have on the moisture state of stockpiled composite product coal. Table 2 gives the corresponding degrees of saturation before and after field atmospheric drying and laboratory wetting. The initial dry density, void ratio, and porosity of the material in the laboratory model tests were 0.818 t/m3, 0.711 and 0.416, respectively. Wetting up due to the simulated rainfall caused the material to slump an average 5% (see Figure 11), increasing the dry density to 0.836 t/m3, and reducing the void ratio and porosity to 0.675 and 0.395, respectively. These latter values were assumed for the field stockpile, to account for densification due to its self-weight. Table 1 indicates that wetting up by rainfall would increase the total moisture content of the composite product coal to an average value of about 11.8%, which would drain down under gravity alone to an average value of about 9.5%, and achieve an average value of about 6.8% following 8 days of hot, dry weather. The corresponding average degrees of saturation of the composite product coal due to wetting up by rainfall, drain
down under gravity alone, and atmospheric drying are 27.8%, 21.8%, and 15.1%, respectively. From Figure 3, the corresponding average matric suctions of the composite product coal due to wetting up by rainfall, drain down under gravity alone, and atmospheric drying are 200 kPa, 1,000 kPa, and 6,000 kPa, respectively. From Figures 4 and 5, the corresponding average hydraulic conductivities of the composite product coal due to wetting up by rainfall, drain down under gravity alone, and atmospheric drying are 1 1011 m/s, 1 1012 m/s, and 6 1014 m/s, respectively. 6 6.1 IMPLICATIONS FOR PRODUCT COAL STOCKPILING AND MANAGEMENT Comparison with reported data
Following 8 days of hot, dry weather, the 11.5 m high composite product coal field stockpile dried to an average total moisture content of 6.8%. This represents a much greater degree of drying than has been reported in the literature. JKMRC (1997) reported an average total moisture content of 10.75% following 5 days of atmospheric drying of a 16 m high stockpile of German Creek composite product coal (containing coarse, fine and ultra-fine size fractions, unlike the composite product coal
517
described herein, which contained only coarse and fine size fractions). A similar result was reported by Williams (2006) following 10 days of atmospheric drying of a 2 m high stockpile of fine-grained Dartbrook composite product coal, which dried to an average total moisture content of 10.6%. Wetting up the composite product coal by simulated rainfall in the laboratory resulted in an average total moisture content of 11.8%. On gravity drainage, this reduced to 9.5%. JKMRC (1997) reported a much higher average total moisture content of 13.1% following 10 mm of rainfall on a 16 m high stockpile of German Creek composite product coal. Williams (2006) reported an average total moisture content of 14.6% following 52 mm of rainfall over 2 days. 6.2 Stockpile dewatering
Product coal stockpiles reclaimed full depth in a series of cuts will produce wetter product coal than stockpiles reclaimed in a series of shallow lifts. 7 CONCLUSIONS
In the interests of ensuring dewatering, product coal stockpiles should be orientated to take advantage of the prevailing wind direction. However, this may also exacerbate dust generation off the stockpile. Drainage installed at the sloping base of stockpiles has been found to be effective in removing water from gravity drainage of the stockpile (Osborne, 2004). However, maintaining the integrity of the drains can be problematic on reclaiming the stockpile. While solar drying of stockpiled product coal is more effective than wind drying, it is limited in depth to about 0.3 m from the surface (Williams, 2006). Product coal stockpiles are reclaimed either full depth in a series of cuts or in a series of shallow lifts. The latter takes advantage of the progressive atmospheric drying of the product coal as it is progressively exposed. 6.3 Stockpile wetting
A combination of physical and unsaturated soil mechanics characterisation testing, atmospheric drying of a field stockpile, and laboratory rainfall simulation of a composite product coal have led to an improved understanding of the ex-plant drying and wetting processes involved. However, while the atmospheric drying represented extreme conditions, the laboratory simulated wetting represented only moderate conditions compared with high rainfall. The results reported in the paper provide implications for product coal stockpiling and stockpile management. ACKNOWLEDGEMENTS The paper was drawn from the results of research carried out as part of ACARP Project C15061. The author gratefully acknowledges ACARP for funding the research, and the company personnel who provided product coal samples and access to their sites. The author is also grateful to Dr Daniel Stolberg and David Truce then of The University of Queensland, who carried out the laboratory and field testing associated with the research. REFERENCES
AS1289. 2000. Methods of Testing Soils for Engineering Purposes. Standards Australia. www.SAIGlobal.com Fredlund, M.D., Fredlund, D.G. and Wilson, G.W. 1997. Prediction of the soil water characteristic curve from grain size distribution and volume mass properties. Proceedings of 3rd Brazilian Symposium on Unsaturated Soils, Rio de Janeiro, Brazil, 2225 April 1997, 12 pp. Fredlund, D.G., Xing, A. and Huang, S. 1994. Predicting the permeability function for unsaturated soils using the soil water characteristic curve. Canadian Geotechnical Journal, 31, 533546. JKMRC. 1997. Investigation of water drainage in a product coal stockpile. JKMRC/AMIRA Project P239C, October 1997. Osborne, D. 2004. Private communication. SoilVision Software. www.SoilVision.com Williams, D.J. 2006. Application of unsaturated soil mechanics to product coal dewatering. Proceedings of 2006 Australian Mining Technology Conference, Hunter Valley, NSW, Australia, 2627 September 2006, 223235. Aus IMM. Williams, D.J. and Whitton, S. 2008. Improving the control of product coal moisture based on unsaturated soil mechanics principles. ACARP Project C15061, September 2008.
Product coal stockpiles located in the shadow of natural hills will not take advantage of enhanced drying by wind action, although dust generation will be reduced. Stockpiles constructed on poorlydrained sites will be ineffective in removing water from gravity drainage of the stockpile. Wet conditions promote the wetting up of product coal stockpiles, particularly at Australian ports due to prevailing wetter conditions there compared with the inland mines (typically increasing the average total moisture content of the composite product coal stockpile at the port by about 0.5% above that at the mine). Stockpile watering to inhibit dust will increase the average total moisture content of the product coal by up to 0.5%. Persistent wet conditions could increase the average total moisture content of a typical composite product coal stockpile by up to 2%.
518
A study of the water retention characteristics and volume change of lime-treated London Clay
X. Zhang, M. Mavroulidou & M.J. Gunn
Faculty of Engineering Science and the Built Environment, London South Bank University, London, UK
Z. Cabarkapa
Geotechnical Consulting Group Ltd, London, UK
ABSTRACT: This paper investigates the effect of lime treatment on the water retention properties of London Clay. A series of contact filter paper tests were performed to obtain the drying Soil Water Retention Curve (SWRC) of statically compacted lime-treated and untreated London Clay specimens respectively. The change in volume with suction during filter paper testing was also determined. In addition to this, further simple tests were performed to assess the shrinkage of samples upon drying. The effect of lime treatment on water retention ability and volumetric change of the tested lime-treated and untreated London Clay soil is shown and discussed. The results showed that lime has a clear favourable effect on the volumetric stability of the soil whereas the effects on water retention properties appear to be less marked. 1 INTRODUCTION field (plasticity characteristics, CBR, unconfined compression tests). With the scope of an increased use of the technique in a wider range of applications, the need has emerged for further data, enabling to understand the properties and behaviour of this type of soil. In particular, testing the soil under unsaturated conditions is necessary due to the paucity of data for this type of soil. Unsaturated conditions are of particular relevance to lime treated soils which are typically compacted (and hence unsaturated) when shallow-mixed in situ. This paper investigates the effect of lime treatment on the water retention properties of London Clay based on Soil Water Retention Curve (SWRC) data obtained from contact filter paper tests. SWRC expressing the variation of water content or alternatively, the degree of saturation with suction, plays an essential role in characterising the hydro-mechanical behaviour of unsaturated soils. Methods have been proposed to predict hydraulic conductivity, volume change, shear strength, diffusion, adsorption, vapour diffusion, thermal conductivity, and a variety of other properties of unsaturated soils, based in part on the information provided by their respective soil-water retention curves (Barbour, 1998). The SWRC is therefore fundamental in predicting and interpreting the behaviour and response of unsaturated soils.
Soil stabilisation with lime has been increasingly used to improve the engineering properties of clayey soils. This can be achieved by either in situ or plant mixing (shallow or deep mixing), or pressure injection of lime slurry into the soil. When lime is added to clayey soils, two different phenomena occur that result in structural change of the soil: cation exchange and pozzolanic reactions. These two mechanisms are referred to as modification and stabilisation respectively, and develop over different timescales. The former mechanism, modification, occurs promptly upon the application of lime. It involves an immediate initial cation exchange between calcium ions and clay minerals, causing the aggregation of soil particles. This is followed by a subsequent development of pozzolanic reactions (provided that lime quantity is sufficient for these to develop). These form stable cementing compounds that gradually improve the strength and compressibility of the soil. This mechanism is referred to as soil stabilisation (Rogers & Glendinning, 1996). Lime treatment has been used in the first place for road and pavement stabilisation (capping layers, sub-base stabilisation). Hence most data in the international literature refer to basic tests commonly used by practitioners in this
519
In addition to SWRC tests a number of simple shrinkage tests are also shown in this paper, complementing observations regarding the effect of lime on the volume change of the soil upon moisture changes. This is of particular relevance, as one of the reasons for using lime treatment is its potential to enhance the volumetric stability of shrinking soils (such as London Clay) while moisture content changes. 2 2.1 MATERIAL, SPECIMEN PREPARATION AND EXPERIMENTAL PROCEDURE London Clay soil properties
reactions. 4% of hydrated lime per dry mix mass was used in the mixes. The necessary percentage of lime for stabilisation was assessed using two different methods, i.e. from plasticity tests and Eades & Grims (1966) method. The latter method defines the optimum lime percentage (in order to trigger pozzolanic reactions between the lime and the soil) as that necessary to achieve a minimum pH value of 12.4. 2.2 Filter paper testing
London Clay from Westminster Bridge in London obtained from a depth of approximately 26 m below ground was used in this study. London Clay is a stiff, overconsolidated clay, formed in the South of England by marine deposition. It is particularly well developed in the London Basin, reaching an average thickness of 130 m. Although at varying percentages according to location, a typical composition of London Clay with respect to its main three constituent minerals would be illite (70%), kaolinite (20%) and montmorillonite (10%) although exact composition changes according to location (Boswell, 1951). Due to the presence of illite and also, in particular, montmorillonite, London Clay is a shrinking clay with annual surface movements due to moisture variations typically of 50 mm or more (Whitlow, 2001). The basic properties of the London Clay soil used in this study are shown in Table 1. The clay was air-dried, pulverised using a rubber pestle and passed through a BS 425 m sieve. For the preparation of untreated London Clay specimens the clay powder was thoroughly mixed with water to achieve a water content 23.4% (which is dry of the Proctor optimum) and left to mellow in sealed bags for 72 h. For the preparation of lime treated specimens, London clay and hydrated lime dry powders were thoroughly mixed and then 25% of water was added (again dry of optimum of the lime treated soil this time). The reason why additional water was used for lime-treated specimens was to ensure that more water was available for pozzolanic
Table 1. Soil type Basic properties of tested London Clay soil. Natural wL w% % wP % wopt % dmax Mg/ m3
IP
Gs
London 31 Clay
63.5 25.8 37.7 2.75 25.5 1.43
A number of statically compacted specimens were prepared at a target dry density of approximately 1.43 Mg/m3 (corresponding to the standard Proctor maximum density of the untreated London Clay). For this, a standard loading frame at a loading rate of 1 mm/min was used. Five specimens of London Clay, and four specimens of lime-treated London Clay were prepared. One specimen of London Clay was tested for the drying SWRC immediately after compaction, whereas the other four were tested after saturation (following a 14-day period of immersion in distilled water). All lime-treated specimens were tested after saturation and parallel water-curing for 28 days. During saturation all specimens were kept within standard oedometer rings (75 mm diameter and 20 mm height approximately) and sandwiched between filter papers and porous stones in a confining frame especially designed to suppress swelling, except two lime-treated specimens which were left to swell freely. The latter two specimens were extruded from the oedometer rings and placed on porous stones standing in water, so that water was slowly absorbed from the bottom of the specimen. Whereas multiple specimens were tested for each type of test to assess repeatability (which was found to be excellent) in this paper results of one specimen from each type is shown, to simplify the graphs for the reader. The specimens presented here are referred to as LCF01 (untreated London Clay, tested after saturation under confined conditions), LCF02 (untreated London Clay, tested immediately after compaction), LLF01 (limetreated London Clay, allowed to swell freely during water curing/saturation) and LLF02 (lime-treated London Clay, tested after water curing/saturation under confined conditions). The reason for testing specimens under two different levels of confinement was to investigate the effect of the stress level in situ corresponding to situations were soil is at shallow depths and greater depths respectively for free swelling and confined specimens respectively. The properties of the compacted specimens before and after saturation are shown in Table 2. Note that at the end of compaction LCF01 specimen had to be transferred to a new ring for saturation. The latter ring was slightly larger in diameter
520
Table 2. testing. Sample LCF01 LCF01 LCF02 LLF01 LLF01 LLF02 LLF02
Properties of filter paper specimens prior to d Mg/m3 1.424 1.326 1.440 1.437 1.157 1.432 1.432 w % 23.42 34.88 23.48 25.37 50.31 25.37 32.68 Sr % 69.14 89.38 71.02 75.36 100 74.88 96.86
Table 3. Properties of compacted specimens for shrinkage tests. Sample LL500 LL1000 LL1500 LC01 LC02 d Mg/m3 1.10 1.27 1.35 1.43 1.65 w % 24.6 24.6 24.6 23.4 23.4 e 1.481 1.156 1.028 0.918 0.669
Stage End of compaction End of saturation End of compaction End of compaction End of saturation End of compaction End of saturation
e 0.932 1.073 0.909 0.906 1.367 0.911 0.911
than the compaction ring and this allowed for some swelling in the radial direction, hence the change in void ratio between compaction conditions and start of test conditions, despite the confinement. The filter paper testing used in the present research to measure matric suction (contact filter paper technique) was performed according to ASTM D 5298-94 (ASTM, 2000). Whatman No. 42 filter paper was used with a calibration formula according to Chandler & Gutierrez (1986). The paper was oven dried at 105C for two hours. It was then put it into a sealed plastic bag to cool at room temperature. The soil specimen was placed between two Perspex disks. Three filter papers were used on each side of the specimen, placed between the soil and Perspex disk interfaces (i.e. six pieces of filter paper in total). The soil specimens were wrapped in three layers of cling film and left in an insulated box for one week at a time. After this period the filter papers were carefully removed with tweezers and placed in an aluminium tin so that they do not lose moisture while their water content was determined. The water content measurement of the filter paper was performed using a Mettler AJ 150 Balance with a 0.0001 g resolution. To determine the dry filter paper mass the paper was again oven-dried at 105 for two hours. The soil specimens were left for air-drying until the new target water content was achieved. They were then wrapped again in cling film (sandwiched within filter papers and Perspex disks) for the new moisture content measurement to be performed one week later. 2.3 Shrinkage testing sample preparation
tests were performed to assess shrinkage upon drying. Three lime treated specimens prepared with the same water content as for SWRC tests were compacted at 500, 1000 and 1500 kPa vertical stress respectively. These are referred to as LL500, LL1000, LL1500 respectively. They were left to slowly air-dry immediately after compaction without curing. Additionally two untreated London Clay specimens (LC01 and LC02) were prepared at the same water content as London Clay specimens used for filter paper tests (i.e. 23.4%) and compacted at two different void ratios (0.918 and 0.669 respectively). The basic properties of the five specimens after compaction are shown in Table 3 below. LC01 was soaked in water under the same confining conditions as specimen LLF02 for comparison purposes. LC02 was left to dry immediately after compaction without carrying out saturation. The volume changes and respective water contents of the above specimens were measured for one week to determine the relationship between void ratio and water content. 3 RESULTS AND DISCUSSION
3.1 Filter paper tests The SWRC of the four specimens are presented in Figures 1 and 2 in terms of degree of saturation and volumetric water content respectively. Each curve point represents the average matric suction value from the two measurement filter papers placed respectively at the top and bottom sides of the specimen. Figure 2 also shows fitted curves to experimental data using MATLAB, according to Fredlund and Xings (1994) equation: s ln(1 / r ) 1 6 ln(1 10 / r ) [ln(e /a )n ]m (1)
In order to obtain a better understanding on the effect of lime on volume change characteristics of London Clay, some additional simple
where volumetric water content; s saturated volumetric water content; suction; r residual suction & a, m, n curve-fitting parameters.
521
Table 4. Curve fitting parameters of the SWRC curves. Sample LCF01 LLF01 LLF02 r % 44.5 55 45 r [kPa] 150000 100000 200000 a 8846 603.3 8100 m 2.615 1.345 2.577 n 0.7433 0.6658 0.7308
Figure 1. Soil water retention curves in terms of degree of saturation.
Figure 3.
Void ratio of the four filter paper samples.
Figure 2. Soil water retention curves in terms of volumetric water content.
The curve-fitting parameters of the SWRC of the four specimens (in terms of volumetric water content versus matric suction) are listed in Table 4. Figure 3 represents the results of the filter paper tests in terms of void ratio versus matric suction. Figure 4 compares the volumetric strains versus matric suction of untreated London Clay sample LCF01 to those of lime-treated sample LLF01 which was compacted and cured at similar conditions as LCF01. From the SWRC and corresponding void ratio and volumetric strain versus suction results, the following observations can be made: LLF02 and LCF01 i.e. the two specimens that were saturated and cured (in the case of lime treated specimen) under confined conditions, show practically the same water retention curves in terms of volumetric water contents. They start with similar volumetric water contents, however the lime treated specimen was able to achieve higher degrees of saturation despite the fact that the porosity of the lime-treated London Clay
Figure 4. Volumetric strain: comparison between untreated (LCF01) and treated (LLF02) London Clay samples.
specimen at low suctions was lower than that of the corresponding London Clay specimen (see Figure 3). This is difficult to explain but it may be linked to pressures generated during confinement suppressing the natural swelling tendency of London Clay (which is higher than that of a lime-treated soil). After suctions higher than 1000 kPa however, LLF02 shows lower degrees of saturation than untreated London Clay (LCF01) despite the similar volumetric water contents and noting that nSr (where n porosity and Sr degree of saturation) this is consistent
522
with the results of Figure 3 showing that void ratio of untreated London Clay change faster with suction so that after 1000 kPa of suction, untreated London Clay has lower void ratios than treated London Clay. However the rates of drying, represented by the slopes of the curves and depending on the uniformity of the specimens are very similar. Note that as in this study compacted granulated clay specimens were used (and in fact compacted dry of optimum), these are likely to have developed a larger pore size network with respect to natural clay soil (Fityus & Buzzi, 2009) and this could be one reason why the differences shown with respect to the limetreated specimen are not that much pronounced. LLF01starts with 100% degree saturation and higher volumetric water contents than all other soil specimens, however its water retention capacity decreases rapidly: desaturation starts at lower suctions than in all other specimens with air-entry values identifiable below 100 kPa (see Figure 1). This shows that for shallow mixing conditions (corresponding to null or low vertical stress levels) in situ lime-treated clay is likely to be often in an unsaturated state, depending on the environmental conditions. This is unlike natural London Clay which is known to be saturated throughout most suction ranges developed in practical engineering situations (in this study there was no data to support this, however this has been often demonstrated in published literature on London Clay, e.g. Croney & Coleman 1954, amongst many other). LLF01, having been allowed to swell during water-curing has the highest void ratios than all other specimens throughout the drying process clearly showing the more open structure of the soil which was free to swell. The lime-treated specimen LLF02 shows clearly lower volumetric strains upon drying throughout all range of suctions, when compared to LCF01. Hence lime-treatment resulted in a clearly higher volumetric stability of the soil when this is subjected to moisture content changes. LCF02, having started drying after compaction without prior saturation, initially has lower volumetric water contents for the same level of suctions with respect to both untreatedsaturated and lime-treated clay specimens. However at suctions above approximately 5,000 kPa volumetric water contents and drying rates (shown from the respective slopes of the SWRC) coincide with those of LCF01 (London claysaturated before compaction) and LLF02 (lime-treated saturated under confined conditions) curves. 3.2 Shrinkage tests
Figure 5. results.
Shrinkage test: comparative void ratio
Figure 6. Shrinkage test: comparative volumetric strains.
the specimens tested for shrinkage as well as those of the two lime treaded specimens used for filter paper testing. The results clearly show that the volumetric strains of all lime treated London Clay specimens are smaller than those of the untreated London Clay specimens. This is consistent with the results based on filter paper tests (see Figure 4) and thus further supports the finding that lime treatment has a favourable effect on the volumetric stability of the shrinking natural clay soil. 4 CONCLUSIONS
Figures 5 and 6 show respectively the void ratios and volumetric strains developing with suction for
A number of filter paper tests were carried out in this study to determine the SWRC of statically compacted London Clay and lime-treated London Clay specimens. A significant finding is that the addition of the lime has a clear favourable effect on the volumetric stability of the soil upon drying as shrinkage is dramatically reduced. The effect of the lime treatment on the soil water retention characteristics is less pronounced. However the
523
tests showed (based on the relationship between suction and degree of saturation) that depending on the field stress state lime treated soil (at shallow depths) can often be in an unsaturated state in situ depending on the environmental conditions, unlike the natural London Clay. Therefore unsaturated soil mechanics concepts and techniques should be applied for the thorough description of the behaviour of this type of soil under field conditions. ACKNOWLEDGEMENTS This study constituted part of research on the hydro-mechanical properties of lime-treated UK clays, funded by the UK Engineering and Physical Sciences Research Council (EPSRC) through grant EP/E037305/1. REFERENCES
American Society for Testing and Materials, 2000 ASTM D 5298-94. Standard test method for measurement of soil potential (suction) using filter paper. Vol. 4.08.
D-18 Committee on Soils and Rocks. West Conshohocken, PA 2000: 10821087. Barbour, S.L. 1998. Nineteenth Canadian Geotechnical Colloquium: The soil-water characteristic curve: a historical perspective. Canadian Geotechnical Journal 35: 873894. Boswell, P.G.H. 1951. The Contribution of Clay Mineralogy to the Study of the Diagenesis of Sediments. Clay Minerals 1(8): 246251. Chandler, R.J. & Gutierrez, C.I. 1986. The filter paper method for suction measurement, Gotechnique 36(2): 265168. Croney, D. & Coleman, J.D. 1954. Soil structure in relation to soil suction (pF). Journal of Soil Science 5(1): 7584. Eades, J.L. & Grim, R.E. 1966. A quick test to determine lime requirements for soil stabilization. Highway Research Record No. 139. Transportation Research Board, Washington, D.C. Fityus, S. & Buzzi, O. 2009. The place of expansive clays in the framework of unsaturated soil mechanics. Applied Clay Science 43(2): 150155. Rogers, C.D.F. & Glendinning, S. 1996. Modification of Clay Soils Using Lime, In Rogers, Glendinning & Nixon (eds), Lime Stabilisation, London: Thomas Telford. Whitlow, R. 2001. Basic Soil Mechanics, 4th Ed., Harlow: Prentice.
524
Comparison of four methods for determining the soil water retention curve
M. Zielinski, P. Sentenac & A. Atique
University of Strathclyde, Glasgow, UK
M. Snchez E. Romero
Texas A & M University, College Station, USA Universitat Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: In the recent years, the relationship between the soil suction and moisture content (water retention curve) became a commonly used tool in assessing the state of desiccation in unsaturated soils. This paper, presents the comparison of four methods for determining the soil water characteristic curve for compacted boulder clay. The laboratory measurements of soil suction and moisture content in compacted clay samples have been performed using: (i) in-direct, (ii) deterministic, and (iii) direct methods. The indirect measurements of soil suction consisted of two independent methods: the direct contact filter paper method (DCFP) and chilled mirror dewpoint potentiameter (WP4). In addition, Mercury Intrusion Porosimetry (MIP) was used determine the water retention curve for the soil used. These three methods were conducted on the relatively small samples and obtained results may not represent the actual water retention curve that can be different from that in the field conditions. Thus, the laboratory results were compared with direct measurements of the soil suction and moisture content taken on the macro-scale embankment model desiccated under fully controlled environment. Two main variables that control desiccation of the soil (soil suction and moisture content) were measured at a depth of 20 cm from the crest of the embankment model. Both, self-refilling tensiometer (SWTS1) and time-domain reflectometry (TDR) have been used to monitor the soil suction and moisture content within embankment body. A good agreement has been found between four methods presented in this study. 1 1.1 INTRODUCTION Summary During the dry spells the shrinkage of soils is commonly associated with the formation of cracks. In order to understand the development of the desiccation cracking with associated failures a basic background of unsaturated soil mechanics must be possessed. Unsaturated soils are mainly controlled by desiccation process. In particular, this process is governed by soil suction, which is dependent on the moisture within the soil through the soil water characteristic curve. Hence, the accurate prediction of the desiccation state of soil was needed. Thus, in recent years, it has become increasingly important for methods capable of measuring or controlling suction efficiently to be developed. Presented study compares four methods for determining soil water retention curve. 1.2 Measurement of soil suction Pore water pressure plays a major role in saturated and unsaturated soil mechanics. Pore water pressure can be divided into: (i) positive pore water
The depth of the unsaturated zone is strongly influenced by environmental factors. Approximately two-thirds of the world has sufficiently hot or arid climatic conditions so that the unsaturated zone forms more than a very shallow surface layer. Moreover, even in regions where unsaturated conditions are limited to a shallow surface layer, many practical problems in geotechnical engineering (i.e. slope stability or pollutant infiltration and migration) are significantly affected by the hydro-mechanical processes occurring in the unsaturated zone. Soil is used extensively as a fill material (i.e. earth-works such as road, railway and flood embankments, earth dams or behind retaining structures) and the unsaturated condition of the fill will crucially affect its behaviour. Seasonal variations in climatic conditions affect the water retention in soils; as a result they swell and shrink.
525
pressure and (ii) negative pore water pressure (suction). Both of these pressures are related to different mechanical soil behaviours and can be responsible for a different soil states and failures. Failures related to positive pore water pressure are more likely to occur in wet climates as shallow slips (Aubeny and Lytton, 2004), where the increasing pore water pressure reduces the strength of the soil. Failures related to negative pore water pressure are more related to the extremely arid, arid and semi-arid climates where the annual evaporation exceeds annual precipitation. Evapotranspiration also plays some role as it causes consolidation and desaturation of the soil mass. Most of the failures that are occurring in unsaturated soils are closely linked with desiccation cracking (Stark and Eid, 1997). When expansive unsaturated soils are subjected to wetting, (ie. a reduction in suction) significant volume changes can occur (swelling). Some of the soils present extreme swelling or expansion when wetted, and some show the significant loss in shear strength upon wetting. Changes in negative pore water pressure, due to heavy rainfalls are the main cause for many of slope failures. This indicates the importance that the negative pore water pressure has in controlling mechanical behaviour in unsaturated soils. There is huge evidence in engineering records about the methods of controlling negative pore water pressure (suction). It can be found that many of them have been developed in the last 50 years when knowledge about unsaturated soils has significantly expanded. In engineering practice, there are two ways to measure soil suction: (i) direct measurements, (ii) indirect measurements. Recently, a deterministic method was also considered as an additional method for obtaining the value of soil suction. There is a wide range of techniques and instruments that can be used for a measurement of soil suction. However, only four of them were in this study. 2 2.1 MATERIAL AND METHODS USED IN PRESENTED STUDY Material
contains 39% of silts and only 9.7% of clay. For each test soil was compacted at 95% of maximum dry density which was found to be d 1.95 Mg/m3. Liquid limit was found to be LL 35.75%, with the Plastic limit PL 16.40% and the Plasticity index ID 19.35%. 2.2 Indirect methodfilter paper
The filter paper method is an inexpensive and relatively simple laboratory technique from which total suction and matric suction can be measured (Bulut et al., 2001). The technique was developed by Gardner (1937), but has not been used by engineers for many years. In recent years the method was revived and found to provide very promising and reliable measurements of soil suction. The method is very low cost making it suitable for routine testing in research laboratories and engineering practice. The filter paper method is commonly used for determining the Soil Water Retention Curve. In this method a circular filter paper (Whatman nr. 42) is placed between two soil samples (see Figure 1). A direct contact is required between the filter paper and soil samples. Additionally, two sacrificial filter papers (Whatman nr. 1) are usually placed below and above the filter. The sample then is wrapped in a cling film and placed into a tightly sealed container for a period of 1 to 2 weeks. This time is required, as the filter paper has to come to the equilibrium with soil. The filter is then carefully removed from the sample and the measurement of the filter paper moisture content is taken. 2.3 Indirect methodWP4 psychrometer Another technique used to measure the soil suction is high range psychrometer (WP4) shown on Figure 2. The working principle of the WP4 is based on the measurement of the relative humidity of the air inside a small sealed chamber where the soil sample is placed (Cardoso et al., 2007).
The comparison of four methods presented in this study was performed using Boulder Clay (Galston Clay) that was previously used for the construction of the flood embankment near Glasgow, UK as a part of flood protection scheme. The same soil was also used for the macro scale laboratory embankment model constructed in the laboratory at Strathclyde University. Boulder Clay contains about 51.3% of coarse soil; with 14.1% of gravels and 37.2 of sand particles. The fine fraction
Figure 1.
Filter paper placed between two soil samples.
526
Figure 2. meter.
WP4Chilled mirror dewpoint potential
WP4 uses the chilled-mirror dewpoint technique to measure the water potential of a sample. This technique is much faster and more accurate than pressure plate and filter paper. As stated above, WP4 measures water potential by determining the relative humidity of the air above the sample in a closed chamber that contains a mirror and a means of detecting condensation on the mirror. At temperature equilibrium, relative humidity is a direct measurement of water potential of the sample. In addition, WP4 features an internal fan that circulates the air within the chamber, thus the equilibrium time is reduced to minimum. WP4 can measure the water potential directly within the range from 0 MPa up to 300 MPa. 2.4 Deterministic methodMIP
Figure 3.
Soil sample intruded by mercury.
of the soil water retention curve. Thus, the entry of the mercury is equivalent to the removal of the water by the non-wetting front of air advance being intruded. The volume of not intruded pores, are used to evaluate the water content or degree of saturation corresponding to matric suction (Romero and Simms, 2008). 2.5 Direct methodSWTS1 tensiometer Alternatively tensiometers were developed to measure directly the negative pore-water pressure that exists in the soil so that the matric suction could be measured directly out in the field, or laboratory. A tensiometer is basically a piezometer that has been re-designed and specifically modified to measure pore water tension (Ridley et al., 2003). The negative pore water pressures are due to the water tension in the air-water menisci in an unsaturated soil. The direct measurement of the matric suction overcomes the problems and assumptions of the other techniques, as the ambient air pressure remains atmospheric. So the pore-air pressure within the soil is atmospheric at the surface being measured. In summary, tensiometer becomes the only one available device at the moment to measure the soil suction directly in the soil. It makes tensiometer a very robust tool and shows its potential in unsaturated soil mechanics. The SWTS1 tensiometer (see Figure 4) used in this study, compromises three different direct measurements. It can measure the following: (i) suction of the unsaturated soils, (ii) positive pore water pressure in saturated soils, (iii) and it can also measure the temperature of the water contained in the soil considered as the temperature of the soil. The range of the sensor varies from 100 kPa to 100 kPa.
The Mercury Intrusion Porosimetry MIP was developed to determine the total pore spaces and pore-size distribution in the porous solids (Diamond, 1970). The principle of Mercury intrusion is to find the pore-size distribution which relates to the volume of intruded pores to the pressure required for intrusion (Romero and Simms, 2008). The sample used for the MIP test requires the removal of water that is contained in the small pores and prevents the entry of mercury. This type of treatment can be done by dehydration process chosen from one of the following techniques: air-drying, oven-drying, freeze-drying or criticalpoint-drying. Freeze-drying process eliminates the surface tension forces caused by air-water interfaces, by manipulating the temperature and pressure conditions. Hence, there is no shrinkage during drying process which can disturb the sample (Romero, 1999). The mercury intrusion process (see Figure 3) is similar to air-intrusion during the desorption path
527
100000 FP 1 FP 2' 10000 FP 2"
Matric suction [kPa]
1000
100
10
Initial d=17.25 kN/m3 Initial d=18.62 kN/m3
Figure 4.
SWTS1 tensiometer.
0.1
Gravimetric moisture content [%]
12
16
20
24
However, in many cases this range is exceeded when the measurements are taken in very dry soils and high suction tensiometers must be considered. The advantage of the SWTS1 sensor is its self-refilling option. When it cavitates, it refills itself once the water gets inside the sensor after rainfall. 3 3.1 SOIL WATER RETENTION CURVE Background
Figure 5.
Soil water retention curve from filter paper.
The Soil Water Retention Curve is the relationship between soil suction and soil water content. This curve describes the thermodynamic potential of the soil water stored in the macro and micro pores with respect to the suction. When the soil contains a relatively small amount of water, the pore water potential is significantly reduced relative to free water, thus producing relatively high soil suction. When the soil contains a relatively big amount of water, the difference between the pore water potential and the potential of free water decreases, thus the soil suction is low. The soil water retention curve can be determined experimentally using one of the techniques described above or derived from in-situ measurements. Alternatively it can be also estimated from the particle size distributions (Fredlund et al., 1997). The SWRC plays an important role in assessing the state of desiccation of unsaturated soils. 4 LABORATORY STUDY AND RESULTS
The filter papers were removed from the samples after one week and the soil suction measurements were taken, whereas the filter papers were replaced in the samples used for the second test. Moreover, before placing the new filter paper two samples from the second test were exposed to atmosphere for about 4 hours in order to initiate drying. Prior to this operation, samples were carefully weighted and measured in order to calculate the moisture content and volume change after drying. The samples were then sealed again and placed inside containers for another week. It can be clearly seen in Figure 5 that the same desorption path is followed by results obtained from both tests. 4.2 Indirect methodWP4-T psychrometer These tests were carried out at UPC in Barcelona. The sample was prepared and compacted in the oedometer ring at desired dry unit weight (d 17.56 kN/m3) and at constant water content of 10.0%. Then the sample was trimmed and sliced into two slices. Both samples were trimmed to the size of sample cup and weighed. The first slice was placed inside the chamber, whereas the second slice was left to air dry in order to find the air dry point. After each measurement, sample was removed and the weighed. The same procedure was applied for an air dried sample. The suction readings were plotted against the moisture content determined from the difference between the initial and final mass of the sample. Despite the fact that the initial water content and void ratio for both samples were different, the same drying path was obtained for both samples, as it is presented on Figure 6. 4.3 Deterministic methodMIP
4.1 Indirect methodfilter paper Two filter paper tests have been carried out in order to study the SWRC of soil used in this research. In the first test, eleven samples were compacted (following compaction curve) at different dry unit weights and at different moisture contents. In the second test two samples were compacted at the desired dry unit weights (d 18.62 kN/m3 and d 17.25 kN/m3) and at constant moisture contents of 14.6% and 18.8% respectively.
MIP test was carried out at UPC in Barcelona using Autopore IV 9500 porosimeter to examine the pore
528
1000000 WP4 1 WP4 2 air dried
10000
1000
100000
Suction [kPa]
10000
Suction [kPa]
d=17.56 kN/m3
100
10
1000
d=16.47 kN/m3
1
100
0.1
10
11
12
Moisture content [%]
10
12
14
16
18
20
22
24
Figure 6.
Soil water retention curve from WP4 test.
Figure 7. Soil water retention curve from MIP test.

1000 1st week drying 2nd week drying 3rd week drying
structure of the soil sample prepared at saturation state. The equipment consists of two operating units: low pressure filling apparatus from approximately 0.67 kPa to 345 kPa measuring macropores with apparent pore diameter from 3.6 m to 360 m where the sample is degassed under vacuum; and the high pressure station for micropores characterisation, where the high pressure (up to 228 MPa) is applied continuously from 0.67 kPa (apparent pore diameter 0.0055 m to 6 m). The MIP equipment requires dehydrated cubical samples measuring preferably less than 1 cm3 (limited by the sample holder). A sample was previously prepared and compacted in the oedometer ring at desired dry unit weight (d 16.47 kN/m3) and at constant water content of 20.03%. From the compacted sample, a small specimen was carefully trimmed using a penknife into a cube measuring around 10 mm on each side. Subsequently this sample was freeze dried to remove the pore water and then kept in a desiccator until testing. The sample was placed in the sample holder, weighed and mercury was allowed to flow into the sample holder. The pressure was raised continuously (up to 345 kPa) to let the mercury intrude the sample. The pressure is then released and the sample is moved to the high pressure station for micropores characterisation. The SWRC is calculated using the apparent pore diameter from the pore size distribution curve. For the saturated moisture content (wsat 22%) the soil water retention curve is presented in Figure 7. 4.4 Direct methodSWTS1 tensiometer
100
Matric suction [kPa]

10 1 16 18 20 22
Gravimetric moisture content [%]
Figure 8. Soil water retention curve from TDR and tensiometer measurements.
This test was carried out on the laboratory embankment model constructed in the concrete flume. The embankment (compacted at 95% maximum dry density-1.85 Mg/m3) was 1 m high, with
the crest of 1 m 1 m and slopes 1:2. Additionally embankment was covered by an environmental chamber equipped with the irrigation and drying systems. One SWTS1 tensiometer and one moisture content probe (TDR) were installed on the crest of the embankment at a depth of 0.2 m and at the same installation angle of 25. In the case of tensiometer installation, the slurry paste was injected into each augered hole in order to provide a good contact between the soil and the ceramic cup. After installation, both sensors were connected to the Delta-T Data Logger that was programmed to take readings every 30 seconds. Following the installation of the sensors, three weeks experiment was carried out with the constant weather conditions applied every week to the embankment model. The inducing desiccation temperature (inside environmental chamber) was set to 20C and one weekly precipitation was intro-
529
1000000
FP1 FP2' FP2" WP4 1 WP4 2 MIP SWTS1 1w SWTS1 2w SWTS1 3w
100000
10000
Suction [kPa]
1000
It is rare that four techniques (filter paper, MIP, WP4 and tensiometer) involving three different measuring and one deterministic methods (indirect, deterministic and direct) will provide such interesting results. Moreover, due to the lack of such observation in the literature, it makes it even more unique and valuable. ACKNOWLEDGMENTS
100
10
0.1
12
16
20
24
Figure 9. tests.
Soil water retention curves obtained from all
The authors gratefully acknowledge the support of the Research Enhancement Group, David Anderson Bequest and Synergy pump prim funding at the University of Strathclyde. The first author also acknowledges the financial support given by AXA Research Fund. REFERENCES
Aubeny, C., Lytton, R., (2004), Shallow slides in compacted high plasticity clay slopes. Journal of geotechnical and geoenvironmental engineering., 130 (7), pp. 717727. Bulut, R., Lytton, R., Wray, W., (2001), Soil suction measurements by filter paper. Proceedings of Geo-Institute Shallow Foundation and Soil Properties Committee Sessions at the ASCE 2001 Civil Engineering Conference., 1, pp. 243261. Cardoso, R., Romero, E., Lima, A., Ferrari, A., (2007), A comparative study of soil suction measurement using two diferent high-range psychrometers. Springer Proceedings in Physics. Experimental Unsaturated Soil Mechanics., 112, pp. 7993. Diamond, S., (1970), Pore size distribution in clays. Clay mineralogy., 18, pp. 723. Fredlund, M.D., Wilson, G.W., Fredlund, D., (1997), Indirect procedures to determine unsaturated soil property functions. Proceedings of the 50th Canadian Geotechnical Conference, Ottawa, Ontario, Canada, October 2022, 1997. Gardner, W.R., (1937), A method of measuring the capillary tension of soil moisture over a wide moisture range. Soil Science Society of America Journal., 43 (4), pp. 277283. Ridley, A., Dineen, K., Burland, J., Vaughan, P., (2003), Soil matrix suction: some examples of its measurement and application in geotechnical engineering. Gotechnique., 53 (2), pp. 241253. Romero, E., (1999), Characterisation and thermo-hydromechanical behavior of unsaturated boom clay: An experimental study. Ph.D. thesis, Universitat Politcnia de Catalunya, Barcelona. Romero, E., Simms, P., (2008), Microstructure investigation in unsaturated soils: A review with special attention to contribution of mercury intrusion porosimetry and environmental scanning electron microscopy. Geotechnical and Geological Engineering., DOI 10.1007/ s1070600892045, pp. 123. Stark, T., Eid, H., (1997), Slope stability analyses in stiff fissured clays. Journal of geotechnical and geoenvironmental engineering., 123 (4), pp. 335343.
duced at the beginning of each week reaching the amount of 55 mm/m2. Three soil water retention curves were obtained from the tensiometer and TDR (one from each week) in order to show the path that SWRC will follow during drying. It can be seen from the measurements presented in Figure 8 that after the rainfall, the same slope of desorption curve was obtained even when the soil volume changed as a result of wetting and drying. 5 COMPARISON OF WATER RETENTION CURVES
Although, four different methods were used to determine SWRC for the soil used, a good match between all of them can be noticed. Moreover, the valuable results were obtained from the measurements that were based on different sizes of samples. The comparison between water retention curves obtained from the laboratory tests and embankment model is presented in Figure 9. 6 CONCLUSIONS
In this paper comparison of four methods for determining the soil water retention curve has been presented. It has been observed that the relationship between soil suction and moisture content (in four methods) matches exactly the same desorption curve prior to soil drying. In presented work, where the size of the samples varied from the 1 cm3 cube to the size of full scale embankment model it has been found that the scale effect plays very little role in determining the soil water retention curve.
530
Thermal and chemical effects
Hydro-mechanical behavior of pre-heated bentonite

Y.F. Arifin T. Schanz
Laboratory of Soil Mechanics, University of Lambung Mangkurat, Banjarbaru, Indonesia Chair and Professor for Foundation Engineering, Soil & Rock Mechanics, Ruhr-Universitt Bochum, Germany
ABSTRACT: This paper focuses on hydro-mechanical behavior (i.e., suction characteristic and swelling pressure) of thermal treated bentonite. Calcium type bentonite (Calcigel) was used in the experimental study. The specimens were heated to temperatures of 100, 200, 300, 400, and 500C. Change in minerallogy of specimen was quantified using X-ray diffraction analysis. Chilled-mirror hygrometer technique was used to determine total suction of the pre-heated specimen at room temperature. Swelling pressure was measured applying constant volume swelling pressure method. The results show that the ability of pre-heated specimen to retain water decreases after heating to temperature higher than 100C due to dehydration and dehydroxylation processes. Pre-heated process does not only influence the value of swelling pressure but also the rate of swelling pressure up built during the test. 1 INTRODUCTION Curve (SWCC) which shows the suction-water content relationship to predict the shear strength of soils. Water permeability of soils in unsaturated condition can also be predicted using statistical model derived from SWCC of soil (Leong & Rahardjo, 1997). Besides suction, swelling characteristic has a good relationship to the hydraulic conductivity of compacted bentonite (Komine, 2004). High swelling pressure of compacted backfill is also effective to minimize excavation damaged zone in the nuclear waste repository. The increase in temperature may cause the smectite mineral to become unstable and transform to more stable silicate phases also known as illitation. This also results in decreasing the ability of the bentonite to retain water. The kinetics of the smectite to illite reaction strongly depends on temperature, time, and K pore-water concentration (Wersin et al. 2006). de la Fuente et al. (2000) found that the transformation to illite occurred to the bentonite undergo heating for 180360 days at temperatures of 120160C which is higher than maximum temperature often considered for bentonite buffer (i.e., 80C) in the design. Drief et al. (2002) found that the transformation from smectite to illite also occurs to the smectite heated up to 50C for 30 days in K-enriched sea water solution. But for K-rich bentonite, a smectite-illite transformation is possible without external supply of K (Drief & Nieto, 2000). This paper presents hydro-mechanical behavior of bentonite pre-heated to the temperatures from 100 to 500C. The hydro-mechanical behavior
High temperature is normally used for stabilizing expansive soils (Joshi et al. 1994, Abu-Zreig et al. 2001, Wang et al. 2008). Joshi et al. (1994) found that as the temperature treatment was increased, shear strength increased particularly between 400500C for Kaolinite, between 500600C for KaoliniteBentonite mixtures, and between 600700C for Bentonite. This result is related to dehydroxilation (i.e., the loss of structural-hydroxyl groups from clays in the form of water) at temperatures of 300500C. Abu-Zreig et al. (2001) found that swelling pressure of three types of clays from Jordan with liquid limits of 4669% starts to decrease at temperature higher than 100C. At temperature of 400C, swelling pressure of the clays decreases up to 90% of the swelling pressure of untreated specimen. Wang et al. (2008) found that swelling potential increases for specimens heated to temperature higher than 100C and starts to decrease at temperatures of 400 and 600C for Kaolinite and Bentonite, respectively. The use of bentonite as a sealing material in High Level Waste (HLW) repository leads to challenge that the bentonite will experienced high temperature by time (Wersin et al. 2007). In this case, suction and swelling characteristics of material are expected unchanged due to temperature effect. Suction constitutes a special interest because many of the geotechnical processes are suction dependent (Fredlund & Rahardjo, 1993). Vanapalli et al. (1996), for instance, used the Soil Water Characteristic
533
includes suction characteristic and swelling pressure of bentonite. Changes in mineralogy of the bentonite due to temperature are also considered. 2 MATERIAL USED
The material used in this study was a calcium-type bentonite, Calcigel (Arifin, 2008). The bentonite used has specific gravity of 2.8, the liquid limit of 180%, and the plastic limit of 33%. The monmorillonite content of the bentonite used is 50%60% with a specific surface area of 525 m2/g. Calcigel has cation exchange capacity of 48 meq/100 g with exchangeable cations Na, Ca2, Mg2, and K of 2, 27, 17, and 0 meq/100 g, respectively. Determined using squeezing pressure technique, osmotic suction of Calcigel was 50 kPa (Arifin & Schanz, 2009). The cations in the extracted soil pore-water obtained in the osmotic suction measurement from the highest to the lowest concentration are Ca2, Na, Mg2, and K. 3 3.1 EXPERIMENTAL PROGRAM Suction characteristic determination
reach water content of 9% and kept in a container for 2 weeks for equilibration. The swelling pressure test was carried out using UPC swelling pressure cells by constant volume test method (Villar et al. 2001, Arifin & Schanz, 2007). The UPC cell was equipped with a load cell which was calibrated prior to use. The statically compacted specimen with a diameter of 50 mm, a height of 20 mm, and dry density of 1.37 Mg/m3 was placed in the cell. Distilled water was circulated through the specimen bondaries to saturate the specimen. The development of swelling pressure was measured from the load cell readings using a handheld volt meter (Arifin & Schanz, 2007). 4 RESULTS AND DISCUSSION
The total suction measurements were performed to the specimen heated at temperatures of 100, 200, 300, 400, and 500C. In this test, the bentonite was placed in oven at 100C for 24 hours. After cooling in desiccator, the specimen was placed in a container and put in an electric furnace that can be used for temperature up to 1000C. In this study, temperatures of 200, 300, 400, and 500C were applied. The specimens were heated for 3 hours at respective temperatures and cooled in a desiccator. Changes in mineralogy after heating were determined using X-ray diffraction analysis. The specimens were placed in the specieal container of chilled-mirror hygrometer (Leong et al. 2003) and soaked in the container with deaired-distilled water for 2 weeks. After soaking, the specimens were dried in a desiccator with salt solution. The relative humidity of the specimens was measured using the chilled-mirror hygrometer. The weights of the specimens were measured using a balance with readability of 0.0001 g immediately after the total suction determination. Afterward, the specimen was placed in a desiccator with salt solution having lower relative humidity or higher total suction. The specimens were placed in the oven with temperature of 100C in order to determine the dry weight of the specimens. 3.2 Swelling pressure measurement
Figure 1 shows the X-ray diffraction result of Calcigel before and after heating at 100 to 500C. As shown in the figure, the peaks of montmorillonite mineral reduce after heating at temperature of 200C. The peak is not shown after heating at temperature of 400C. Figure 2 shows the weight loss of specimen due to heating. As shown in the figure, the weight of specimen decreases by increasing temperature (i.e., at 100200C and at 400C). Dramatic decrease in weight loss at 100200C is due to dehydration (i.e., the removal of inter-layer water). The weight loss at higher temperature (i.e., 400C) can be attributed to dehydroxylation (i.e., the loss of structural-hydroxyl groups from clays in the form of water) (Joshi et al. 1994). This process results in unrecognizing montmorillonite mineral in X-ray diffraction test for specimen heated to temperature 400C. Wang et al. (2008) stated that as temperature increases to 400C, particles start to coagulate together due to electro-chemical bonds. As a consequence, larger particles form and result in a
1600 1400 1200 Montmorillonite
Intensity cps
1000 800 600 400 200 0 0 20 40 60
500C 400C 300C 200C 100C Non treated specimen
80
2theta ()
For swelling pressure tests, the pre-heated specimens were mixed with deaired-distilled water to
Figure 1. X-ray diffraction result of Calcigel before and after heating.
534
Temperature (oC) 0 0.0% 0.5% 1.0% Loss of weight 1.5% 2.0% 2.5% 3.0% 3.5% 4.0% 100 200 300 400 500 600
1000000
Non treated specimen 100C 200C 300C

100000
dehydration
Total suction (kPa)
400C 500C
10000
1000
dehydroxylation
50%
40% 30% 20% 10%
100 0 10 20 30 40 50
Figure 2.
Weight loss versus heating temperature.
Water content (%)
Figure 3. Effect of heat pretreatment on the suction characteristic curves of Calcigel.
drop in clay size content. Wang et al. (2008) also stated that this formation of larger particles due to electro-chemical bonds appears to be irreversible. Considering very small amount of K was found in the extracted soil pore-water and in the CEC determination, transformation from montmorillonite (smectite) to illite does not occur in this type of bentonite. Therefore, decreasing the peaks of montmorillonite is due to removal of water from specimen. Figure 3 shows suction characteristic curves of pre-heated Calcigel. These curves also show the ability of soil to retain water or water retention curve. The experimental data were fitted with solid line. Dash lines show the alteration of the suction characteristic curves of the specimen compared to that of non heated specimen, with reference to the water content. Figure 3 shows that no difference in the total suction versus water content between non-treated specimen and specimen pre-heated at 100C was observed. This result confirms that the changes in the suction characteristic of compacted bentonite due to elevated temperature less than 100C are reversible. The figure also shows that the ability of specimen to retain water reduces approximately 15% for the pre-heated specimen at 200C. The reductions increase up to 25%, 35%, and 40% after the specimens were pre-heated at temperatures of 300, 400, and 500C, respectively. These occur due to coagulation of particles by electro-chemical bonds that results in decreasing the ability of the bentonite to retain water. However, some coagulation are still reversible because the specimens still retain some amount of water. Figure 4 shows swelling pressure development versus time plotted in semi-logarithmic scale for unheated and pre-heated specimens. From the
1200 Non treated sp. 1000 Swelling pressure (kPa) 100 C 200 C 300 C 400 C 500 C 600
800
400
200
0 0.1 1 10 100 1000 10000 100000 Time (minutes)
Figure 4. Calcigel.
Swelling pressure development of pre-heated
figure, it can be found two types of swelling pressure developments. First type is one-step swelling pressure development and second type is two-step swelling pressure development. The one step swelling pressure development occurs to the untreated specimen and pre-heated specimens at 100, 200, and 300C. One-step swelling pressure development means that rapid increase in swelling pressure occurs only once. For instance, rapid swelling pressure development for untreated specimen occurs at 52000 minutes. For two-step swelling pressure development, rapid swelling pressure occurs twice. For instance, swelling pressure development for preheated specimen at 400C occurs at 110 minutes and 50050000 minutes.
535
Agus & Schanz (2005) found that the rate of swelling pressure development is a function of the initial total suction of the compacted specimen. The higher the initial total suction of the specimen, the faster the swelling pressure development (Agus & Schanz, 2005). However, from Figure 3, it can be seen that the initial total suction of specimens preheated at 400 and 500C is less than those of specimen pre-heated at 100, 200, and 300C. Therefore, the faster and the two-steps swelling pressure development of pre-heated specimen at 400 and 500C are due to dehydration and dehydroxylation processes occured to the specimens. First swelling pressure development occurs to fullfil the loss of structural hydroxil groups or dehydroxylation and the second occurs due to osmotic mechanism. Considering high swelling pressure obtained from both specimens, it is possible that not all electrochemical bonds due to dehydroxylation process in the specimen is irreversible. Complete dehydroxylation process may not be achived in this study. For bentonite, Joshi et al. (1994) and Wang et al. (2008) reported that dehydroxylation was defined at a temperature of 500C and completed by 700C. Irreversibility of specimen pre-heated to the temperatures of 400 and 500C results in decreasing swelling pressure obtained. 5 CONCLUSION
REFERENCES
Abu-Zreig, M.M., Al-Akhras, N.M. & Attom, M.F. 2001. Influence of heat treatment on the behavior of clayey soils. Applied Clay Science 20:129135. Agus, S.S. & Schanz, T. 2005. Swelling pressure and total suction of compacted Bentonite-sand mixtures. Bilsel, H and Nalbantoglu, Z (eds) Proceeding of International Conference on Problematic Soils. North Cyprus. Vol. 1: 6170. Arifin, Y.F. 2008. Thermo-hydro-mechanical behavior of compacted bentonite-sand mixtures: An experimental study. PhD dissertation. Bauhaus-Universitaet Weimar. Germany. Arifin, Y.F. & Schanz, T. 2007. Modified isochoric cell for temperature controlled swelling pressure tests. In Tom Schanz (ed.) Mechanics of Unsaturated Soils: Proc. 2nd International Conference, Weimar, March 2007. Berlin: Springer. Vol. 1: 229241. Arifin, Y.F. & Schanz, T. 2009. Osmotic suction of highly plastic clays. Acta geotecnica 4(8): 177191. de la Fuente, S., Cuadros, J., Fiore, S. & Linares, J. 2000. Electron microscopy study of volcanic tuff alteration to illite-smectite under hydrothermal conditions. Clays and Clay Minerals 48(3): 339350. Drief, A., Martinez-Ruiz, F., Nieto, F. & Sanchez, N.V. 2002. Transmission electron microscopy evidence for experimental illitization of smectite in K-enriched seawater solution at 50C and basic pH. Clays and Clay Minerals 50(6): 746756. Drief, A. & Nieto, F. 2000. Chemical composition of smectites formed in clastic sediments. Implications for the smectite-illite transformation. Clay minerals 35(4): 665678. Fredlund, D.G. & Rahardjo H. 1993. Soil mechanics for unsaturated soils, John Willey & Son: Canada. Joshi, R.C., Achari, G., Horsfield, D. & Nagaraj, T.S. 1994. Effect of heat treatment on strength of clays. Journal of Geotechnical Engineering 120(6): 10801088. Komine, H. 2004. Simplified evolution on hydraulic conductivities of sand-bentonite mixture backfill. Applied Clay Science 26:1316. Leong, E.C. & Rahardjo, H. 1997. Permeability functions for unsaturated soils. Journal of Geotechnical and Geoenvironmental Engineeing 123(12): 11181826. Leong, E.C., Tripathy, S. & Rahardjo, H. 2003. Total suction measurement of unsaturated soils with a device using the chilled-mirror dew-point technique. Gotechnique 53(2): 173182. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction. Canadian Geotechnical Journal 33:379392. Villar, M.V., Rivas, P., Campos, R., Lloret, A., Romero, E. & Mariano, A. 2001. First report on thermo-hydromechanical laboratory tests. Report 70-IMA-L-0-86. Wang, M.C., Jao, G. & Ghajal, M.S. 2008. Heating effect on swelling behavior of expansive soils. Geomechanics and Geoengineering 3(2): 121127. Wersin, P., Johnson, L.H. & McKinley I.G. 2006. Performance of the bentonite barrier at temperatures beyond 100C: A critical review. Physics and Chemistry of the Earth 32(8): 780788.
The hydro-mechanical behavior of pre-heated specimen at temperatures 100500C has been presented. The following conclusions can be drawn based on the results of suction characteristic and swelling pressure investigation on pre-heated Calcigel. The peaks of montmorillonite mineral in the bentonite used in this study alter for specimen heated at the temperature higher than 100C and fade for higher than 400C. Dehydroxylation was well defined at temperature of 400C. No evidence of mineralogical change is found to the bentonite used in this study. There is no change in total suction characteristic curves for specimens heated to temperature of 100C. The ability of pre-heated specimen to retain water decreases after heating to temperature higher than 100C due to dehydration and dehydroxylation processes. Pre-heated process does not only influence the value of swelling pressure but also the development of the pressure during the test. Two types of swelling pressure development (i.e., one-step and two-step swelling pressure development) were found. Two-step swelling pressure development occurs to the specimen heated at temperature higher than 400C.
536
Lime stabilisation of pyroclastic soils

M. Cecconi, V. Pane & F. Marmottini
DICA, University of Perugia, Italy
G. Russo, P. Croce & S. dal Vecchio

DiMSAT, University of Cassino, Italy
ABSTRACT: Natural pozzolanas are very reactive with calcium oxide (CaO) provided by lime, the effectiveness of the stabilisation mainly depending on the amount of low alkali alluminosilicate glass phase. Nevertheless there is a lack of systematic studies on the suitability of pyroclastic soils to lime stabilisation, and the effects of the treatment parameters on the engineering properties of the stabilised material. In this paper the preliminary results of experimental research on lime stabilisation of pyroclastic soils are presented. Two pyroclastic soils have been considered in this study. The first is a sand with gravel belonging to a pyroclastic flow deposit of the Colli Albani volcanic complex, and the second is a silty sand originated by the late activity of the Somma-Vesuvius. The results on natural and lime stabilised samples allow a first overview on the suitability and the effectiveness of lime stabilisation of pyroclastic soils. 1 INTRODUCTION the Somma-Vesuvius volcano. These soils belong to deposits including fallout and flow pyroclastic soils/rocks such as pozzolanas and tuffs. Their main features are briefly outlined in the following. The geotechnical investigation on roman pozzolanas started about 15 years ago, driven by the need of analysing the stability of sub-vertical cuts and underground cavities which are rather frequent in many areas of the city. The main outcomes of the study of the mechanical properties of the deposit of Pozzolana Nera (PN) in saturated conditions are described in detail in Cecconi & Viggiani (2001), while the main results of the experimental work on the hydraulic and mechanical properties of the material in unsaturated conditions are discussed in Cattoni et al. (2007). The PN is a coarse-grained pyroclastic weak rock originating from pyroclastic flows, chaotically deposited at high temperature in response to gravity and then cooled. At the scale of the laboratory sample natural fragments of pozzolanas contain crystals, glass shards and pumices, as well as lithic fragments in highly variable proportion. The micro-structure consists of sub-angular grains of very variable size with a rough surface; intrinsic inter-particle bonds, probably due to the original material continuity during deposition, are made of the same constituents of grains and aggregates. Therefore, bond deterioration and grain crushing upon loadingwhich may occur at very low stress levelscan be considered as undistinguishable physical features. The Monteforte (MF) pyroclastic soil is a weathered and humified ashy soil belonging to
The construction of large infrastructures in Central and Southern Italy introduces the relevant problem of the allocation of the excavated material not suitable for earthworks (Croce & Russo, 2002). In this perspective, some improvement techniques, such as lime stabilisation, have revealed to be essential for the reuse of the natural materials. Pyroclastic soils are very frequent in Central and Southern Italy. It is well known that natural pozzolanas are very reactive with calcium oxide (CaO) provided by lime addition. Nevertheless, pyroclastic soils are generally considered as problematic soils due to their nature, heterogeneity, microstructural features, and their complex unsaturated hydro-mechanical behaviour (e.g. structural collapse upon wetting) and generally they are not used for lime stabilization. All these circumstances have motivated the present research on lime stabilisation of pyroclastic soils, considering the lack of systematic studies on this topic in literature. At this stage, the experimental study is aimed at investigating firstly the soil suitability to the treatment and secondly the effects of treatment parameters on the engineering properties of the stabilised materials. The ongoing experimental work will be dedicated to the hydromechanical behaviour of the stabilised soils. Two pyroclastic soils have been considered in this study, as the representative units of two wide volcanic regions, the volcanic complex of Colli Albani involving the south-east area of the city of Roma and the Campanian region with
537
the stratigraphic succession of Somma-Vesuvius eruptions products. The MF soil characterizes the upper layers (stratum n.2) of the succession in the test site of Monte Faggeto, about 40 km northwest of the volcano Somma-Vesuvius, where an ongoing experimental research project on mudflows in pyroclastic soils (Evangelista et al., 2008) is actually carried on. The relevant hydro-mechanical properties of MF soil are discussed in Papa (2007), Papa et al. (2008) and Nicotera et al. (2008). 2 EXPERIMENTAL WORK
When lime is added to soil, cation exchange and the pozzolanic reactions (Bell, 1996) take place modifying relevantly the microstructure and improving the mechanical properties of the treated soil. These two mechanisms, referred to as modification and stabilisation of treated soils respectively (Rogers & Glendinning, 1996), develop simultaneously but with different time scale. The cation exchange between calcium ions and clayey minerals takes place very quickly, inducing the flocculation of the fine-grained fraction. At a macroscopic scale, the grain size distribution of the treated soil shows a reduction of the finer fraction, and the plasticity index is considerably reduced. The development of pozzolanic reactions, with the formation of stable compounds as hydrated calcium silicates and hydrated alumina silicates, is slower than cation exchange on time scale, and it is responsible for the improvement of the mechanical properties of the treated soil in terms of increase of shear strength and reduction of the compressibility (Russo & Croce, 2001, Croce & Russo, 2003). The first set of analyses of the ongoing laboratory research is focused on the suitability of the improvement technique on pyroclastic soils. Chemical and physical analyses (pH measurements, grain size distributions, specific gravities, Atterberg limits) have been carried out on both natural and treated soils. The results of the first stage of analyses allowed the set up of the treatment parameters (lime percent by weight, initial water content, curing time) in order to gain an effective improvement of the soil. The following stage of the experimental work is focused on the hydro-mechanical behaviour of the unsaturated and treated pyroclastic soil by means of oedometer test, triaxial saturated and unsaturated tests. In the following the oedometer tests results will be dealt with some details. Microstructural analyses have been finally considered in this experimental work. Different kinds of analyses (MIP, SEM, DTA, X ray diffraction, BET) have been planned and partially developed at this stage. These results will not be discussed in this paper.
All the samples were prepared selecting the passing percent to the 2 mm sieve of the oven dried soil. Lime treated samples were prepared by hand mixing the soil with a fixed amount of quicklime powder and distilled water, allowing the quicklime to hydrate for 24 hours. pH measurements were carried on stabilised samples prepared with different amounts of CaO varying in the range 0.0715%, repeating the measurement at increasing curing time. Also, grain size distribution, specific gravity, plastic limit, liquid limit of the treated soils were determined at various curing time. The lime treated unsaturated samples for oedometer tests were prepared following two different procedures: 1. the soil was prepared at fixed water content and added with a percent in weight of quicklime (for PN w 20%, lime content 10%). The mixture was then dynamically compacted inside the oedometer mould; 2. the soil was mixed with lime in dry conditions. As soon as the cell was assembled and the load device applied, water was added and the soil submerged into. Very quickly a vertical stress (v 50 kPa) was applied, in order to prevent the sample from any swelling. By adopting these two procedures, the initial voids ratio e0 of PN was found to vary in a rather wide range of values 1.1 1.25, even if much care was given in the sample-preparation technique. In all tests, vertical stress was conventionally applied in successive steps (v/v 1). However, stress increments were applied each hour, in view of the fact that the 90% of settlement systematically took place in less than one minute. 3 RESULTS
Measurements of pH of lime treated soils have been carried out on finer fractions of pyroclastic soils. Figure 1a and 1b shows the experimental values of pH obtained respectively for PN and MF. The addition of lime leads to an increase of pH towards the 12.4 value that is the pH of saturated lime water. The amount of lime required is generally high due to the poor clay fraction of the investigated soils. The pH of soils at different curing times tends to reduce (Figure 2a and 2b) due to the consumption of calcium ions during pozzolanic reactions (Rogers & Glendinning, 2000). The lower is the initial amount of CaO, the faster is the pH reduction. The percentage of 10% in weight of lime for PN and 7% for MF provided the maintenance of a stable pH for long times, allowing the full development of pozzolanic reactions to take place. These values were then selected in order to prepare the lime stabilised specimens.
538
12
12
15
12
11
11
10
pH pH
%CaO
9 Pozzolana Nera (Rome)
0 days 14 days 28 days 42 days
10
15% CaO 11% CaO 7% CaO 3% CaO
9
0 3 6 % CaO 9 12 15
10
20 30 time [days]
40
50
13 12 11 10
pH
13
1% CaO 3% CaO 7% CaO 5% CaO
12
11
pH
9 8 7 6
Monteforte
t=0 gg t=14 gg t=28 gg t=60 gg
10
% CaO
8
0
9
10
20
30
time [days]
40
50
60
% CaO
Figure 1. b) MF.
pH measurements for stabilised soils: a) PN,
Figure 2. pH measurements at increasing curing time: a) PN, b) MF. Table 1. Specific gravity measurements. Sample Natural 10% lime 10% lime 10% lime Natural 7% lime 7% lime 7% lime 7% lime Curing time / 0 days 14 days 28 days / 0 days 7 days 14 days 28 days Gs 2.670 2.682 2.621 2.623 2.701 2.697 2.610 2.680 2.710
Measurements of specific gravity were carried out on increasingly finer, powdered natural material of PN with a Helium picnometer. An average value Gs 2.67 was found; data also revealed that the closed porosity, within the clasts, is very small. MF specific gravity was measured by means of conventional picnometers, filled with water or CCl4, allowing the removal of entrapped air bubbles by applying the vacuum. Measurements were then repeated on the stabilised materials at different curing time. The obtained values of Gs are shown in Table 1. For PN soil the reduction of Gs with the addition of lime occurs in the first couple of weeks. Then, Gs assumes approximately a constant value. A initial reduction of Gs is showed by MF soil in the short term, while an increase to values higher than the initial ones occurs in the long term. The latter is probably due to the cementation of grains as the pozzolanic reactions develop.
Soil PN PN PN PN MF MF MF MF MF
Grain size distributions were determined for both natural and stabilised PN and MF soils, the latter at increasing curing times of 0, 7, 14 days (0, 7, 14, 28, 42 days for MF). Figures 3a and 3b show that the effects the lime addition are instantaneous, leading to the reduction of the fine fraction and the increase of the coarser fraction. The formation of
539
100 90 80 70
P [%]
wL [%]
65
28 days
60 7 days 14 days 55 0 days natural 50
60 50 40 30 20 10 0 0.01
100 90 80 70
natural 0 days 7 days 14 days
0.1
D [mm]
10
Figure 4. MF soil.
Liquid limit for natural and 7% lime stabilised
1.3
P [%]
60 50 40 30 20 10 0 0.001 0.01 0.1 D [mm] 1 natural 0 days 7 days 14 days 42 days 10 100
1.1 1.2
voids ratio, e
28 days
1.0
0.9
Figure 3. Grading curves for natural and stabilised soils: a) PN; b) MF.
0.8
larger aggregates is mainly due to the flocculation of grains induced by ionic exchange on the short term, and this behaviour also persists for longer curing times. The liquid limit wL of stabilised MF samples was measured at increasing curing time. In Figure 4 the results show that an increase of wL takes place with increasing curing time, as a consequence of the modification of the solids surface induced by lime addition. The plastic limit wP was not detectable for both natural and treated samples. Figure 5 and 6 show the results obtained from oedometer tests carried out on PN natural and treated samples prepared following the procedure 1), with curing time of 1 and 3 weeks. Data are plotted in terms of voids ratio, e, and volume strain, v, as a function of vertical stress. The initial response on first loading is characterised by gradual yield as the vertical stress increases. For vertical stresses in excess of about 2 MPa, larger volume strains occur along a normal compression line ( ), whose slope depends on the curing time, as shown in Figure 5 representing the s evaluated for not-treated, 1 week and 3 weeks cured samples. The compression index Cc (obtained
0.7 10 100 1000 10000
v (kPa)
Figure 5. Compressibility curves from oedometer tests on PN: voids ratio e vs vertical stress, v.
from the last two points of the compression curve) decreases with curing time, varying from 0.35 (nottreated PN) to 0.26 (21 days of curing time), thus revealing that the stiffness of the stabilised samples increases with time. On unloading, the behaviour of stabilised samples is stiffer than that of not-treated samples, with an average swelling index of Cs 0.012 (Cs 0.019 for not-treated pozzolana). Finally, the same data are also reported in Figure 6 in terms of volume strains, in order to avoid any effect of the initial voids ratio on the observed response. It is noted that the very large volume strains accumulated on first loading from the natural/not-treated material are noticeably inhibited when lime stabilisation is provided. Again, the curing time plays a significant effect.
540
0.0 2.0 4.0
main topics: the changes induced by lime at the microstructure level, and the hydromechanical behaviour of unsaturated treated soils. ACKNOWLEDGEMENTS The experimental work performed at University of Perugia by Sara Pappalardo and Tania Ricci and at University of Cassino by Anastasia Capotosto is gratefully acknowledged. REFERENCES
10 100 1000 10000
volume strain, ev (%)
6.0 8.0 10.0 12.0 14.0 16.0 18.0
v (kPa)
Figure 6. Compressibility curves from oedometer tests on PN: volume strain v vs vertical stress, v.
CONCLUSIONS
Two pyroclastic soils have been investigated at this stage of the research, the first coming from the Colli Albani volcano activity and the second belonging to the stratigraphic succession of Somma-Vesuvius products. The first experimental results point out the suitability of lime stabilisation for the investigated pyroclastic soils. The ongoing research activity has been focused on the suitability of the soils investigated to lime stabilisation, mainly based on pH tests at different lime content and at increasing curing time. The high percents of lime required (10% for PN and 7% for MF) depend on the relatively large characteristic dimension of grains. These lime percents maintain high the alkalinity of pore water, allowing the full development of pozzolanic reactions for both the investigated soils. The modifications induced by lime are detected at macrostructure level by changes of specific weights, grain size distributions and liquid limits of natural soils. The treatment parameters (percent of quicklime, initial water content, curing time) have been set up for each soil in order to make the lime stabilisation effective. The results obtained from one dimensional compression tests on treated and unsaturated samples showed a relevant increase of stiffness (i.e. decrease of Cc upon loading) already on the short term, with further increase corresponding to higher curing times. Further investigations are needed to assess the effectiveness of lime stabilisation of pyroclastic soils. At present, the research is focused on two
Bell. Cattoni E., Cecconi M., Pane V. 2007. Geotechnical properties of an unsaturated pyroclastic soil from Roma. Bull. Eng. Geol. Environ., 66, 403414. Cecconi M., Viggiani G.M.B. 2001. Structural features and mechanical behaviour of a pyroclastic weak rock. Int. J. Numer. Anal. Meth. Geomech., 25, 15251557. Croce P., Russo G. 2002. Reimpiego dei terreni di scavo mediante stabilizzazione a calce. Proc. XXI CNG, LAquila, 2002, 387394, Patron Editore. Croce P., Russo G., 2003. Experimental investigation on lime stabilised soils. Proc. XIII European Conference of Soil Mechanics and Geotechnical Engineering (ECSMGE), Prague. Evangelista A., Nicotera M.V., Papa R., Urciuoli G. 2008. Field investigations on triggering mechanisms of fast landslides in unsaturated pyroclastic soils. Unsaturated soils: advances in geo-engineering. Toll et al. (eds.). Taylor & Francis Group, London, 909915. Nicotera M.V., Papa R., Urciuoli G., Russo G. 2008. Caratterizzazione in condizioni di parziale saturazione di una serie stratigrafica di terreni suscettibili di colata di fango. Incontro Annuale dei Ricercatori di Geotecnica (IARG), Catania. Papa R. 2007. Indagine sperimentale sulla coltre piroclastica di un versante della Campania. PhD Thesis, University of Napoli Federico II. Napoli, Italy. Papa R., Evangelista A., Nicotera M.V., Urciuoli G. 2008. Mechanical properties of pyroclastic soils affected by landslide phenomena. Unsaturated soils: advances in geo-engineering. Toll et al. (eds.). Taylor & Francis Group, London, pp. 917923. Rogers C.D.F., Glendinning S. (1996). Modification of clay soils using lime. In Lime Stabilisation, C.D.F. Rogers, S. Glendinning and N. Dixon (eds.), pp. 99112, London, Thomas Telford. Rogers C.D.F., Glendinning S. (2000). Lime Requirement for Stabilization. Journal of the Transportation Research Board, 1721/2000, pp. 99112. Russo G., Croce P. 2001. Indagini sperimentali sui terreni stabilizzati a calceIncontro Annuale dei Ricercatori di Geotecnica (IARG), Milano. Tedesco D.V. 2007. Hydro-mechanical behaviour of limestabilised soils. PhD Thesis, University of Cassino. Cassino, Italy.
541
Effect of temperature on collapse potential of kaolin clay

A. Haghighi, G. Medero & P. Woodward L. Laloui
School of the Built Environment, Heriot-Watt University, Edinburgh, UK Swiss Federal Institute of Technology Lausanne, EPFL, Lausanne, Switzerland
ABSTRACT: The understanding of the fundamental behaviour of clays subjected to high temperatures is essential when working with clay barriers in nuclear waste storage, drilling of deep offshore wells or using clay liners on landfills. The effect of temperature on the soils response under different wetting and stress paths has been studied on this research project. A series of single and double temperaturecontrolled oedometer tests were performed on kaolin clay samples at different initial suction values. Initial suction of the studied material, at different temperatures, was obtained using the filter paper methodology. The oedometer tests were undertaken to characterise the thermo-hydro-mechanical behaviour of the kaolin clay and to quantify the collapse potential in non-isothermal conditions. Based on the experimental results, it was possible to observe a significant influence of the temperature magnitude as well as the initial level of suction on the materials behaviour. 1 INTRODUCTION the influence of temperature on the volume change behaviour of compacted clays. A common conclusion from these studies is that an increase in temperature on normally consolidated clays causes a reduction in the void ratio of the soil, while heating may induce expansion for over consolidated clays. The results of these studies also indicate that the thermal volumetric behaviour of unsaturated soils is strongly affected by suction and overconsolidation ratio. The effect of temperature on water retention curves has been studied by several authors (Romero et al., 2001; Villar & Lloret, 2004; Tang & Cui, 2005; Francois & Laloui, 2008; Uchaipichat & Khalili, 2009; Haghighi et al., 2010) who all observed a slight decrease in retention capacity of soil with increasing temperature. It was concluded that at the same suction value, lower degree of saturation at higher temperatures is due to the reduction in the surface tension of water with increasing temperature. Furthermore, the results of experimental studies of thermal effects on hydraulic conductivity of compacted clays showed that permeability increases with an increase in temperature, due to the decrease in water viscosity (Cho et al., 1999; Delage et al., 2000). Several investigations have been carried out over the past few years to examine the parameters affecting the collapse potential of compacted soils (Pereira & Fredlund, 2000; Rao & Revanasiddappa, 2003; Lim & Miller, 2004; Jotisankasa et al., 2007); however, little data is available in the literature on
Over the past few decades, geotechnical researchers have become very interested in the study of thermal effects on the hydro-mechanical properties of soils. One of the main reasons is due to the interest in storage and disposal of high-level nuclear wastes. Heavily compacted clays surrounding waste disposal are subjected to long term elevated temperature, and this can significantly affect the physical and mechanical properties of clay (Romero et al., 2003; Tang et al., 2008; Francois & Laloui, 2008). The need for an improved understanding of the effects of elevated temperature on soil behaviour is important in many other applications such as the burial of high voltage cables, drilling of deep offshore wells, and foundations subjected to temperature changes. Temperature effect is also relevant in the case of compacted clay used as a hydraulic barrier in landfills in order to keep the waste confined and to minimize the escape of contaminants to groundwater and surrounding area (Yesiller et al., 2005; Bouazza et al., 2008). Clay liners on landfills may be subjected to high temperature of up to 70 C, as a result of biological decomposition of waste containment (Rowe, 2005; Bouazza et al., 2006). To achieve such an understanding it is necessary to incorporate the effect of temperature into the current understanding of soil behaviour. Many researchers (e.g. Sultan et al., 2002; Cekerevac & Laloui 2004; Romero et al., 2005) have performed laboratory studies to investigate
543
Dry density, d (KN/m3)
the effect of temperature on the collapse potential of compacted clays. Collapse potential is an important parameter to be considered when examining the volume change characteristics of unsaturated compacted clays. The results of these studies indicate that the collapse potential is directly related to the matric suction of compacted material and its degree of relative compaction. Lawton et al. (1992) previously demonstrated that pre-wetted moisture content governs the collapse potential, and that exist a critical degree of saturation above which collapse potential becomes negligible. Medero et al. (2003) performed a series of oedometer tests consisting of loading an artificial collapsible soil at constant water content up to a specified vertical stress followed by flooding (i.e. reducing suction) of the sample to observe the corresponding volume change that takes place in the sample. The results of their investigation indicate that the amount of collapse increases to a maximum value, beyond which the collapse tends to decrease for increasing values of vertical stress. It was concluded that this was due to the progressive breakage of metastable structure of the soil skeleton under higher loads. This paper presents the results of a temperaturecontrolled oedometer tests performed on kaolin clay samples at different initial suction values. The main purpose of this work is to investigate the effect of initial suction together with temperature on the collapse potential of compacted soil samples. The filter paper method was used to obtain initial suction of the soil samples at different temperatures. 2 2.1 EXPERIMENTAL PROGRAMME Studied material
Table 1. Soil properties used in this study. Property Liquid limit: % Plastic limit: % Specific gravity Silt fraction: % Clay fraction: % Value 55.0 31.4 2.64 58 42
Clay
100
Silt
Percentage finer (%)
80
60
40
20
0 0.0001
0.001
0.01
0.1
Particle diameter (mm)
Figure 1. clay.
15.7 15.6 15.5 15.4 15.3
Particle size distribution of the studied kaolin
Zero-air-voids line
Sr = 90 % 15.2 15.1 19 20 21 22 23 24 25 26 27 Sr = 80 %
Laboratory tests were performed on kaolin clay with properties shown in Table 1. Figure 1 presents the particle size distribution of the studied material, and shows about 98% by weight is smaller than 0.02 mm and 42% is smaller than 0.002 mm. The mineralogical composition of the material obtained from x-ray diffraction analysis was found to be 82% Kaolinite, 15% Quartz, and 3% Illite and/or Mica. Figure 2 shows the compaction curve obtained following the British Standard BS-1377. From the compaction curve, the optimal water content was found to be 24% and the maximum dry density was 15.58 KN/m3. 2.2 Performed tests The effect of temperature on the soil behaviour was studied by running isothermal oedometer tests at two different temperatures. The experimental
Water content, w (%)
Figure 2.
Studied kaolin clay compaction curve.
approach involved performing a series of single and double oedometer tests on soil samples at different initial suction values and same initial void ratio. Table 2 summarizes the characteristics of the tested soil samples under several thermo-hydromechanical paths. The table also presents initial and final properties of the soil samples. All the oedometer soil samples were prepared by mixing distilled water thoroughly into the ovendried soil to get a homogonous material, which was then statically compacted into the oedometer ring to obtain the desired initial void ratio. The oedometer apparatus was modified in order to be used in
544
Table 2. List of tests in temperature-controlled oedometer apparatus. Initial state T (C) 20 e0 2.02 2.02 2.02 2.02 2.03 2.01 2.01 2.02 1.99 1.98 1.99 1.98 1.98 1.98 1.99 1.99 d (KN/m3) 8.58 8.58 8.57 8.56 8.54 8.59 8.60 8.59 8.66 8.68 8.65 8.69 8.68 8.69 8.65 8.66 w (%) 18.7 18.7 18.8 18.9 19.2 18.5 18.4 18.6 7.6 7.4 7.7 7.3 7.4 7.2 7.8 7.6 Sr0 (%) 24.5 24.5 24.6 24.7 25.0 24.3 24.2 24.4 10.1 9.9 10.2 9.7 9.9 9.6 10.3 10.1 s0 (KPa) 426 Final state ef 1.35 1.08 1.07 1.10 1.33 1.05 1.07 1.06 1.37 1.07 1.09 1.08 1.34 1.04 1.04 1.09 w (%) 16.7 40.1 39.3 40.3 14.4 37.8 37.6 38.4 6.3 39.2 40.1 40.3 3.2 39.1 39.2 40.4 Srf (%) 32.5 98.3 96.6 96.4 28.6 95.1 92.4 95.2 12.1 96.7 97.1 98.5 6.3 99.3 99.5 97.8 Test condition constant water content wetted at 5 KPa wetted at 100 KPa wetted at 200 KPa constant water content wetted at 5 KPa wetted at 100 KPa wetted at 200 KPa constant water content wetted at 5 KPa wetted at 100 KPa wetted at 200 KPa constant water content wetted at 5 KPa wetted at 100 KPa wetted at 200 KPa
50
371
20
747
50
603
a temperature chamber. The temperature chamber allows controlling temperature to 0.1 C. Initial matric suction of the soil samples was obtained at 20 C and 50 C, using filter paper methodology. The procedure used in this study for soil suction measurement is similar to that proposed by Bulut et al. (2001) and followed by Haghighi et al. (2010). Haghighi et al. (2010) proposed a unique calibration equation, for Whatman No. 42 filter paper, in which the effect of temperature on filter paper moisture content is taken into account. The proposed calibration equation can be used for both matric and total suction measurement. The measurement of initial matric suction (s0) was made by placing a piece of Whatman No. 42 filter paper against the soil samples in an air-tight glass jar at relatively constant temperature. The soil samples were prepared to the same water contents and dry densities as the oedometer tests. One filter paper was sandwiched between two larger size filter papers and placed in between the two soil samples to ensure hydraulic continuity between the soil pore water and the water absorbed in the filter papers. The two external papers are to prevent contamination of the middle filter paper by soil. The arrangement of the soil samples and the filter papers were placed inside the glass jars. The glass jars were sealed properly and placed inside the constant temperature chamber for a period of 14 days to ensure that equilibrium was reached. After this period, the glass jars were carefully opened and the water contents of filter papers were then measured using a scientific balance weighing accurately to 0.0001 g. The suction of the soil samples was calculated from the water content of filter papers, using
the calibration equation proposed by Haghighi et al. (2010). Collapse potential of compacted kaolin clay has been studied in this experimental programme for stress level between 5 to 400 KPa, using the double and single point oedometer tests. It was observed that the majority of collapse occurred within the first five minutes of loading upon wetting on the single point oedometer tests. Collapse potential was determined based on deformation values after stabilization period. A similar investigation was carried out by other authors (e.g. Delage et al., 2005; Medero et al., 2009) for different collapsible materials. The collapse potential of the soil samples were found according to the following equation based on the proposition of Jennings and Knight (1975) and defined by ASTM D 5333 (ASTM 2003): Collapse potential ei e f 1 e0 (1)
In this equation e0 refers to initial void ratio and ei and ef are the values of void ratio obtained from the oedometer curves at as-compacted and wetted condition respectively, under the same applied vertical stress. Jennings and Knight (1957) proposed the double oedometer test method which consists of testing two nominally identical samples. One sample is initially saturated with water under a small seating load and allowed to collapse. After equilibrium is reached, the soil sample is loaded according to standard incremental loading procedure. The other sample is tested at the as-compacted water content
545
using standard incremental loading procedure. The vertical strain difference between the saturated and as-compacted samples was determined as the collapse potential of the sample. The collapse potential of the soil samples was also investigated from single point oedometer test results. The test was performed according to ASTM D 5333. In this method the soil sample was incrementally loaded at its natural water content until reaching the desired vertical stress; then inundated and deformation readings were recorded. The vertical strain difference due to wetting under applied vertical stress after stabilizing is determined as the collapse potential of the test specimen. 2.3 Results and discussion
Figure 3 and Figure 4 show the results of single and double oedometer tests performed on compacted material prepared to dry density of 8.6 KN/m3 (corresponding to initial void ratio of 2) and as-compacted water contents of 7% and 18%. The figures show the void ratio of the soil specimens versus vertical overburden pressure. Initial void ratios of the samples varied slightly, and the results of individual test have
2.0 1.8 e0 = 2.0 w0 = 7 %
been adjusted slightly to show all tests starting from the average initial void ratio of 2.01. From these figures, it can be noticed that the soil samples are highly collapsible, with significant collapse occurring even on wetting at low stress values. In tests performed at constant water content, the suction would be expected to reduce as the void ratio decreases and degree of saturation increases during each load increment. From analysis of the results, it was observed that heating induces a thermal hardening phenomenon which results in an increase of the stress value at which yield occurs (i.e. preconsolidation pressure of the soil decreases with increasing temperature). Moreover, it was observed that elastic and plastic compressibility parameters can be considered independent of temperature. Similar behaviour was observed by Cekerevac & Laloui (2004), and Tang et al. (2008). The influence of temperature on the collapse potential (CP) of the material is presented in Figure 5 and Figure 6. From the figures it can be noticed that the collapse potential of the material tends to decrease with increasing temperature. The effect of temperature is however less evident as the applied vertical stress increases. Another important
20
s0 = 426 (KPa)
Collapse potential (%)
18 16 14 12 10 8 0 100 200 300

s0 = 371 (KPa)
e0 = 2.0 w0 = 18 %
Double oedometer test at 20C Double oedometer test at 50C
Void ratio
1.6 1.4 1.2 1.0 0.8 1 10 100 1000

constant water content 20C wetted at 5 KPa 20C wetted at 100 KPa 20C wetted at 200 KPa 20C constant water content 50C wetted at 5 KPa 50C wetted at 100 KPa 50C wetted at 200 KPa 50C
Vertical Stress (KPa)
400
500
Vertical stress (KPa)
Figure 3. Single and double oedometer test results for e0 2 and w0 7%, at 20 and 50 C.
2.0 1.8
Figure 5. Collapse potential of samples at 20 and 50 C.
e0 = 2.0 w0 = 18 %
20
s0 = 747 (KPa)
18 16 14 12
s0 = 603 (KPa)
e0 = 2.0 w0 = 7 %
Double oedometer test at 20C Double oedometer test at 50C
Void ratio
1.6 1.4 1.2 1.0 0.8 1

constant water content - 20C wetted at 5 KPa - 20C wetted at 100 KPa - 20C wetted at 200 KPa - 20C constant water content - 50C wetted at 5 KPa - 50C wetted at 100 KPa - 50C wetted at 200 KPa - 50C
10 8 0 100 200 300 400 500
10
100
1000
Vertical Stress (KPa)
Figure 4. Single and double oedometer test results for e0 2 and w0 18%, at 20 and 50 C.
Figure 6. Collapse potential of samples at 20 and 50 C.
546
observation is that collapse potential of the samples increases with increasing applied vertical stress up to a maximum value, beyond which the collapse potential decreases for increasing values of vertical stress. A similar behaviour has been reported by Sun et al. (2004), and Medero et al. (2009). The maximum CP was observed at approximately 40 KPa of vertical stress, for the soil samples with initial water content of 7% and initial suction of 747 KPa at 20 C and initial suction of 603 KPa at 50 C. Moreover, the maximum CP was observed at approximately 25 KPa of vertical stress, for the soil samples with initial water content of 18% and initial suction of 426 KPa at 20 C and initial suction of 371 KPa at 50 C. The initial matric suction of the soil samples was obtained using the procedure described above; and reported in Table 2. All the soil samples were prepared with the same initial void ratio e0 2, and initial water content of 7% and 18%. From the results it can be deduced that soil suction is lower at higher temperature for the soil samples for the same void ratio and water content. Figure 7 shows the obtained CP for the samples with initial suction of 747 KPa and 426 KPa at
20
s0 = 426 (KPa)
12.8
e0 = 2.0 Vertical stress: 200 KPa T = 20 C
12.4
CP (%)
12.0
11.6
T = 50 C
11.2 200 400 600 800
Initial matric suction (KPa)
Figure 9. CP of soil samples at different initial suction values at vertical stress of 200 KPa.
20 C. The results demonstrate that vertical stress at which maximum CP occur increases with increasing initial suction value. Figure 8 and Figure 9 show the obtained CP, at different initial suction values, at vertical stress of 100 and 200 KPa respectively. From the figures it can be observed that CP is higher for the sample with higher initial suction values. The similar behaviour was observed for both single and double oedometer test results. 3 CONCLUSION
e0 = 2.0 T = 20 C
Double oedometer, wi:7% wetted at 100 KPa wetted at 200 KPa Double oedometer, wi:18% wetted at 100 KPa wetted at 200 KPa
18 16 14 12 10 8 0 100 200
s0 = 747 (KPa)
The main conclusions that can be drawn on the basis of the test results are as follows: 1. A series of single and double temperature controlled oedometer tests performed revealed that elastic and plastic compressibility parameters are independent of temperature. For the studied material, it was observed that collapse increases with increasing vertical stress up to a maximum value, followed by a decrease with increasing applied vertical stress. 2. It was observed that collapse potential of the studied kaolin clay tends to decrease with increasing temperature. 3. Collapse potential of the material increases with increased initial suction values. Also it was observed that vertical stress at which maximum CP occur increased with increasing initial suction value. 4. Filter paper method was used for soil suction measurement at different temperatures. It was observed that soil suction is lower at higher temperature for the soil samples with the same void ratio and water content. REFERENCES
American Society for Testing and Materials (ASTM). 2003. Test method for measurement of collapse potential of soils. Annual book of ASTM standards D 5333, Philadelphia.
300
400
Figure 7. Collapse potential of samples at 20 C s0 747 KPa, and s0 426 KPa.

15.6
e0 = 2.0 Vertical stress: 100 KPa
15.0
CP (%)
T = 20 C
14.4
13.8
T = 50 C
13.2 200 400 600 800
Initial matric suction (KPa)
Figure 8. CP of soil samples at different initial suction values at vertical stress of 100 KPa.
547
Bouazza, A., Nahlawi, H., Murdoch, C. & Alyward, M. 2006. Temperature monitoring in a green waste landfill cell, Geoenvironmental Engineering Conference, Japan. Birle, E., Heyer, D. & Vogt, N. 2008. Influence of the initial water content and dry density on the soil-water retention curve and the shrinkage behaviour of a compacted clay. Acta Geotechnica, 3 (3): 191200. Bouazza, A., Abuel-Naga, H.M., Gates, W.P. & Laloui, L. 2008. Temperature Effects on Volume Change and Hydraulic Properties of Geosynthetic Clay Liners. The First Pan American Geosynthetics Conference & Exhibition: 102109. Bulut, R., Lytton, R.L. & Warren, W.K. 2001. Suction measurements by filter paper method. American Society of Civil Engineers, Geotechnical Special Publication, 115: 243261. Cekerevac, C. & Laloui, L. 2004. Experimental study of thermal effects on the mechanical behaviour of a clay. Int. J. Numer. Anal. Meth. Geomech, 28: 209228. Cho, W.J., Lee, J.O. & Chun, K.S. 1999. The temperature effects on hydraulic conductivity of compacted bentonite. Applied Clay Science, 14: 4758. Franois, B. & Laloui, L. 2008. ACMEG-TS: A constitutive model for unsaturated soils under non-isothermal conditions. International Journal of Numerical and Analytical Methods in Geomechanics, 32: 19551988. Delage, P., Sultan, N. & Cui, Y.J. 2000. On the thermal consolidation of Boom clay. Canadian Geotechnical Journal, 37: 343354. Delage, P., Cui, Y.J. & Antonie, P. 2005. Geotechnical problems related with loess deposits in Northern France. Proceeding of international conference on problematic soils. Cyprus. Haghighi, A., Medero, G., Mercier. B. & Woodward, P. 2010. Temperature effects on filter paper suction measurement. Under preparation. Jennings, J.E. & Knight, K. 1957. The additional settlement of foundations due to a collapse of structure of sandy sub-soils on wetting. Proceeding of International conference on Soil Mechanics and Foundation Engineering. London: 316319. Jennings, J.E. & Knight, K. 1975. A guide to construction on or with materials exhibiting additional settlement due to collapse of grain structure. Proc. Regional Conf. for Africa on Soil Mechanics & Foundation Engineering, Rotterdam: 99105. Jotisankasa, A. Ridley, A. & Coop, M. 2007. Collapse behaviour of compacted silty clay in suction monitored oedometer apparatus. Journal of Geotechnical and Geoenvironmental Engineering, 133 (7): 867886. Lawton, E.C., Fragaszy, R.J. & Hetherington, M.D. 1992. Review of Wetting-Induced Collapse in Compacted Soil. Journal of Geotechnical Engineering, 118 (9): 13761393. Lim, Y.Y. & Miller, G.A. 2004. Wetting-Induced compression of compacted Oklahoma soils. Journal of
Geotechnical and Geoenvironmental Engineering, 130 (10): 10141023. Medero, G.M., Schnaid, F., Gehling, W.Y.Y. & Gallipoli, D. 2003. Analysis of the mechanical response of an artificial collapsible soil. Proc., From Experimental Evidence towards Numerical Modelling of Unsaturated Soils, Berlin: 135145. Medero, G.M., Schnaid, F. & Gehling, W.Y.Y. 2009. Oedometer behaviour of an artificial cemented highly collapsible soil. Journal of Geotechnical and Geoenvironmental Engineering, 135 (6): 840843. Pereira, J.H.F. & Fredlund, D.G. 2000. Volume change behaviour of collapsible compacted gneiss soil. Journal of Geotechnical and Geoenvironmental Engineering, 126 (10): 907916. Rao, S.M. & Revanasiddappa, K. 2003. Role of soil structure and matric suction in collapse of a compacted clay soil, Journal of Geotechnical Testing, 26: 102110. Romero, E., Gens, A. & Lloret, A. 2001. Temperature effects on the hydraulic behaviour of an unsaturated clay. Geotechnical and Geological Engineering, 19: 311332. Romero, E., Gens, A. & Lloret, A. 2003. Suction effects on a compacted clay under non-isothermal condition. Gotechnique, 53 (1): 6581. Romero, E., Villar, M.V. & Lloret, A. 2005. Thermohydro-mechanical behaviour of two heavily overconsolidated clays. Engineering Geology 81 (3): 255268. Rowe, R.K. 2005. Long term performance of contaminant barrier systems. Gotechnique, 55 (9): 631678. Sultan, N., Delage, P. & Cui, Y.J. 2002. Temperature effects on the volume change behaviour of Boom clay. Engineering Geology, 64: 135145. Sun, D.A., Matsuoka, H. & Xu, Y.F. 2004. Collapse behaviour of compacted clays in suction-controlled triaxial tests. Geotechnical Testing Journal, 27: 362370. Tang, A.M., Cui, Y.J. & Barnel, N. 2008. Thermomechanical behaviour of a compacted swelling clay. Gotechnique, 58 (1): 4454. Tang, A.M. & Cui, Y.J. 2005. Controlling suction by the vapour equilibrium technique at different temperatures and its application in determining the water retention properties of MX80 clay. Canadian Geotechnical Journal, 42 (1): 287296. Uchaipichat, A. & Khalili, N. 2009. Experimental investigation of thermo-hydro-mechanical behaviour of an unsaturated silt. Gotechnique, 59 (4): 339353. Villar, M.V. & Lloret, A. 2004. Influence of temperature on the hydro-mechanical behaviour of a compacted bentonite. Applied Clay Science, 26: 337350. Yesiller, N., Hanson, J.L. & Liu, W.L. 2005. Heat generation in municipal solid waste landfills. Journal of Geotechnical and Geoenvironmental Engineering, 131 (11): 13301344.
548
Influence of freeze-thaw action on mechanical behavior of unsaturated crushable volcanic soil

T. Ishikawa & S. Miura K. Itou
Hokkaido University, Sapporo, Japan Nagoya Railroad Co., Ltd., Nagoya, Japan
ABSTRACT: This paper examines the effect of freeze-thaw action on the deformation-strength characteristics of crushable volcanic coarse-grained soil in unsaturated condition. An unsaturated triaxial apparatus, which can simulate freeze-thaw sequence in real foundations, was developed and a series of monotonic triaxial compression tests for a crushable volcanic coarse-grained soil exposed to freezethaw action were performed under various degrees of saturation. Test results showed that the freeze-thaw action made the amount of particle breakage under consolidation and shear increase for crushable volcanic coarse-grained soils, and its shear strength and deformation modulus under unsaturated conditions decreased by being exposed to freeze-thaw action. These indicate that the freeze-thaw action has strong influences on the deformation-strength characteristics of a crushable volcanic soil in unsaturated conditions even if the soil lacks in frost heave characteristics. 1 INTRODUCTION triaxial compression tests for saturated specimens, the increase in degree of saturation and the cyclic freeze-thaw action induce severe particle breakage during freeze-thawing and shearing in crushable volcanic coarse-grained soils, thereby decreasing the shear strength of the soil. The objective of this study is to evaluate the effect of freeze-thaw action on the deformation-strength characteristics of crushable volcanic coarse-grained soil in unsaturated condition. In this paper, a series of monotonic triaxial compression tests for a crushable volcanic coarse-grained soil exposed to freezethaw action were performed under various degrees of saturation by using an unsaturated triaxial apparatus (Ishikawa et al. 2009), which can simulate freeze-thaw sequence in real foundations.
In cold regions like Hokkaido, the north area of Japan, the degradation in the physical properties and mechanical characteristics of soil ground caused by freeze-thaw action threatens to aggravate natural disasters such as slope failure at cut slopes or landslide at natural slopes in snow-melting season. For example, the subsurface failure often observed in snow-melting season is deemed to be caused by the increase in degree of saturation arising from snow-melting and the change in hydro-mechanical characteristics resulting from freeze-thaw action. Moreover, volcanic coarse-grained soils, which have caused complicated geotechnical engineering problems, are widely distributed across Hokkaido. The most important property in geotechnical engineering problems is that the mechanical behavior of volcanic coarse-grained soils is greatly affected by the degree of particle breakage (Miura & Yagi 2003). Crushing of volcanic soil particles can be observed even under relatively low stress levels such as subsurface layer. Accordingly, it is indispensable for establishing a precise predictive method of slope failure in Hokkaido to examine the influence of the freeze-thaw action on the hydro-mechanical behavior of unsaturated soil ground. To date, other papers of the authors (Ishikawa et al. 2007, 2008) have revealed that from the results of unsaturated triaxial compression tests for non freeze-thawed specimens and freeze-thawing
2 2.1
METHODOLOGY Freeze-thawing triaxial apparatus for unsaturated soils
A schematic diagram of a freeze-thawing triaxial apparatus for unsaturated soils is shown in Figure 1. One of key features for the apparatus is the ability to replicate one-dimensional frost heave in a triaxial specimen. The apparatus has a cooling system to control the temperature of cap and pedestal, respectively. So, any temperature difference between cap and pedestal can be arbitrarily set to
549
the specimen. Moreover, the temperature inside the triaxial cell is kept constant by circulating cold water through an opening between an inner cell and an outer cell. Besides, since the specimen is covered with an acrylic cylindrical cell for frost heave simulation during freezing and thawing and a lateral displacement of the specimen is constrained by the cell, one-dimensional frost heave can be replicated in the specimen. The apparatus can thus apply a freeze-thaw sequence to a triaxial specimen as experienced by in-situ soils in cold regions. On the back of freeze-thaw process, a triaxial compression test can be performed for coarse-grained soils under various degrees of saturation and loading conditions. The apparatus can control pore air pressure (ua) and pore water pressure (uw) during testing separately so as to set a triaxial specimen under a prescribed degree of saturation. Pore water pressure is applied to a specimen through a hydrophilic acrylic copolymer membrane filter (Versapor membrane filter, Air Entry Value (AEV) 110 kPa) attached to the pedestal, while the pore air pressure is applied through a hydrophobic polyflon filter attached to the cap. Note that a triaxial compression test can also be performed under saturated condition by removing a versapor membrane filter on a porous metal as shown in Figure 1. Axial load can be applied to the specimen by strain control method with a Direct Drive Motor (DDM). The measurement of stress and strain in freeze-thawing triaxial compression test was conducted as follows. The axial stress (a) is measured with a load cell installed inside the triaxial cell, and the axial strain (a) is measured with an external displacement transducer. The volumetric change (v) of a specimen is measured with a differential pressure gauge connected to both an inner cell and an outer cell, while the volume of drainage (Vw) during testing is measured with a double tube burette connected to the pedestal. 2.2 Soil specimens
100 Triaxial compression test 90 =49kPa, s=60kPa net : Initial 80 : After test of Nf=0 70 : After test of Nf=1 60 50 40 3 s = 2.34g/cm 30 3 dmax= 0.62g/cm 20 3 dmax= 0.46g/cm 10 Fc = 0.90% 0 0.01 0.1 1 10 Grain size(mm)
Figure 2. Physical properties and grain-size distribution curves.
Hokkaido. Belonging primarily to Shikotsu primary tephra deposited by the eruption of Shikotsu caldera 31,00034,000 years ago, Kashiwabara volcanic soil was extracted from approximately 1.5 m under ground level (lower than freezing depth). Physical properties and grain-size distribution curves of Kashiwabara volcanic soil are shown in Figure 2. Kashiwabara volcanic soil characteristically contains both coarse and fine fractions, and it has low values of both the dry densities (dmax, dmin) because its constituent particles are very porous on account of having large number of intra-particle voids. Furthermore, past studies have revealed that Kashiwabara volcanic soil shows remarkable particle crushability even under relatively low stress levels. 2.2.2 Preparation of test specimens Preparation of test specimens for Kashiwabara volcanic soil was conducted as follows. Cylindrical specimens, which sizes were initially 170 mm in height and 70 mm in diameter, were prepared by using the air pluviation method with oven-dried Kashiwabara volcanic soil. To ensure an experimental accuracy from the aspect ratio of the maximum particle size versus the specimen diameter, the grain size above 9.5 mm was screened out from the original soil while preparing the specimens for triaxial compression tests. The initial dry densities of specimens (d0) were adjusted so that the dry density after consolidation (dc) becomes 0.54 g/cm3, equal to the in-situ dry density (din-situ 0.53 0.55 g/cm3). Allowable variations of d0 were limited to within 5% of the prescribed density. 3 EXPERIMENTAL PROGRAM
2.2.1 Soil sample A test material was taken from a natural deposit of volcanic coarse-grained soil in the island of
Direct Drive Motor
(a)
Cap Pedestral
Dial Gauge to Constant Temperature Device for Cap Outer Cell Load Cell Thermometer for Cap Antifreezing Fluid Specimen with Membrane Cooling Water Frost Heave Cell Partition to Constant Temperature Device for Pedestral
(b)
70mm 170 mm
Inner Cell Poluflon Filter Versapor Membrane Filter
3.1 Freeze-thawing triaxial compression tests 3.1.1 Freeze-thaw process A freeze-thaw process was conducted as follows. First, an acrylic cylindrical cell for frost heave was installed on a soil specimen enclosed in the rubber membrane, and de-aired water was added from
Figure 1. Freeze-thawing triaxial apparatus. (a: freezethaw process, b: shear process).
550
Percent finer (%)
the bottom end of the specimen until the degree of saturation reached 80% or over. The state of the specimen after water permeation is called as quasi-saturation, hereafter. Next, the specimen was one-dimensionally consolidated, to a of 12.2 kPa by loading a weight on the top of the specimen, in consideration of overburden pressure by surface soil. Subsequently, the specimen was frozen from the upper part and thawed from the lower part with the cooling system while allowing unfrozen water in the specimen to inflow and outflow through a water plumbing path of pedestal. The temperature gradient through the specimen was maintained at 0.1C/mm, and the constant freezing and thawing velocity (U) was 1.6C/h. In this paper, the abovementioned series of operations is defined as 1 freezethaw process cycle. After the freeze-thaw process, the cell for frost heave was removed from the specimen. 3.1.2 Consolidation process A consolidation process was conducted against non freeze-thawed specimens (number of freeze-thaw process cycle, Nf 0) and freeze-thawed specimens (Nf 1) as follows. In case of unsaturated conditions, the quasi-saturated specimen was isotropically consolidated under a prescribed net normal stress (net c ua) of 49 kPa for 4 hours in fully drained condition by applying a designated confining pressure (c) of 249 kPa, pore air pressure (ua) of 200 kPa and pore water pressure (uw) of 200 kPa until there was no longer a tendency for the change in axial displacement or drainage volume. Subsequently, an unsaturated specimen under an intended suction (S ua uw) was produced by decreasing uw in stages while keeping both c and ua constant. Note that the suctions were set at 5 kPa, 15 kPa, 35 kPa and 60 kPa, which corresponded to the degrees of saturation (Sr) of approximately 49%, 42%, 34% and 27%, respectively. Water retention curves were drawn on the basis of the relationships between suction and volume of drainage during this process. On the other hands, a saturated specimen was achieved through applying negative pore water pressure, namely the double vacuuming method (Ampadu & Tatsuoka 1993), and subsequently permeating de-aired water into the quasi-saturated specimen. A back pressure of 200 kPa was then applied to ensure saturation and achieve a B-value of 0.96 at least. Next, the specimen was isotropically consolidated under a prescribed effective confining pressure (c) of 49 kPa for 4 hours in fully drained condition by applying a designated confining pressure (c) of 249 kPa and pore water pressure (uw) of 200 kPa. 3.1.3 Shear process Upon attaining an equilibrium condition in the consolidation process, a freeze-thawed specimen was continuously sheared by applying an axial
deviator stress (q) at a designated constant axial strain rate under fully drained condition (CD test) while all other testing parameters were held constant. The axial strain rate of Kashiwabara volcanic soil was set equal to 0.2%/min, and the axial compression was terminated after reaching an axial strain (a) of about 25%. In addition, ordinary triaxial compression tests were also performed under the similar experimental conditions but without the freeze-thaw process for comparison. 3.2 Particle size analysis
After testing, particle size analysis was performed for all the tested specimens to assess the degree of particle breakage that transpired throughout the test. The degree of particle breakage that the specimen undergoes was measured in terms of the increment of fine content (Fc), which is the change between the fines content of the sheared and the original specimen. The fines content (Fc) is simply the percent finer corresponding to a grain size of 0.075 mm diameter in its particle size distribution curve. 4 4.1 RESULTS AND DISCUSSIONS Behavior during freeze-thaw process
In freeze-thawing triaxial apparatus for unsaturated soils, unfrozen water in the specimen flows in and out through a versapor membrane filter attached to a water plumbing path of the pedestal during freeze-thaw process. However, the coefficient of permeability of the versapor membrane filter (k 4.4 108 m/s) is much lower than that of a porous metal used in the test method for frost susceptibility of soils (e.g. JGS 0172). To examine the influence in the coefficient of permeability of a water plumbing path on the behavior of a soil specimen during a freeze-thaw process, Figures 3 show the relations in a freeze-thaw process between temperatures of cap (Tc) and pedestal (Tp), volume of drainage (v), axial displacement of cap (u) and elapsed time (t), by comparing the results under the usage of a versapor membrane filter with those under the usage of a porous metal. Note that as the origin point of the axial displacement is designated to the point before loading, the settlement due to overburden pressure is initially observed. The frost heave can hardly be recognized in Figure 3 irrespective of the pedestal type, while a small settlement is observed in both soil specimens after thawing and the density increases. The thaw settlement in Kashiwabara volcanic soil is considered to be caused by the particle breakage due to freeze-thaw action. Moreover, other behavior of soil specimens during freeze-thaw process such as the volumes of drainage agrees well with Figures 3(a) and 3(b), quantitatively. These results
551
Axial displacement (mm)
20 15 10 5 0 -5 -10 -15 -20 0 10 20
(a)
30 40 50 Elapsed time (h)
Porous metal : Cap temp. : Pedestal temp. : Drainage : Axial disp.
60
70
0.0 -0.5 -1.0 -1.5 -2.0 -2.5 -3.0 -3.5 80
20 15 10 5 0 -5 -10 -15 -20
(b)
0 10 20
Versapor membrane filter : Cap temp. : Pedestal temp. : Drainage : Axial disp.
30 40 50 Elapsed time (h)
60
70
2.0 1.5 1.0 0.5 0.0 -0.5 -1.0 -1.5 -2.0 80
Figure 3. filter).
Comparison of frost heave behaviors between pedestal types. (a: porous metal, b: versapor membrane
100
Suction, s (kPa)
indicate that Kashiwabara volcanic soil is a non frost-heaving geotechnical material, and that a unsaturated triaxial apparatus adopting the pressure membrane method have highly conformity with conventional frost susceptible tests in terms of frost heave of volcanic coarse-grained soils. 4.2 Effects of freeze-thaw action SWCC
80 60 40 20 0
Kashiwabara volcanic soil net= 49kPa,
dc= 0.54g/cm
3
s=60kPa
s=35kPa s=15kPa s=5kPa
Nf= 0 cycle Nf= 1 cycle
Figure 4 shows the relationships of freeze-thawed specimens and non freeze-thawed specimens after isotropic consolidation between suction (S) and degree of saturation (Sr) in triaxial compression tests under various degrees of saturation. The relations correspond to the Soil-Water Characteristic Curves (SWCCs) during the drying process under net 49 kPa. At low degree of saturation (Sr 50%), suction arises in the specimen, and it sharply increases with decreasing Sr. However, even if suction increases in the range over S 60 kPa, the degree of saturation tends to converge at a specific value, namely residual degree of saturation (Sr0), of about Sr0 25% for non freeze-thawed specimens and 30% for freeze-thawed specimens, respectively. Moreover, for plots with the same degree of saturation, the suction of a freeze-thawed specimen is higher than that of a non freeze-thawed one, while the shape of the S Sr relation is almost unchanged. These results indicate that the freezethaw action has a strong influence on the water retention characteristics of crushable volcanic soils even though non frost-heaving geomaterials. It can be considered that the difference in water retentivity between two specimens is due to the increase in the fine content caused by the freeze-thaw of the specimen as shown in Figure 2. 4.3 Effects of freeze-thaw action on shear behavior
10
20
30
40
50
60

Figure 4. curves. Physical properties and grain-size distribution
Figures 5 show typical relationships of freezethawed and non freeze-thawed specimens between axial deviator stress (q), volumetric strain (v), and axial strain (a) in CD tests under different suction (S). Kashiwabara volcanic soil exhibits strain hardening until the deviator stress reaches peak
at an axial strain of about 15%, and then turns strain softening like dense sand and over consolidated clay, regardless of the presence or absence of freeze-thaw process. On the other hand, a tendency of the specimen volume easy to contract in freezethawed specimens is observed in comparison with non freeze-thawed specimens. Figure 6 shows the relationships of both types of specimens between peak shear strength (qf) and degree of saturation (Sr). For plots with the same suction, the peak shear strength of a freezethawed specimen drops as compared with that of the non freeze-thawed one irrespective of degree of saturation despite the fact that the increase in the density was caused by freeze-thawing. Besides, for plots of non freeze-thawed specimens, the peak shear strength of the dry specimen is the greatest, the saturated specimen the least, while the strength of the unsaturated specimens is intermediate between these two extremes. The decreasing tendency of peak shear strength with increasing Sr for freeze-thawed specimens is similar to that for non freeze-thawed ones. Hence, as degree of saturation increases the shear strength of the volcanic soil decreases regardless of freeze-thaw history. Figure 7 shows the relationships of both types of specimens between the secant deformation modulus (Ef) at failure, which were obtained from the stress-strain curves as shown in Figure 5, and degree of saturation (Sr). Both secant deformation moduli decrease with increasing Sr irrespective of
552
Cap temp. ( C) o Pedestal temp. ( C) Drainage (ml)
Cap temp. ( C) o Pedestal temp. ( C) Drainage (ml)
Deformation modulus at failure, Ef (Mpa)
300 Deviator stress, q (kPa) 250 200 150 100 50 0 0
(a)
30 25 20 15 10 5 0 0 20 40 60
Kashiwabara volcanic soil
net= 49kPa, dc= 0.54g/cm
3
Kashiwabara volcanic soil net=49kPa, s=5kPa, 0.2%/min : Nf=0 (Sr=47.9%) : Nf=1 (Sr=50.2%)
v
300 Deviator stress, q (kPa) 250 200 150 100 50 0 0
10 15 20 Axial strain, a (%)

q
8 6 4 2 0 25
Volumetric strain, v (%)
(b)
80
100

net=49kPa, s=60kPa, 0.2%/min
Kashiwabara volcanic soil : Nf=0 (Sr=27.3%)
v
: Nf=1 (Sr=32.4%)
10 15 20 Axial strain, a (%)
8 6 4 2 0 25
Figure 7. Influence of freeze-thaw action on secant deformation modulus at failure.
4.4
Relations of particle breakage to strength
Figure 5. Results of freeze-thawing triaxial compression tests. (a: S = 5 kPa, b: S = 60 kPa).

300
250 200 150 100 50 0 0 20

3
40
60
80
100
Figure 6. strength.
Influence of freeze-thaw action on peak shear
freeze-thaw history, and the secant deformation modulus of a freeze-thawed specimen is higher than that of a non freeze-thawed one with the same degree of saturation over all the range of Sr. These results indicate a possibility that the rise in water retentivity of freeze-thawed Kashiwabara volcanic soil caused by the increment of the fine content results in the increase in degree of saturation under the same suction, and thereby brought about the further degradation of the mechanical properties though the effect is small in comparison with that of freeze-thaw action. Accordingly, the freeze-thaw action has a profound influence on the deformation-strength characteristics of unsaturated crushable volcanic coarse-grained soils.
The degradation in the deformation-strength characteristics of a crushable volcanic coarse-grained soil may have to do with the increase in the fine content arising from freeze-thaw action and submergence (Ishikawa et al. 2007, 2008). Figure 8 shows the relations of freeze-thawed and non freezethawed specimens between increment of fine content (Fc) through a test and degree of saturation (Sr). It illustrates that particle breakage through a triaxial compression test of Kashiwabara volcanic soil becomes more pronounced by the freeze-thaw action. The graph also shows a consistent pattern of particle breakage, which indicates that saturated specimens were crushed to a greater extent than dry specimens during triaxial testing, with unsaturated specimens being intermediate. The tendencies similar to the above-mentioned results are observed in the relations of peak shear strength and secant deformation modulus to degree of saturation as shown in Figure 6 and Figure 7, respectively. Accordingly, the particle breakage is closely related to the mechanical behavior of crushable volcanic soils under not only saturated condition but also unsaturated ones. Figure 9 and Figure 10 show the relations of both types of specimens between increment of fine content (Fc) through a test and peak shear strength (qf) and the relations between Fc and secant deformation modulus (Ef) at failure, respectively. It is recognized that the peak shear strength and secant deformation modulus at failure decrease inversely with the increase in the increment of fine content, and the relations seems to approximate a line with a slight dispersion regardless of the presence or absence of freeze-thaw process, though plots for freeze-thawed specimens are found at the lower right as compared with those of non freeze-thawed ones. From these results, it seems reasonable to conclude that the decrease in peak strength and shear modulus of
Maximum deviator stress, qf (kPa)
553
5 Increment of fine content, Fc (%)

Maximum deviator stress, qf (kPa)
300 250 200 150 100 50 0 0 1 2 3

3
4 3 2 1 0
Kashiwabara volcanic soil 3 net= 49kPa, dc= 0.54g/cm Nf= 0 cycle Nf= 1 cycle
20 40 60 80 Degree of saturation, Sr (%)
100
Figure 9. strength.
Increment of fine content, Fc (%)
Influence of freeze-thaw action on peak shear
Figure 8. breakage.
Influence of freeze-thaw action on particle
Deformation modulus at failure, Ef (Mpa)
crushable coarse-grained volcanic soils suffered from freeze-thaw action mainly originates in the increment of particle breakage during consolidation and shear processes irrespective of the degree of saturation. 5 CONCLUSIONS
30 25 20 15 10 5 0 0 1 2 3
3
The following findings can be obtained; The freeze-thaw action made the amount of particle breakage under consolidation and shear increase for crushable volcanic coarse-grained soils The increase in the fine content caused by freeze-thaw results in the rise in water retentivity of crushable coarse-grained volcanic soils. The shear strength and deformation modulus under unsaturated conditions decrease with the increment of particle breakage by freeze-thaw action. These indicate that the freeze-thaw action has a strong influence on the deformation-strength characteristics of a crushable volcanic coarse-grained soil in unsaturated condition even if the soil lacks in frost heave characteristics. However, there is room for further investigation as to the difference in the frost heave characteristics between saturated soils and unsaturated soils, and the influence on the deformation-strength characteristics. REFERENCES
Ampadu, S.K. & Tatsuoka, F. 1993. Effect of setting method on the behaviour of clays in triaxial compression from saturation to undrained shear. Soils and Foundations, 33(2): 1434. Ishikawa, T., Miura, S., Ito, K. & Ozaki, Y. 2008. Influence of freeze-thaw action on mechanical behavior of
Increment of fine content, Fc (%)
Figure 10. Influence of freeze-thaw action on particle breakage. saturated crushable volcanic soil. In Susan, E.B. et. al. (eds), Deformation Characteristics of Geomaterials, Proc. intern. symp., Atlanta, 2224 September 2008.: 557564. Rotterdam: Millpress. Ishikawa, T., Miura, S. & Tokoro, T. 2009. Effect evaluation of freeze-thaw action on hydro-mechanical behavior of unsaturated granular materials. In Hamza, M. et al. (eds), 17th International Conference on Soil Mechanics and Geotechnical Engineering, Proc. intern. conf., Vol. 1, Alexandria, 59 October 2009.: 833836. Rotterdam: Millpress. Ishikawa, T., Ozaki, Y. & Miura, S. 2007. Influence of degree of saturation on strength characteristics of crushable volcanic coarse-grained soils, In Zongze, Y. et al. (eds), The 3rd Asian Conference on Unsaturated Soils, Proc. intern. conf., Nanjing, 2123 April 2007.: 223229. Beijing: Science Press. Miura, S. & Yagi, K. 2003. Mechanical behavior and particle crushing of volcanic coarse-grained soils in Japan. In Tan, et al. (eds), Characterisation and Engineering Properties of Natural Soils, Proc. intern. wks., Singapore, 24 December 2002. 11691203. Lisse, Swets and Zeitlinger.
554
Influence of temperature on the water retention capacity of soils analysis using solution thermodynamics
A.C. Jacinto & A. Ledesma
Technical University of CatalunyaUPC, Barcelona, Spain
M.V. Villar & R. Gmez-Espina
Research Centre for Energy, Environment and TechnologyCIEMAT, Madrid, Spain
ABSTRACT: Water potential is the energy of water in a soil resulting from the different forces acting upon it. Water potential is a measure of the capacity of the soil to hold water in its pores. Experimental results show that the soil water retention capacity tends to reduce with increasing temperature. The paper presents a formulation developed using concepts of thermodynamics of adsorption that relate changes in the energy status of water in soil with changes in temperature. Experimental results of water retention capacity obtained on compacted samples of an expansive clay have been analysed using this approach and that deduced using the capillary model (i.e. considering the changes in the surface tension of water by temperature increments). The results suggest that mechanisms other than the change in the surface tension affect the retention capacity of water in soils as temperature increases. 1 INTRODUCTION variation of soil suction with respect to temperature, for constant water content, as s T
w
Soils consist of a solid skeleton with pores in between. The pores have different sizes, shapes and spatial distributions and provide the space for storage and transport of liquid and gas. Water molecules in soils experience many different forces. However, it is possible to calculate the potential energy of a unit quantity of water resulting from the forces acting upon it. This energy is a measure of the capacity of the soil to hold water in its pores. In Soil Mechanics, the equivalent of the energy potential expressed on volume basis and with opposite sign is called suction s (Yong 1999, Gens & Olivella 2001). Although it is generally acknowledged that suction in soils is not exclusively a capillary process, a first analysis of experimental results can be carried out if it is assumed that the soil suction coincides with the capillary pressure given by s Pc Pg Pl 2 gl cos sl r (1)
s gl gl T
(2)
where Pc is capillary pressure, gl is the surface tension between fluids (liquid and gas), sl is the angle at which the liquid meets the capillary, and r is the radius of the capillary tube. Philip & de Vries (1957) assumed that the surface tension is the only temperature dependent parameter in Equation 1. By mathematical derivation of this equation it is possible to obtain the
where w is the water content. Experimental results of temperature effects on water retention capacity in sand and silt loam described in the literature could not be explained from changes in the surface tension (Haridasan & Jensen 1972, Hopmans & Dane 1986, Nimmo & Miller 1986, Constantz 1991, She & Sleep 1998). Generally, the temperature effects were larger than those predicted by using Equation 2. The temperature influence on the water retention capacity of clayey soils received special attention in the last decades (Romero et al. 2001, Villar & Lloret 2004, Villar et al. 2006, Villar & GmezEspina 2007). In compacted samples of clayey soils, it is assumed that the temperature dependence on surface tension is the main mechanism affecting the soil water retention capacity when capillary effects dominate, that is, for high water contents. However, for low water contents, the main temperature effects on the water retention capacity are due to changes in the interaction mechanisms between clay particles and water (Romero et al. 2001). As an alternative to the analysis using the capillary model, thermodynamic of adsorption gives a tool to study the physical adsorption of water by soils (Hill 1950, Myers 2002, Myers & Monson 2002). For an
555
adsorbent such as clay, adsorption occurs in the different pores present in the material. Surface thermodynamics is able to explain the physical adsorption of thin films on solid surfaces but its application to porous materials is of very limited utility because of the complex geometry of the interfaces in the systems. Nevertheless, the thermodynamics of adsorption can be treated as a special case of solution thermodynamics for which the adsorbates are the solutes and the adsorbent is the solvent. This work investigates the influence of high temperatures on the water retention capacity of compacted bentonite. Experimental data of water retention capacity obtained on compacted samples of an expansive clay (MX-80 bentonite) were analysed. The results suggest that mechanisms other than the change in the surface tension affect the retention capacity of water in soils as temperature increases. 2 THERMODYNAMIC OF ADSORPTION
flexible adsorbents for which the density varies with the amount adsorbed (as it is the case of swelling clays) or under the application of external loads. In what follows, all the extensive variables are written per unit mass of adsorbent and called mass extensive variables. For simplification, no distinction is made between the symbols used to indicate variables on solid mass basis (Myers 2002, Myers & Monson 2002). Because all variables in the internal energy are extensive quantities, it is possible to apply the Eulers homogeneous function theorem to Equation 3 and obtain U TS PV i ni
i 1 c
(4)
The excess energy Ue is given by U e U U g U s (5)
This Section presents a summary of the thermodynamic approach used in this work. A detailed treatment of this subject can be found in Myers (2002) and Myers & Monson (2002). Thermodynamic equations for bulk solutions are extended to adsorption by adding a term corresponding to the solid adsorbent, which in a sense can be viewed as a solvent (Myers & Monson 2002). The differential of the internal energy becomes dU TdS PdV i dni dm
i 1 c
The excess energy is the total energy of the system (gas phase plus solid phase) minus the energy of the gas phase (Ug) minus the energy of the clean solid adsorbent (Us) at the equilibrium temperature and pressure. The excess volume Ve results V e V V g V s 0 (6)
From Equations 46, the internal energy of the adsorbed phase Ue is U e TS e i nie
i 1 c
(3)
(7)
where the intensive variables are the absolute temperature T, the external pressure P, the chemical potential of the adsorbates i, and the chemical potential of the solid adsorbent defined on a mass basis. The extensive variables are the internal energy U, the entropy S, the amount of moles of each adsorbate ni, and the mass of solid adsorbent m. Equation 3 applies to a solid phase containing c species of solutes (adsorbates) inside the pores of the adsorbent. The chemical potential of the adsorbent changes with the amount of material adsorbed. A porous medium can be considered as composed of a matrix and a porous space (Coussy 2004). In this approach it is assumed that the connected porous space (the space through which two points can be joined by a path lying entirely within the medium) coincides with the gas phase volume Vg. The matrix is composed of both a solid part and a possible occluded porosity, whether saturated or not, and its volume is given by Vs. Therefore, the total volume of the system V is divided into the gas phase volume and the solid phase volume. Both volumes can undergo changes when the thermodynamic variables are changed. Equation 3 therefore can be applied to
where Se is the entropy of the adsorbed phase, nie is the amount adsorbed of the i component presents in the gas phase, and the surface potential or chemical potential of the solid adsorbent relative to its pure standard state at the equilibrium pressure and temperature. They are given by Se S Sg Ss s
s
(8)
If no adsorption occurs, is equal to and the surface potential is zero. Using the Legendre transformations for the auxiliary functions in the adsorbed phase, the following results (Myers 2002) Ue He
c
(9a) (9b)
G e Ae H e TS e i nie
i 1
where He, Ae, and Ge mean enthalpy, Helmholtz free energy, and Gibbs free energy respectively. The free energy of the adsorbed phase consists of the free energy of nie moles of each species in equilibrium with the gaseous phase plus the surface potential , which is zero if no adsorption takes place.
556
Suction (MPa)
The excess integral functions for the adsorbed phase are defined relative to the perfect gas reference state at the same temperature (Myers 2002, Myers & Monson 2002). G e H e T S e nie i i0 H e nie hi hi0 T 2
i 1 c c c
1.0E+03 20C 25C 1.0E+02 40C 40C 80C 80C
i 1
(10a) (10b) (10c)
1.0E+01
S e nie si si0 T P , yi i 1
T T P , yi
1.0E+00 0.0
10.0 20.0 Water content (%)
30.0
The quantities i0, si0, and hi0 refer to the molar values in the perfect gas reference state. The overline notation means partial molar variables in the bulk gas phase. Differentiation of Equations 10a-c gives the differential functions for a component i in the adsorbed phase (Myers 2002, Myers & Monson 2002) f G e i i0 RT ln i0 gie e fi ni T ,ne
j
Figure 1. Experimental values of soil suction against water content at different temperatures and dry densities. Empty symbols 1.60 Mg/m3, filled symbols 1.75 Mg/m3.
top and bottom in order to allow water exchange in the vapour phase with the surrounding atmosphere, whose RH was controlled inside desiccators placed in ovens (Villar & Lloret 2004). 3.2 Test results
(11a)
H ln fi (11b) hi hi0 R hi e e 1 T ne ,ne n i T , ne i j j

e
The calculation of the potential of water in soil (or soil suction) on the basis of relative humidity is accomplished through the thermodynamics relationship (Edlefsen & Anderson 1943) M M s (12) RH exp w w exp w l RT l RT with Mw the molar mass of water, R the gas constant and w the water potential. The symbol is used here to indicate that the soil water potential is defined relative to the perfect gas reference state at the same temperature. Figure 1 presents the relation between the soil suction and the water content in samples tested under isochoric conditions and at different temperatures (Villar et al. 2006, Villar & Gmez-Espina 2007). The difference among the values was more important in the high water content zone and the trend indicates that at saturated state the water content decreased when the temperature increases. 4 4.1 ANALYSIS OF EXPERIMENTAL RESULTS Total and liquid water content
S e f si si0 sie RT ln i0 e T fi ne ,ne ni T ,nej i j (11c) where g e , h e , and s e are the differential Gibbs free energy, enthalpy, and entropy respectively, and fi and fi0 are the fugacity and the standard-state fugacity respectively. 3 3.1 EXPERIMENTAL RESULTS Material and experimental methods
The MX-80 bentonite is extracted from Wyoming (USA). It is a worldwide known commercial material supplied in the form of powder homoionised to sodium. The MX-80 bentonite consists mainly of montmorillonite (6582%). Two methods were applied to determine the retention capacity at constant volume and at different temperatures. One of them consisted in measuring the relative humidity (RH) of clay blocks at particular water contents while they were kept inside stainless steel hermetic cells whose temperature could be regulated (Villar et al. 2006). In the other method the clay was compacted inside cells specially designed to avoid the swelling of the clay in wetting paths. The cells were perforated on
The liquid degree of saturation in soils, commonly referred to as degree of saturation, is defined as Sl
d w l
(13)
where is the porosity, d the soil dry density, and l the liquid density. The initial water content in a soil (corresponding to a given temperature and pressure) defines
557
a suction value (or soil water potential). This suction can be evaluated through the relative humidity measured in the soil pores by using Equation 12. At ambient temperature, the degree of saturation in the soil is defined by the mass of liquid water. However, if the soil sample is heated, liquid water evaporates until a dynamic equilibrium is obtained between this and the water in the gas phase (vapour). If the sample is in a closed recipient the water content will be constant but, as more liquid water becomes vapour, the liquid degree of saturation will decrease. By considering the initial liquid degree of saturation and porosity of a soil sample, it is possible to define the initial volume occupied by the liquid phase (Vl) and gas phase (Vg). From the initial conditions, the moles of vapour and air can also be determined. It is assumed that the gas phase behaves as an ideal gas. Therefore, it is possible to determine the 0 variation of the saturated vapour density v . Then, if the relative humidity at different temperatures is determined experimentally, it is possible to calculate the vapour density v using the relation
1.0E+03
Suction (MPa)
1.0E+02
1.0E+01 25C 30C 1.0E+00 0.0 40C 40C 60C 60C 80C 80C 20.0
5.0 10.0 15.0 Liquid water content (%)
Figure 2. Experimental values of suction against liquid water content at different temperatures and dry densities. Continuous lines correspond to a dry density of 1.60 Mg/m3 and dashed lines to a dry density of 1.75 Mg/m3.
w g e h e T s e g e s e T n e
(16)
The differential entropy can be calculated as (17)
RH v v0
(14)
The gas phase volume and vapour density define the mass of water in the gaseous phase (vapour w mass) mg . Finally, the water mass in the liquid phase mlw is calculated as
w mlw mw mg
(15)
with ne the amount adsorbed. The differential entropy is the slope in a plot of the water potential as a function of the temperature (at constant amount adsorbed). The differential enthalpy can be obtained dividing Equation 16 by T 2 and rearranging terms, which gives h e T 2 g e T T n e (18)
where mw is the total mass of water in the sample. In the cell method, the soil water potential was fixed and once equilibrium attained, the water content was gravimetrically determined. In the sensor/ cell method the total water content was initially fixed and then the water potential was determined at different temperatures. In the latter case, although the total water content was constant, as the temperature increased more liquid water became vapour (water in the gas phase). Although the two methods used to obtain the laboratory data have different theoretical principles, the results obtained with both methodologies were found to be coherent (Villar et al. 2006, Villar & Gmez-Espina 2007). The concepts described above were applied to the results shown in Figure 1. Figure 2 shows values of liquid water content and the corresponding values of suction for different temperatures and sample dry density. Straight lines in the figures were used to define suction values at intermediate liquid water contents. 4.2 Analysis using solution thermodynamics
When the chemical potential of soil water is determined through RH, it is convenient express Equation 18 in the following form ln RH h e R 1 T ne (19)
Using the solution thermodynamic approach, the following expression for the potential of soil water can be deduced (Eq. 11a)
By use of Equation 19 and the slope in a plot of ln(RH) as a function of (1/T) it is possible to calculate the differential enthalpy. Figure 3 is a plot of the water potential as a function of temperature, at different liquid water contents and for each dry density considered in the experiments. Figure 4 presents values of ln(RH) as a function of the inverse of temperature for the same cases considered before. The slopes of the straight lines in Figure 3 with changed sign represent the value of the differential entropy at a given water content. From the slopes in Figure 4 the differential enthalpy can be calculated using Equation 19. The values obtained are independent on the temperature, but vary with the water content (Table 1).
558
0.0 Water potential (kJ/kg) -50.0 -100.0 -150.0 -200.0 -250.0 280.0 5.0% 7.5% 10.0% 12.5% 15.0% 300.0 320.0 340.0 Temperature (K) 360.0 5.0% 7.5% 10.0% 12.5% 15.0% 380.0
2 can be written for the dependence of the water potential on temperature changes (the contact angle is assumed temperature independent). w gl w T w gl T (21)
The surface tension of most liquid-gas interfaces decreases with increasing temperature in a nearly linear fashion (Adamson 1990). For water, it is assumed that this dependence is given by
Figure 3. Water potential against temperature at different water contents. Continuous lines correspond to dry density of 1.60 Mg/m3 and dashed lines to dry density of 1.75 Mg/m3.
0.0
gl a bT
(22)
where a and b are constants to be determined by fitting experimental results. From Equations 21 and 22 a relation between the potential of water in soil and temperature changes is obtained d w b dT w a bT (23)
-0.6 ln (RH)
-1.2
5.0% 7.5% 10.0% 12.5% 15.0%
5.0% 7.5% 10.0% 12.5% 15.0% 3.0E-03 1/T (1/K) 3.2E-03 3.4E-03
Then, integrating the above expression yields

r w w
-1.8 2.6E-03
gl
r gl
2.8E-03
(24)
Figure 4. ln(RH) against 1/T at different water contents. Continuous lines correspond to dry density of 1.60 Mg/m3 and dashed lines to dry density of 1.75 Mg/m3. Table 1. Values of the differential entropy and enthalpy. 1.60 Mg/m3 w% 5.0 7.5 10.0 12.5 15.0 s kJ/(K kg) 2.7 1.8 1.2 0.8 0.5
e
1.75 Mg/m3 h kJ/kg 1008.5 672.1 441.7 287.8 186.4

e
s kJ/(K kg) 1.7 1.3 0.9 0.7 0.5
h kJ/kg 706.8 512.7 371.8 269.5 195.3
with w and wr being the water potentials, and gl and glr the surface tensions respectively, at temperature T and at reference temperature Tr. Given a value of water potential and temperature, Equation 24 defines the evolution of the potential of water in soil as the temperature changes. Considering the expression adopted for the surface tension (Eq. 22), Equation 24 can be rewritten in a similar form to Equation 16 w 0 T 1 with the coefficients 0 and 1 defined as (25)
0
r w r a 1 rw b r gl gl
(26)
4.3
Analysis using the capillary model
The capillary potential of water in soil expressed on mass basis c can be deduced from Equation (1) c 2 gl cos sl
l r
(20)
Table 2 shows the values of coefficients 0 and 1 obtained deduced from the experimental results presented above. These coefficients were calculated considering a reference value of the water chemir cal potential (w ) at each water content and for a reference temperature (Tr) equal to 25C. Coefficients a and b were assumed equal to 0.118 N/m and 1.54 104 N/(K m) respectively (Adamson 1990). 4.4 Comparison of results The differences between the corresponding results in Table 1 and 2 can be partially explained by
with l the liquid density. Since this model assumes that the capillary potential coincides with the water potential (i.e. c w), an expression similar to Equation
559
Table 2.
Values of the coefficients 0 and 1. 1.60 Mg/m3 1.75 Mg/m3 0 kJ/kg 353.7 227.3 146.1 93.9 60.4 1 kJ/(K kg) 0.4 0.3 0.2 0.2 0.1 0 kJ/kg 303.4 221.9 162.3 118.7 86.8
w% 5.0 7.5 10.0 12.5 15.0
1 kJ/(K kg) 0.5 0.3 0.2 0.1 0.1
Thermodynamics of adsorption is an alternative tool to analyse the influence of temperature on the retention capacity of soils. The solution obtained using thermodynamic concepts is formally equal to that which results from applying the capillary model. Because of that, the temperature effect on the soil retention capacity can be incorporated into numerical analyses using the same strategy as that traditionally used when applying the capillary model. REFERENCES
Adamson, A.W. 1990. Physical chemistry of surfaces. Fifth edition. New York, John Wiley & Sons. Constantz, J. 1991. Comparison of isothermal and isobaric water retention paths in nonswelling porous materials. Water Resour. Res. 27(12): 31653170. Coussy, O. 2004. Poromechanics. John Wiley & Sons, Chichester. Edlefsen, N.E. & Anderson, A.B.C. 1943. Thermodynamics of soil moisture. Hilgardia 15(2): 31298. Gens, A. & Olivella, S. 2001. THM phenomena in saturated and unsaturated porous media. Fundamentals and formulation. Revue Franaise de Gnie Civil 5(6): 693717. Haridasan, M. & Jensen, R.D. 1972. Effect of temperature on pressure head-water content relationship and conductivity of two soils. Soil Sci. Soc. Amer. Proc. 36: 703708. Hill, T.L. 1950. Statistical mechanics of adsorption. IX. Adsorption thermodynamics and solution thermodynamics. J. Chem. Physics 18(3): 246256. Hopmans, J.W. & Dane J.H. 1986. Temperature dependence of soil water retention curves. Soil Sci. Soc. Am. J. 50: 562567. Myers, A.L. 2002. Thermodynamics of adsorption in porous materials. Am. Inst. Chem. Eng. 48(1): 145160. Myers, A.L. & Monson, P.A. 2002. Adsorption in porous materials at high pressure: theory and experiment. Langmuir 18: 1026110273. Nimmo, J.R. & Miller, E.E. 1986. The temperature dependence of isothermal moisture vs. potential characteristics of soils. Soil Sci. Soc. Am. J. 50: 11051113. Philip, J.R. & de Vries, D.A. 1957. Moisture movement in porous materials under temperature gradients. Trans. Am. Geophys. Union 38(2): 222232. Romero, E., Gens, A. & Lloret, A. 2001. Temperature effects on the hydraulic behaviour of an unsaturated clay. Geotech. Geol. Eng. 19: 311332. She, H.Y. & Sleep, B.E. 1998. The effect of temperature on capillary pressure-saturation relationships for airwater and perchloroethylene-water systems. Water Resour. Res. 34(10): 25872597. Villar, M.V. & Lloret, A. 2004. Influence of temperature on the hydro-mechanical behaviour of a compacted bentonite. Appl. Clay Sci. 26: 337350. Villar, M.V. & Gmez-Espina, R. 2007. Retention curves of two bentonites at high temperature. Experimental Unsaturated Soil Mechanics. Springer Proceedings in Physics 112: 267274. Villar, M.V., Gmez-Espina, R. & Martn, P.L. 2006. Behaviour of MX-80 bentonite at unsaturated conditions and under thermo-hydraulic gradient: p. 51 Technical Report 1081, Madrid, Ciemat.
comparing Equations 16 and 25. Although they are formally similar, the coefficients in Equation 16 include the interaction between the adsorbed water and the soil. This last interaction is considered through the surface potential , which varies as a function of the amount adsorbed (Section 2). Recalling Equations 17 and 24, it is possible to write w w T T
w
(27)
where the expression of the coefficient depends on the case considered. Thus, when the results are analysed using solution thermodynamics the following expression is obtained for this coefficient

h e s e
(28)
Values of were calculated for the two densities analysed in this work. Although both h e and s e depend on the water content, there is not a clear tendency about the dependence of on that variable. For the low density samples a value of 374.3 2.1 K was obtained whereas in the case of the high density samples the value was 410.0 1.9 K. When the capillary model is considered

1 a 0 b
(29)
and in this case will be a constant, as the coefficients a and b are independent on the water content. Replacing the values adopted for a and b in Equation 29 a value of equal to 766.2 K is obtained. 5 CONCLUSIONS
The influence of temperature on the water retention capacity of compacted samples of MX-80 bentonite has been analysed using experimental data reported in the literature. The retention capacity decreases with temperature, more than predicted by the change in surface tension.
560
Influence of temperature rise under undrained condition on the shear behavior of unsaturated sandy soil considering solubility of gaseous components
Y. Katsuno, M. Iwabuchi, K. Ozawa, R. Uzuoka & M. Kazama
Tohoku University, Sendai, Japan
ABSTRACT: Both monotonic and cyclic shear strength of unsaturated soils are significantly affected by an initial effective stress condition and a degree of saturation. In this paper, we have studied influence of temperature change under closed condition on the shear behavior of unsaturated sandy soil experimentally. In addition, to consider the effects of solubility of gaseous component to the pore fluid, air and CO2 are used as a gaseous material. The solubility of CO2 into water is larger than ordinary air. Because the solubility of gaseous material varies with the temperature and the pressure, not only initial effective stress state but also the stress state during shearing process under closed condition are influenced. From the test results, it was found that the shear strength for the specimen subjected to temperature rise was smaller and that the shear strength for the specimen with CO2 was smaller than that with air. 1 1.1 INTRODUCTION Background and purpose Hilf (1948) developed the expression for the change in the pore gas pressure of unsaturated soil subjected to shear loading, taking Boyles law and Henrys law into consideration. Grozic et al. proposed a constitutive model, using the Hilfs equation, and investigated the behavior of unsaturated soil subjected to monotonic shearing. They supposed loose sand at several degrees of saturation, and compared the experimental results and analytical one. When the degree of saturation is over 90%, the model provided reproducible results. With the same model, they analyzed the monotonic shear behavior of soil with CH4 and CO2 as pore gases. (The solubility of CO2 is more than 10 times higher than that of CH4 gas.) In this paper, we discuss the influences of temperature changes on shearing behavior of unsaturated sandy soils, focusing on stress state change caused by the degree of solubility of the pore gases. For comparison purpose, we used CO2 as a pore gas material. 1.2 Fundamental relationships Effective stress of unsaturated soil is expressed by Bishops equation below. ( ug) (ug uw) (1)
In recent years, problems concerning thermal environment of ground such as waste repository has been one of the geotechnical engineering subjects. Dumping grounds for radioactive wastes or industrial waste disposal plant, deformation of ground during the mining of methane hydrate are also the other examples. In those grounds, liquid and air phases are subjected to the temperature change, and it results in the change of effective stress state of the soils. Furthermore in the country with high seismic activity like Japan, the seismic stability of the waste repository should be confirmed. It is known that even if unsaturated soils liquify during the earthquake, if the degree of saturation is high enough (Unno et al. 2008, and Kazama et al. 2007). It is necessary to evaluate the influence of thermal environment on the shear strength of unsaturated soils. When temperature changes, it leads to changes in volume and pressure of pore gas of soils. The variations of volume and pressure of pore gas are governed by Boyle-Charles law and Henrys law. On the other hand, the volume of soil is affected by shear loading because soils have a dilative/compressive nature. It results in changes in volume and pressure of pore gas. Thus, interaction between temperature change and the shearing behavior of soils should be clarlified. However, so far, few attempts have been made to study on this subject.
where, effective stress, total stress, ug: pore gas pressure (including water vapor pressure), uw: pore water pressure, suction parameter (in this research, we used the degree of saturation Sr).
561
Figure 1 shows the solubility of CO2 and air into water at certain temperature, that is solubility curve. In case of relatively low solubility like air and CO2, Henrys law is consistent under the testing condition here. Therefore, the amount of dissolution is proportional to the partial pressure of pore gas P (ug uwv). The volume Vgd of dissolved gas into pore water at a certain temperature and partial pressure is obtained theoretically. Change in state of pore gas is expressed by Boyle-Charles law. In undrained condition, we can use the equation (2) at any time. (u g u wv )( Vg Vgd ) T constant (2)
2 2.1
EXPERIMENTAL METHOD Testing triaxial apparatus
where, uwv: water vapor pressure, Vg: volume of pore gas, Vgd: volume of dissolved gas (supposed to be in air phase), T temperature. In an equilibrium state, partial pressure of dry gas equals that of dissolved gas. Supposed that dissolution and the equilibrium state of pore gas, the pore gas pressure after changes in temprature T or changes in volume Vg is expressed by equation (3). ug
Vg0 Vgd0 T0 (u g0 u wv0 ) ( V V g0 g ) Vgd T0 T
u wv
(3)
Figure 2 shows the schematic diagram of testing apparatus for cylindrical specimens 50 mm in diameter and 100 mm in height. A glass fiber filter and ceramic disk with an AEV of 50 or 200 kPa air were installed at the top and bottom of specimen, respectively. A step motor was used for strain-controlled monotonic/cyclic loading, and the volume change of the whole specimen was measured directly from the differential pressure meter between the inner cell and outer water head, and its measurement capacity was 25% volumetric strain. Continuous measurements of the pore water and gas pressure were taken to obtain the soil suction. The pore air pressure during shear was measured by the air transducer attached directly above the specimen. A solenoid valve was attached in order to avoid the effect of gas compressibility in the pipe line. The volume of the inner pipe from the top of the specimen to the solenoid valve was 0.18 cm3, which is considered small enough to allow for accuracy. Temperature of a specimen can be controlled by band-heater attached to the inner cell. 2.2 Physical property of testing sandy soils Figure 3 shows the grain size distributions of a fine sand (Toyoura sand) and a volcanic sand (Tsukidate sand) used in this study. The fine sand has no fine content and a specific gravity of 2.643. The minimum and maximum void ratios are 0.609 and 0.967, respectively. On the other hand, the volcanic sand has about 30% fine content and the
where, index 0 in the equation indicates the quantity of initial state when the pore gas pressure is in the equilibrium condition. Here, water vapor pressure uwv can be obtained from thoeretical relationship between suction and relative humidity. Multiplying ratio (P/P0) is expressed like the equation below when a change in volume of soil skelton V or a change in temperature of specimen T occurs. Vg0 Vgd0 P T0 T P0 T0 ( Vg0 Vg ) Vgd
Air CO2
(4)
Volumetric coefficient of solubility (CC: gas/CC: liquid)
0 0 20 40 Temperature (C) 60
Figure 1.
Solubility of air and CO2 into water.
Figure 2.
Testing apparatus.
562
100 Percent finer by weight (%) 80 60 40 20 0 0.001

Volcanic sand (Tsukidate soil) Fine sand (Toyoura sand)
Table 1.
Test cases. Drainage condition Gas Shear D D U U D D U D D U D D U U U U Air CO2 Air CO2 Air Air Air Air Air Air Air CO2 Air CO2 Air CO2 M M M M C C C C C C C C C C C C
net Sr T Case Sand [kPa] [%] [C] 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 F F F F V V V V V V F F F F F F 20 20 20 20 20 20 20 20 20 20 20 20 20 20 40 40 90 90 90 90 40 40 40 80 80 80 90 90 90 90 90 90 20 20 20 40 20 40 20 20 40 20 40 20 20 40 20 40 20 20 20 40 20 40 20 40 20 40
0.01
0.1 1 Particle size (mm)
10
100
Figure 3. Grain size distribution curves.
specific gravity of 2.478 because of the porous nature of the soil particles. 2.3 Test conditions
Table 1 depicts the test conditions of the triaxial tests. The initial soil conditions were the testing parameters: the initial degree of saturation (40, 80, 90%), the net stress (net0 20 kPa and 40 kPa in target value), which corresponds to shallow ground. 2.4 Specimen preparation method
M: monotonic shear C: Cyclic shear. All shearing processes were under undrained condition. F: Fine sand (Toyoura sand) V: Volcanic sand (Tsukidate sand). For temperature change process D: drained, U: undrained.
Axial strain (%)
2 0 0.2% 0.4% 2 0 3600 0.8% 1.2% 1.6%
The initial stress condition including the soil water condition was achieved in the following way. To begin with, mold was filled with distilled water and dried sand is poured into the mold through the water. Here, in case of CO2 specimen, the soda water, which was made by adding CO2 gas to deaired water under 200 kPa, was used instead of distilled water. After falling the sand into the soda water and setting the upper pedestal, residual air in the upper portion was replaced by CO2 gas. In the consolidation process, the confining pressure is applied step by step. Because of the difficulty in controlling the gas pressure to achieve the prescribed initial degree of saturation, we controlled the drained water volume from the initial condition (i.e. through the drying process). That is, when the target degree of saturation was achieved during the consolidation process, the water route was closed and the initial pore water pressure is equal to the atmospheric pressure 101.3 kPa. After that, we waited the achievement of equilibrium state. We considered that the equilibrium state was achieved when the fluctuation of the pore air pressure and hydraulic pressure had been less than 1.0 kPa for a half day. 2.5 Loading conditions
2.0% 10800
7200 Time (s)
Figure 4.
Applied cyclic axial strain.
Series of strain-controlled monotonic/cyclic triaxial tests were conducted under the undrained
condition. For our purposes, an undrained condition is one in which the migration of pore water and air to the outside of specimen is not allowed during shearing. In other words, this condition assumes that the loading rate is more rapid than the migration rate of pore water and gas. After achieving the initial condition, cyclic axial strain was applied. Figure 5 shows the cyclic shear history used. The single amplitudes of the axial strains of the sinusoidal wave were 0.2, 0.4, 0.8, 1.2, 1.6, and 2.0% with every ten cycles. The concept of strain controlled cyclic shear test is shown in the literature by the last author (Kazama et al., 2000). The frequency of the sinusoidal wave is 0.005 Hz. This loading rate is slow enough to achieve an equilibrium condition between air and water pressure. This point was confirmed from preliminary tests.
563
Multiplying ratio (P/P0)
Air (Theory) Air (Experiment, Case13) CO2 (Theory) CO2 (Experiment, Case14)
Table 2. Test condition of specimen after consolidation (just before shear loading). ug Case [kPa] 1 2 3 4 91.2 84.1 104.8 99.0 161.0 164.4 166.6 121.6 121.8 120.0 87.8 81.2 94.3 108.4 94.8 112.9 uw s [kPa] [kPa] e 88.3 82.7 104.7 98.5 2.9 1.4 0.1 0.5 0.756 0.757 0.754 0.755 0.981 0.980 0.963 0.891 0.943 0.963 0.756 0.752 0.748 0.756 0.743 0.747 Sr [%] P c [kPa] [kPa] 112.1 104.0 113.2 104.6 180.4 181.5 171.1 143.5 142.7 127.3 107.7 102.2 104.7 110.8 121.6 132.5
1.5
91.2 23.3 90.2 21.0 89.9 12.1 90.0 5.9 41.3 41.4 44.2 79.6 76.4 73.6 90.0 90.6 91.5 90.6 91.5 91.2 41.7 35.9 14.2 28.0 28.9 14.4 21.7 21.9 11.6 3.9 28.2 21.1
1 20 30 Temperature ( ) 40
Figure 5. Example of pore gas pressure ratio variation during temperature rise process under closed condition.
5 6 7 8 9 10 11 12 13 14 15 16
107.6 53.4 120.9 40.6 141.9 24.7 114.3 7.3 113.1 8.7 112.3 7.6 85.8 80.2 94.2 108.0 94.7 112.6 2.0 1.0 0.1 0.4 0.1 0.3
3 3.1
TEST RESULTS AND DISCUSSION Initial condition
Before monotonic or cyclic shear stress is applied, from an initial thermo-mechanical equilibrium condition (20C, a given degree of saturation, atmospheric pressure as a gas pressure and net stress 20 kPa), temperature increased stepwise up to 40C under closed condition and achieved new equilibrium condition. Table 2 shows the condition of specimen just before the shear loading (after consolidation and temperature rise). Figure 5 shows the example of pore gas pressure variation during temperature rise process for Case-13 (air) and Case-14 (CO2). In the figure, solid line and short dashed line show the theoretical lines in which the initial state at a 20 degree was assumed to be a fully dissolved condition. The difference between theory and experiment indicates that initial condition was not fully dissolved condition as we considered. Figure 6 shows the variation of the partial pressure of pore gas with the variation of temperature. It is found from the figure (a) that the increase of pressure of CO2 case is larger than that of air case. It is because the amount from which the gas that has dissolved to water CO2 more than air comes out in the vapor phase is large when the temperature rises. The amount of pressure increase depends on the drainage condition and the degree of saturation as shown in the figure (b) It results in the decrease of effective stress as shown in Figure 7. 3.2 Undrained monotonic shear test
s: suction.
Change in partial pressure of gas (kPa)
(a) Fine sand, Undrained

15 Case14 (CO2, Sr=90%)
(b) Volcanic sand, Air

15 Case10 (Sr=80%, U) Case7 (Sr=40%, U) Case9 (Sr=80%, D) Case6 (Sr=40%, D)
Case13 10 (Air, Sr=90%)
10
Figure 6. Variation of partial pressure of pore gas with the variation of temperature.
Effective mean normal stress (kPa)
(a) Fine sand, Undrained

20
40
(b) Volcanic sand, Air
Case6 (Sr=40%, D) Case7 30 (Sr=40%, U)
10
Case13 (Air, Sr=90%) Case14 (CO2, Sr=90%) 30 Temperature ( ) 40
20
0 20
10 20
Case9 (Sr=80%, D) Case10 (Sr=80%, U) 30 Temperature ( ) 40
Figure 7. Variation of effective stress with the variation of temperature.
Figure 8 shows the relationship between the axial strain and the deviator stress in triaxial compression tests under undrained monotonic loading.
Figure 9 shows the corresponding effective stress paths. It can be seen that stiffness at a small strain level when starting from larger initial effective stress is higher. In addition to this, not having received
564
Fine sand, Sr=90% Deviator stress (kPa) 150

20
20
40 Fine sand Deviator stress (kPa)
20 40 40
100
20
20
50
0 0
Case1 (Air) Case2 (CO2) Case3 (Air) Case4 (CO2)
20 2 0
Case12 (CO2)
5 10 Axial strain (%)
15
Axial strain (%)
Figure 8. Stress and strain relationship during monotonic undrained shear.
Figure 10. Typical example of stress and strain relationship during cyclic shear for case-12.
Fine sand, Sr=90%
Change in pore gas pressure (kPa)
Fine sand, Sr=90% Deviator stress (kPa) 150

20 20 40
20
Initial effective stress 20
100
20 40
10 Case11 (Air) Case12 (CO2) 0 0 4000 8000 12000
50
0 0
Case1 (Air) Case2 (CO2) Case3 (Air) Case4 (CO2)
50 100 Effective mean normal stress (kPa)
Time (s)
Figure 9. Effective stress paths of monotonic undrained shear.
Figure 11. Time histories of the pore gas pressure increase.
the temperature rise demonstrates strong shear strength. And the specimen with CO2 gas demonstrates weak shear strength comparing to that of the specimen with air. If all cases are assumed to have the same soil particle structure before shearing, dilation nature is considered to be the same. Therefore, the cause of effective stress difference between air and CO2 is the degree of the pore gas pressure reduction during the shearing process. 3.3 Undrained cyclic shear test
Figure 10 and 11 show typical examples of the relationship between the axial strain and the deviator stress, and change in pore gas pressure during cyclic shear test, respectively. As shown in the figures, in the constant strain amplitude test, the shear stress demonstrated gradually decreases corresponding to the effective stress reduction. In the
figure 11, we have to pay attention to that pore gas pressure gradually increases and finally reaches the initial effective stress. This stage is regarded as zero effective stress condition that is complete liquefaction state (Unno et al., 2008). It is also interesting in figure 11 that rising process of the pore gas pressure is different in CO2 and air. Undrained cyclic shear strength can be evaluated from a dissipation energy obtained from stressstrain relationship (Kazama et al., 2000). The dissipation energy indicates the ductility nature of the soil against to liquefaction and is defined by the following equation. W (t ) (t ) dt
0 t
(5)
where (t) and (t) are the deviator stress and strain velocity at a time t.
565
12 Volcanic sand, Air Case5 (20 , D)

Dissipation energy (kPa)
Sr=40%
degree before shearing. Conclusions obtained from this study are summarized as follows: 1. On the effective stress state before shearing The effective stress decreased by temperature rise under closed condition because of the pore gas pressure increase. The degree of pore gas pressure rise for CO2 specimen was much larger than that for air specimen. It reflected the difference of solubility of the gas into the pore water from a certain temperature to the temperature. 2. On the undrained monotonic shear strength From the reason above, when subjected to temperature rise under undrained condition, the shear strength of the specimen with CO2 as a pore gas was lower than the specimen with pore air. Furthermore, under the constant temperature condition, the shear strength of the specimen with pore air was larger than that of the specimen with CO2, because the effective stress of the specimen with CO2 did not decrease so much comparing to the specimen with pore air during dilation process in shearing. 3. On the undrained cyclic shearing Undrained cyclic shear strength was evaluated by the dissipation energy obtained from stress and strain relationship during cyclic shearing. The undrained cyclic shear strength of the specimen with CO2 was lower than that of the specimen with pore air because of the same reason described above. From the test results, the influence of temperature rise under undrained on the shear strength is much larger, when the pore gas has higher solubility decrement with the temperature and lower effective stress condition. REFERENCES
Hilf, J.W. Estimeting construction pore pressure in rolled earth dams. In Proceedings of the 2nd International Conference on Soil Mechanics and Foundation Engineering Rotterdam, The Netherlands, Vol. 3, pp. 234240, 1948. Grozic, J.L.H., Imam, S.M., Robertson, R.P.K. and Morgenstern, N.R. Constitutive modeling of gassy sand behaviour Can. Geotech. J., 42, pp. 812829, 2005. Kazama, M. and Unno, T. Earthquake-induced mudflow mechanism from a viewpoint of unsaturated soil dynamics, Experimental Unsaturated Soil Mechanics, 112 Springer proceedings in Physics, T. Schanz (Ed.), pp. 437444, 2007. Kazama, M., Yamaguchi, A. and Yanagisawa, E. Liquefaction resistance from a ductility viewpoint, Soils and Foundations, Vol. 40, No. 6, pp. 47-60, 2000. Unno, T., Kazama, M., Uzuoka, R. and Sento N. Liquefaction of Unsaturated Sand Considering the Pore Air Pressure and Volume Compressibility of Soil Particle Skeleton, Soils and Foundations, Vol. 48, No. 1, pp. 8799, 2008.
Case6 (40 , D)
Case7 (40 , U) 4 Case8 (20 , D) Case9 (40 , D) Sr=80% Case10 (40 , U) 0 0 250 Cumulative shear strain (%) 500
Figure 12. Dissipation energy consumed during cyclic shear process for volcanic sand.
Dissipation energy (kPa) Fine sand, Sr=90%
net =40kPa
6 4 2 0
Case 15 (40 Case 11 (20
, Air) , Air) , CO 2) , CO 2) , Air) , CO 2)
Case 12 (20 Case 16 (40 Case 13 (40 Case 14 (40
250 Cumulative shear strain (%)
500
Figure 13. Dissipation energy consumed during cyclic shear process for fine sand.
It is understood form the equation that the dissipation energy is shear work consumed during cyclic shear. Figure 12 and 13 show the dissipation energy consumed during the test. It can be seen from the figure 12 that the dissipation energy was not influenced by temperature rise under the drained condition. On the contrary to this, the dissipation energy in the case subjected to temperature rise under the undrained condition is less than that in the case under drained condition. If we compare the energy consumed between air and CO2 specimen in figure 13, the dissipation energy in case of CO2 specimens are less than that in case of air specimens under the same test conditions. The reason is the same as the case of monotonic loading test. 4 CONCLUSIONS
The authors have carried out both monotonic and cyclic shear tests of unsaturated sandy soils subjected to the temperature rise from 20 degree to 40
566
Thermal pulse effects on the stiffness degradation of unsaturated clayey materials

A. Lima, J.A. Pineda & E. Romero
Department of Geotechnical Engineering and Geosciences, Universitat Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: This paper describes the results of an experimental programme carried out to evaluate the influence of relatively fast thermal pulses on the stiffness degradation of two Tertiary argillaceous rocks (a plastic clay rock and a stiff claystone). Heating pulses were performed using a specially developed setup installed in a microwave oven. Low-strain shear moduli using bender elements were measured before and after each of the heating pulses to track the evolution of the thermally induced degradation. An important stiffness reduction was observed on both samples. In addition, mercury intrusion porosimetry tests to analyse the pore size distribution changes between unaltered and post-mortem samples were performed to have further insight into the degradation phenomena undergone by the materials. 1 INTRODUCTION technique for studying thermal effects under water undrained conditionsthe heating pulse is faster than the rock capacity to drain the excess pore water pressureand their consequences on rock degradation. Two types of Tertiary rocks are studied, namely a plastic clay rock and a stiffer claystone. However, the application of this technique is not straightforward and an important effort is required to overcome experimental difficulties for simultaneously measuring temperature and pore pressures changes in the sample. The paper presents preliminary heating/ cooling results, in which the pore pressure evolution has not been measured. Nevertheless, information on water pressure changes induced on heating under nearly constant volume and carried out on the same plastic clay rock is presented with the aim of complementing the coupled evidence (temperature and pore pressure). Stiffness degradation is evaluated in terms of the low-strain shear modulus changes undergone by the materials before and after each of the heating pulses using bender elements tests under unconfined conditions. As complementary information into the degradation phenomena undergone, the study also analyses the pore size distribution changes experienced by the materials along the different heating pulses performed. 2 TESTED MATERIALS
Degradation of clayey materials is usually related to weathering processes leading to irreversible changes in the material hydro-mechanical properties. Environmental changes on the exposed surface (wetting-drying cycles), stress changes (loading-unloading processes) and thermal effects (heating-cooling paths) are widely recognised to trigger degradation processes. In the case of thermal effects, heating-cooling processes induce pore water pressure changes, which depend on the rate of temperature increase/decrease, on soil compressibility and thermal-expansion coefficient, on water permeability, porosity, saturation degree, as well as on the hydraulic boundary conditions applied. The unsaturated case is more insidious because thermal effects induce changes in both water and gas pressures. Thermal impact and their effects on previously damaged zones may play an important role on low permeability geological host formations for disposal of High Level Nuclear Waste. In addition, relatively fast processes of temperature increase and heat-generated pore pressures can occur in rapidly deforming shear band during fast sliding landslides inducing frictional heat input (Vardoulakis 2002, Pinyol & Alonso 2010). Most of the experimental research related with thermal effects on clayey materials has been carried out under relatively low temperature rates. Recent examples of these laboratory tests can be found in Muoz et al. (2009) and Lima et al. (2009a, b). Particularly, this study focuses on relatively fast heating pulses under unstressed conditions using microwaves, which comes up as an interesting
Two different Tertiary argillaceous materials were used in the study: a plastic clay rock (Boom clay, Mol, Belgium) and a stiff sulphate-bearing claystone (Lilla claystone from lower Ebro Basin in Spain).
567
Table 1. Property Density
Main properties of tested argillaceous rocks. Boom clay 1.99 to 2.05 Mg/m3 1.65 to 1.71 Mg/m3 21 to 25% 2 to 4 MPa 2.67 Mg/m3 0.560 to 0.618 0.358 to 0.382 91 to 100% 56% 27% 2.7 1012 m/s 0.7 to 1.1 GPa Lilla claystone 2.57 to 2.61 Mg/m3 2.47 to 2.51 Mg/m3 3 to 4% 50 to 80 MPa 2.75 to 2.82 Mg/m3 0.110.13 0.090.11 80 to 85% 23% 5% 6.8 1013 m/s 9 to 12 GPa
0.84 0.72 0.60 Pore size density function, -(enw/(logx)) 0.48 0.36 0.24 0.12 0.00
(a)
Boom clay e = 0.56
Dry density Water content Total suction Density of solids Void ratio Porosity Degree of saturation Liquid limit Plasticity index Vertical water permeability Low-strain shear modulus (bender elements)
(b) 0.04 0.03 0.02 0.01 0.00 1
Lilla claystone e = 0.12
10 100 1000 Entrance pore size, x (nm)
10000
Table 1 summarises the basic characterisation and the main volumetric and gravimetric properties of the kaolinitic-illitic Boom clay, which is slightly overconsolidated (Lima 2010). Lilla claystone is composed mainly of two components: the host argillaceous matrix (illite, paligorskite, domolite and quartz) and the sulphated crystalline fraction (anhydrite and gypsum). Table 1 presents the main properties of this claystone (Pineda 2010). The pore size distribution network of the two materials was studied using mercury intrusion porosimetry (MIP) tests carried out on freeze dried unaltered samples. Figure 1 shows the pore size distribution curves for both materials. For Boom clay a monomodal distribution with dominant pore size around 90 nm is observed. In the case of Lilla claystone, it also tends to a monomodal distribution although the dominant pore mode was not detected (the dominant pore mode is lower than 7 nm, the minimum detectable pore size with MIP). 3 THERMAL PULSE TESTS
Figure 1. Pore size distributions for unaltered a) Boom clay and b) Lilla claystone.
Thermal effects have been studied by applying fast heating pulses using microwaves and by measuring the evolution of temperature inside the sample during heating and subsequent thermal dissipation. This process is carried out under constant water content and under unstressed conditions. Experimental difficulties arise when steel objects including miniature probes are irradiated with microwaves.
Only temperature changes could be determined in this preliminary stage of the research. Figure 2 shows the experimental setup of the thermal pulse test. The objective is to apply a thermal pulse of specific characteristics (power and time) using a modified commercial microwave oven (maximum power 1400 W). A continuous record of temperature is made using a type K thermocouple housed in a carbon fibre tube. The thermocouple is installed through the oven wall and inserted into the middle of the sample (10 mm in depth). A digital thermometer with data acquisition is used for recording the temperature measurements. In order to avoid moisture losses during the tests, each sample is previously vacuum sealed using a polyethylene bag with a vapour water transferability of 1.8 g/m2/24 h at 23C and 85% of relative humidity. The experimental protocol consists of different stages. In first a step, the preparation of the polyethylene bag and the insertion of the thermocouple is performed. A small hole of 1 mm diameter is drilled through the bag to pass the thermocouple, which is carefully sealed using epoxy glue. In a second step, the sample is installed inside the bag and the thermocouple inserted 10 mm in depth through a hole previously drilled into the sample using a mini drilling machine. Afterwards, vacuum is applied and the polyethylene bag thermally sealed.
568
Table 2.
Characteristics of heating pulses applied. Pulse 1 Pulse 2 P (W) 1400 1400 t (s) 28 20 Pulse 3 P T (W) (s) 140 28 140 20 Pulse 4 P (W) t (s)
P Material (W) BC LC
t (s)
1400 10 1400 20
1400 20
2 1. digital thermometer with data acquisition 2. microwave oven 3. sample under vacuum in polyethylene bag 4. thermocouple protected by carbon fibre tube
Figure 2. Scheme of the experimental setup and vacuum-sealed sample placed inside the microwave oven.
The sealed sample is placed into the microwave oven and the thermocouple passed through the oven wall, in which a 1-mm hole has been drilled. Interferences inside the microwave oven are avoided by housing the thermocouple with carbon fibre tube. In a final stage, thermal pulses are applied. Each test is finished after thermal dissipation to room temperature (22C). Afterwards, samples are vacuum-sealed again and stored during one day for humidity equalisation before determining shear stiffness and continuing with the next pulse. Table 2 shows the characteristics of the heating pulses (power and duration) applied on each material (BC Boom clay and LC Lilla claystone, respectively). Samples with dimensions of 38 mm in diameter and 76-mm high were carefully trimmed to avoid alterations. The initial state of Boom clay sample was drier (w 16.5%) then the natural one, which corresponded to a total suction 8 MPa. The initial dry density was d 1.70 Mg/m3 (void ratio around 0.57), which was associated with a degree of saturation Sr 77%. The initial state of Lilla sample was w 3.7%, 60 MPa, d 2.49 Mg/m3, e 0.12 and Sr 85% (similar to Boom clay). The maximum difference in water content measured after each pulse was 0.3% for Boom clay and 0.01% for Lilla claystone. After final dismantling, a maximum volume change of 0.1% was measured for Boom clay and lower than 0.05% for Lilla specimen. During heating, pore water pressure increases due to its larger thermal expansion coefficient. The change of pore water pressure on thermal loading with no change in boundary confining stress can be analysed assuming volumetric compatibility
between soil matrix and their constituents (air, liquid and solid) using compressibility and thermal expansion coefficients (refer for example to Agar et al. 1986, Vaziri & Byrne 1990). Important material and state parameters linked to this fast temperature increase and pressure build-up are the soil and constituent compressibility values, the soil and constituent thermal-expansion coefficients, the porosity and the degree of saturation. It is also largely dependent on the rate and range of temperature change, the stress state, the water permeability, the hydraulic boundary conditions applied and the induced damage on pore pressure increase. Under no change in boundary confining stress, Romero (1999) estimated a maximum quasi-undrained pressurisation coefficient of 0.005 MPa/C for compacted Boom clay at Sr 94%. Vardoulakis (2002) estimated a value of 0.06 MPa/C for saturated Boom clay. A value of 0.189 MPa/C has been recorded by Lima et al. (2009a) under saturated and nearly isochoric conditions (constant volume heating cell). Particularly, Figure 3 presents the evolution of water pressure and temperature at approximately the same point in this heating cell. As observed, the variability in the pressure response makes necessary the use of thermo-hydro-mechanical analyses using numerical tools to compensate for the lack of information of the experimental setup.
1.8 38
Pore water pressure (MPa)
1.6
34 Temperature (C)
1.4
30
1.2 heating phase Pw1 bottom 25 mm T2 bottom 25mm 95 112.5 130 147.5 165 Time (min) 182.5 200
26
22
0.8
18
Figure 3. Time evolution of water pressure and temperature during a heating path under nearly isochoric conditions on saturated Boom clay (Lima et al. 2009a).
569
STIFFNESS EVALUATION
Temperature, T (C)
350 300 250 200 150 100 50 0 300
(a)
Boom clay 1st pulse 2nd pulse 3rd pulse
In order to evaluate thermal effects on the stiffness response of Boom clay and Lilla claystone, the low-strain shear modulus G0 under unconfined conditions was determined by bender elements. In this technique, two polarised piezoceramic transducers (one transmitter and one receiver) are used to transmit and capture a dynamic signal which travels through the soil sample. The time delay (or travel time) between emitted and received signals is used to determine the shear wave velocity (Vs) where the travel length is commonly taken as the tip-to-tip distance between piezoceramics. Thus, the low-strain shear modulus (G0) can be obtained from the total density (t) and the shear wave velocity (G0 tVs2). A sine pulse with a high input apparent frequency (50 kHz) was used in to minimise near-field effects. Amplitude of input wave was equal to 20 Vpp. The experimental procedure followed Pineda et al. (2008a, b). This technique was corroborated with redundant measurements at constant water content using resonant column (shear strains 103 %) at two different confining stresses of 200 kPa and 800 kPa after pulse 1 on Boom clay, as well as by using unoading/reloading stiffness data in unconfined compression tests performed on Lilla samples under unaltered and final (after the different heating pulses) conditions. Resonant column tests were not performed on Lilla samples due to their high stiffness. 5 EXPERIMENTAL RESULTS
(b)
Lilla claystone 1st pulse 2nd pulse 3rd pulse 4th pulse
Temperature, T (C)
250 200 150 100 50 0 1 10

Short-circuit
Time (s)
100
1000
10000
Figure 4. Time evolutions of temperature along the different pulses performed.

Lilla claystone (w=3.7%) BE 1 RC 800 kPa 1 RC 200 kPa 1 1
Initial states
Figure 4 shows the time evolution of temperature during the application of the different pulses (three pulses on Boom clay and four on Lilla claystone). Boom clay consistently underwent larger temperature peaks at increasing power input and durations. On the contrary, Lilla claystone displayed the maximum temperature peak along the first pulse (320C) with a progressive decrease of the peak intensity with subsequent pulses (pulses 2 and 3 showed peak temperatures around 135C, while for the fourth pulse the peak value was not obtained due a short-circuit problem). In this last case, it appeared that water evaporation losses undergone by this stiff and low porosity material on first heating were not uniformly re-absorb during the subsequent cooling processwater was mainly absorbed at the surface, despite the equalisation stage carried out. Under a slightly drier condition in the inner part of the specimen, microwaves were not efficient in increasing temperature. Water evaporation was clearly observed in both materials since the polyethylene bag inflated during the temperature increase path. In the case
G/G0
0.8
2 3 4 2
0.6 Boom clay (w=16.5%) BE RC (800 kPa) RC (200 kPa) 0.4 0 100 200 300
Peak Temperature, Tpeak (C)
400
Figure 5. Variation of low-strain shear moduli with peak temperature for the different pulses applied.
of Boom clay, despite undergoing a similar process, the larger water storage capacity prevented this phenomenon from being detected. Figure 5 shows the variation of shear moduli with peak temperature for the different pulses
570
60 55 Unconfined compression strength, qu (MPa) 50 45 40 35 30 25 20 15 10 5 0 0 0.01 0.02 Axial strain, a (-) 0.03 0.04 Lilla ST-2 (Unaltered) Lilla S-10 (Thermally degraded) Eu-r(S-10) = 3172 MPa
0.84 Pore size density function, - enw/(logx)

Eu (ST-2) = 4441 MPa
Unaltered Boom clay After thermal pulses
0.72 0.60 0.48 0.36 0.24 0.12 0.00 1 10 100
1000
10000
Entrance pore size, x (nm)
Figure 6. Unconfined compression tests on thermally degraded and unaltered Lilla claystone samples.
Figure 7. Pore size distribution tests on Boom clay sample under unaltered condition and after undergoing the different heating pulses.
0.09
Pore size density function, -(enw/(logx))
performed on both materials. Stiffness variation is plotted in terms of the stiffness ratio (G/Go) determined after each of the heating pulses. A clear monotonic decrease is detected, which sustains the degradation undergone by both materials. Resonant column tests with confining pressure performed on Boom clay after the first cycle displayed a slightly larger value compared to bender element readings. In the case of Lilla claystone, the decrease in stiffness was not so large compared to Boom clay measurements. Probably, the lower storage capacity and porosity of the claystone, as well as its difficulty in uniformly re-absorbing water, is affecting this degradation. Unconfined compression tests on thermal degraded (after four pulses) and unaltered samples were performed on Lilla claystone. Figure 6 shows the stress-strain curves obtained, in which an unloading/reloading cycle relatively far from failure was performed to determine the elastic Young moduli for both conditions (Eu-r). The results are indicated in the figure. A stiffness reduction of 29%, in agreement with bender element results, is detected. 6 MICROSTRUCTURAL STUDY OF INDUCED DEGRADATION
0.08 0.07 0.06 0.05 0.04 0.03 0.02 0.01 0 1 10
Lilla claystone Unaltered sample After thermal pulses
100
1000
10000
100000
Figure 8. Pore size distribution tests on Lilla claystone under unaltered condition and after undergoing the different heating pulses.
unaltered state. A clear shifting of the dominant pore size towards a larger size is detected in Boom clay sample after heating. Results presented in Figure 8 for Lilla claystone show a similar trend towards slightly larger dominant micropores and the appearance of some macroporosity after the heating pulses. 7 SUMMARY AND CONCLUSIONS
A final microstructural study evaluating the pore size distribution changes using mercury intrusion porosimetry between the unaltered state and after undergoing the different thermal pulses, was also performed on both materials to get further insight into the degradation phenomena. Figure 7 shows the pore size distribution curve of the thermally degraded Boom clay compared to its initial
The effects of very fast heating pulses under unstressed conditions on the stiffness response of two unsaturated clayey materials have been presented. The tested materials were plastic Boom clay and a stiff Lilla claystone under relatively high degrees of saturation (around 80%). A new
571
experimental setup was developed to induce relatively fast heating pulses using microwaves and by monitoring temperature changes with a thermocouple inserted into the specimen. The samples were perfectly sealed to avoid water content changes during the heating and cooling processes. Nevertheless, some non-uniformity aspects regarding water content distribution inside Lilla claystone sample, affected the efficiency of the microwaves to increase temperature. No evolution of the pore pressure was done due to the difficulty in installing pressure transducers inside the microwave oven. Nevertheless, complementary information on water pressure changes induced on heating under nearly constant volume and carried out on the same saturated plastic clay was also given. Despite the limitations of the technique, the setup comes up as an interesting way for studying thermal effects under water undrained conditions the heating pulse is faster than the rock capacity to drain the excess pore water pressureand their consequences on rock degradation. A dynamic technique (bender elements) was used to monitor stiffness degradation and low-strain shear moduli changes undergone by the materials before and after each of the heating pulses. The experimental results clearly showed the stiffness degradation on increasing heating pulses for both materials. Redundant measurements of stiffness using resonant column and unconfined compression tests corroborated its reduction with thermal cycling. As complementary information into the degradation phenomena undergone, the study also included the pore size distribution changes experienced by the materials after the different heating pulses performed. A clear shifting of the dominant micropore size towards larger values was detected in both damaged samples after the heating cycles. In the case of Lilla claystone some emergence of macroporosity was also observed after the heating pulses. ACKNOWLEDGEMENTS The authors acknowledge the financial support provided by EIG-EURIDICE/SCK.CEN through a PhD collaboration with UPC and ALBAN PROGRAMME grants from the European Commission, EU Programme of High Level Scholarships for Latin America, id number E04D027285CO. REFERENCES
Agar, J.G., Morgenstern, N.R. & Scott, J.D. (1986). Thermal expansion and pore pressure generation in oil sands. Can. Geotech. J., 23: 327333.
Lima, A. (2010). Thermo-hydro-mechanical behaviour of natural Boom clay. PhD Thesis, Universitat Politcnica de Catalunya, Barcelona. Lima, A., Romero, E., Gens, A., Muoz, J. & Li, X.L. (2009a). Heating pulse tests under constant volume on natural Boom clay. Proc. Int. Symposium on Unsaturated Soil Mechanics and Deep Geological Nuclear Waste DisposalUNSAT-WASTE 2009, August 2428, 2009, Shanghai, China. Y.-J. Cui, W.-M. Ye, B. Chen, Y.-G. Chen (eds.). Book of proceedings (Tongji University): 136141. This paper has been accepted for publication in the Journal of Chinese Rock Mechanics and Engineering. Lima, A., Romero, E., Vaunat, J., Gens, A. & Li, X.L. (2009b). Heating pulse tests under constant volume on natural boom clay. Experimental results and numerical simulations. In: Impact of Thermo-HydroMechanical-Chemical (THMC) processes on the safety of underground radioactive waste repositories. An international conference and workshop in the framework of the European Commission TIMODAZ and THERESA projects, 29th September1st October 2009, Luxembourg. The final proceedings will be published by the European Commission. Muoz, J.J., Alonso, E.E. & Lloret, A. (2009). Thermohydraulic characterization of soft rock by means of heating pulse tests. Geotechnique 59(4): 293306. Pineda, J.A. (2010). Degradation of argillaceous rocks: an experimental study. PhD Thesis, Universitat Politcnica de Catalunya, Barcelona. Pineda, J.A., Lima, A. & Romero, E. (2008a). Influence of hydraulic paths on the low-strain shear modulus of a stiff clay. Proc. First European Conference on Unsaturated Soils: Advances in Geo-Engineering, July 24, 2008, Durham, UK. D. Toll, C. Augarde, D. Gallipoli & S. Wheeler (eds.). Taylor & Francis Group, London: 519523. Pineda, J.A., Arroyo, M., Romero, E. & Alonso, E.E. (2008b). Dynamic tracking of hydraulically induced degradation of a shale. Proc. 4th Int. Symp. Def. Characteristics of Geomaterials, 2224 September, 2008, Atlanta, USA. S.E. Burns, P.W. Mayne & J.C. Santamarina (eds.). IOS Press, Amsterdam: 809816. Pinyol, N.M. & Alonso, E.E. (2010). Fast planar slides. A closed-form thermo-hydro-mechanical solution. Int. J. Numer. Anal. Meth. Geomech. 34(1): 2752. Romero, E. (1999). Characterisation and thermo-hydromechanical behaviour of unsaturated Boom clay: an experimental study. PhD Thesis, Universitat Politcnica de Catalunya, Barcelona. Vardoulakis, I. (2002). Dynamic thermo-poro-mechanical analysis of catastrophic landslides. Gotechnique 52(3): 157171. Vaziri, H.H. & Byrne, P.M. (1990). Numerical analysis of soil sand under nonisothermal conditions. Can. Geotech. J., 27: 802812.
572
Effects of a microbiological compound for the stabilisation of compacted soils on their microstructure and hydro-mechanical behaviour
L. Morales & E. Garzn E. Romero C. Jommi
Universidad de Almera, Almera, Spain Universitat Politcnica de Catalunya, Barcelona, Spain Politecnico di Milano, Milano, Italy
ABSTRACT: The contribution focuses on the effects of a new microbiological treatment on the geotechnical properties of soils adopted for linear earth construction in the Iberian Peninsula. To this purpose, at the Universidad de Almeria, different soils are being treated with a product of microbiological base, to improve the mechanical properties of the soil, relying on the capacity of some microorganisms to precipitate substances with stabilising characteristics, in particular calcium carbonate. Different aspects related to the microbiological stabilisation are analysed, in view of the performance of the soil as unsaturated construction material. Changes in the soil microstructure are evaluated first, via the results of mercury intrusion porosimetry. The consequent influence on the retention behaviour of the soil, on its shear strength and on its volumetric stiffness is then discussed. 1 INTRODUCTION The aforementioned reasons stimulated the adoption of microorganisms as potential tool in soil stabilisation, among the different applications proposed in the last decade (Weiner & Addadi 1997, Tiano et al. 1999, Gray 2001, Bang et al. 2001). In this framework, at the Universidad de Almeria, an experimental study was initiated, with the aim of verifying the stabilising effects of a microbiological treatment on different Spanish soils already adopted as construction materials. In this contribution, a comprehensive overview is given of the effects of the microbiological treatment on a silty soil, starting from microstructure observations and passing to the hydro-mechanical aspects of relevance for earth construction. 2 MICROBIOLOGICAL TREATMENT
Linear earthworks in Spain have acquired increasing importance in the last years. The economical investment of the Ministerio de Fomento (the Spanish Ministry of Public Works) on routes and urban roads has increased, from 2.120 million Euros in 2003, to 4.315 million Euros in 2006, representing about 41% of the total investment in civil engineering works, after railways infrastructures (MFOM 2006). Earthworks construction constitutes a political priority and a social demand. At the same time, increasing attention to environmental problems and the Spanish commitments for reduction of CO2 emission are promoting the research for alternative construction materials and procedures, to reduce energetic costs and environmental impact to a minimum. The problem is not easy to tackle. In fact cement stabilisation of earth constructions, which was widely adopted in Andalusia, has high energetic cost, also in terms of CO2 emissions. Besides, there is a concern about the risks for workers and environment regarding chemical compounds (ANCADE Manufacturers National Association of Limes and Derivatives of Spain, Hkkinen & Vares 1998).
Soil improvement via micro-biological treatments relies on the capacity of some microorganisms to precipitate substances with stabilising characteristics, in particular calcium carbonate. Boquet et al. (1973) suggest that almost all bacteria are capable of precipitating CaCO3, yet the precise role of microbes in carbonate precipitation process is not totally clear.
573
Considerable research on carbonate precipitation by bacteria has been performed using ureolytic bacteria (Dick et al. 2006, Bang et al. 2001, Stocks-Fischer et al. 1999), which are able to influence precipitation of calcium carbonate by the production of a urease enzyme. This enzyme catalyses the hydrolysis of urea (CO(NH2)2) to CO32 and ammonium, resulting in an increase of pH and carbonate concentration in the bacterial environment (Stocks-Fischer et al. 1999):
2 CO(NH2)2 2H2O 2 NH4 CO3
(1)
In the presence of calcium ions, the solution can become supersaturated with respect to calcium carbonate leading to its precipitation: CO32 Ca2 CaCO3 (2)
Precipitation of calcium carbonate crystals occurs by heterogeneous nucleation on the bacterial cell wall once supersaturation is achieved. Contrarily to other methodologies (for an extensive overview, see Castanier et al. 1999), production of carbonate by hydrolysis of urea can be easily controlled, and it allows for the production of high concentrations of carbonate within a short amount of time. In contrast with the research directly related to microbial aspects, minor attention has been given to the influence of the chemical parameters, i.e. calcium ions concentration and nutrients concentration on the efficiency of the bio-deposition treatment. Few studies indicate that the concentrations of these components influence the amount and type of precipitates that are formed (Jimenez-Lopez et al. 2008, De Muynck et al. 2009). The latter aspect is investigated at the Universidad of Almera, by changing the amount of bacteria introduced in soils with different initial natural calcite content.
Iberian Peninsula. The grain size distributions of the natural (B-5) and of the treated (BT-5 and BT-5_2D) materials are presented in Figure 1. The comparison shows that the microbiological treatment tends to aggregate the soil, so that eventually an apparent coarser distribution results. If the bacteria dose is doubled, slightly finer aggregates result. The most relevant physical and chemical parameters to complete the characterisation of the natural and of the treated soils are summarised in Table 1. The data show that the effects of bacteria are increasing calcite content of the soil, and reducing the soil specific surface and its plasticity. The reason for the latter result could be the addition of a non-plastic component to the natural soil, or the result of the more aggregated structure promoted during the treatment. Standard Proctor compaction performed on the natural soil gave an optimum dry density of D 1.85 Mg/m3, which can be achieved with a water content of wopt 0.15. The natural soil samples were prepared by Proctor compaction, on the dry side of the optimum, at a water content between wopt and (wopt 2%), to be representative of the soil compacted in the field, and providing, at the same time, an initial structure potentially prone to collapse. Different choices could be pursued to compare the performance of the treated soil to that of the
ASTM: #4#10#16#40#100#200
100
B-5
80 % passing 60 40 20
BT-5 BT-5_2D
SOIL, MICROBIOLOGICAL TREATMENT AND SAMPLES PREPARATION
10
0 .1
0 .0 1
particle size (mm)
A microorganism of the Bacillaceae family is used in the precipitation of carbonates minerals in the soil investigated. The microorganisms are added to the natural samples together with compaction water content, and the soil samples are left ageing for seven days at least in a humid chamber, after which the treatment is stopped by increasing the temperature. The results of treatments performed with a bacteria reference dose (T) and double dose (T_2D) will be compared. The natural soil used in the tests presented is a silty-clayey sand from the south-east of the
Figure 1. Grain size distribution of the natural soil B-5 and of the treated soil with single (BT-5) and double (BT-5_2D) dose. Table 1. Physical characteristics of the samples. Ss(m2/g) wL PI CaCO3 g (%) (%) (%) (Mg/m3) BET Langmuir 2.73 2.74 2.68 33.1 45.9 31.8 43.7
Sample
B-5 48. 30. 4.5 BT-5 49. 29. 6.4 BT-5_2D 33. 10. 9.0
574
natural soil. In the tests presented in the following, the samples were prepared by Proctor compaction at the same water content and with the same energy spent for the natural soil, following a possible field practice. This results in samples having, in general, higher initial void ratio than that of the natural soil, due to the higher stiffness of the treated soil. Some of the treated samples, after Proctor compaction, were statically re-compacted to analyse the influence of the initial void ratio on their behaviour (BT-5_R). 4 MICROSTRUCTURE INVESTIGATION
1000
MIP B-5 MIP BT-5
100
MIP BT-5_R WP4 B-5 WP4 BT-5_R
10 suction (MPa)
0.1 equivalent diameter 3 m 30 m 0.01
The microstructure features of the natural and the treated soil were studied by means of Scanning Electron Microscopy and Mercury Intrusion Porosimetry (MIP). The pore size distributions of relevant samples, derived from MIP data, are reported in Figure 2. The comparison between the data for natural soil and those for the treated soil shows that the appreciable effect of bacteria is to reduce the porosity in the range 3 m 30 m. Besides, compaction at the same energy is less effective in reducing the pores of larger dimensions. The peak of the largest pores for the treated soil is around 200 m, while that of the natural material, compacted with equivalent energy is around 50 m. Re-compaction of the treated soil, i.e. increasing compaction energy, may reduce the larger pores, while the small ones remain unaltered. MIP data were elaborated to provide an estimation for the retention curve, and the elaborations are compared with direct measurements obtained by a dew point probe (WP4) on compacted samples. For suctions equal or higher than 10 MPa, the two data sets tend to coincide, well describing the intra-aggregate porosity. For lower suctions,
0.6
MIP B-5 MIP BT-5
0.001
0.2
0.4
0 .6
0.8
degreeof saturation, Sr (---)
Figure 3. Water retention data from WP4 for natural and treated soil and retention curve estimations from MIP data.
points elaborated from MIP describe the pore size distribution of the sample frozen at its compaction water content, while the WP4 data come from samples which may swell interacting with water. The overall comparison suggests that the water retention domain of the natural and the treated soils should differ one from the other mostly in the suction range 0.01 0.1 MPa, with the treated soil showing higher retention capacity in this suction range. The shape of the actual retention domains will be better described by the WP4 data. 5 MECHANICAL BEHAVIOUR
3 m 30 m
pore size density function (---)
MIP BT-5_R
0.4
Direct shear tests and oedometer tests were performed on natural and treated samples to analyse the effects of the treatment on the mechanical behaviour of the soil. The tests were performed at the as-compacted and at the saturation water contents, simulating possible saturation in the field due to water infiltration. 5.1 Shear strength
0.2
0 .00 1
0.01
equivalent entrance diameter, D (m)
0.1
10
10 0
10 0 0
Figure 2. Pore Size Distributions of the natural and treated samples, from MIP data.
The data of direct shear tests performed on saturated samples are compared in Figure 4. The comparison shows that the treated soil is more contractant than the natural one, due to its higher initial void ratio. Nonetheless, the shear strength which can be mobilised is slightly higher at the stress levels investigated. The result is confirmed by the shear strength envelopes, depicted in Figure 5, which give a friction angle of about 38 for the natural soil, and of 40 for the treated one.
575
1 60
(kP a)
B -5
1 20 80 40 0 0 -0.2 -0.4 -0.6 -0.8 0
B T -5
150 kP a
vertica l disp., v (m m ) sh ear stress,
100 kP a 50 kP a
150 kP a 150 kPa 50 kPa 100 kP a 100 kPa 50 kPa
0.4
0.8
1 .2
1.6
relative ho rizontal disp la ce m ent,
h (m m )
Figure 4. Direct shear data for the natural sample B-5 and the stabilised sample BT-5: (a) mobilised shear strength and (b) vertical displacement.
150
' = 37.9 ' = 40.1

shear stress, (kPa) 100
B-5 BT-5
50
B-5 BT-5
50
100
15 0
20 0
effective normal stress, 'n (kPa)
Figure 5. Shear strength envelopes for the natural (B-5) and the treated (BT-5) samples.
with the reference bacteria dose, and on the soil treated with a double dose of bacteria. Oedometer tests were run also on the soil re-compacted statically, after treatment and first Proctor compaction, in order to reduce the void ratio. All tests were conducted following the same protocol. After mounting the samples in the oedometer cell, the vertical load was increased in steps, until a vertical net stress of v 50 kPa or v 100 kPa. Once reached the desired stress level, the samples were flooded at constant vertical stress for 24 hours. Afterwards, the samples were loaded in steps of 24 hours each, up to about 1 MPa or 2 MPa, depending on the test. Finally the samples were unloaded, again in steps of 24 hours each. The data of the tests in which the samples were flooded at v 50 kPa are shown in Figure 6. The initial void ratios of the different samples are slightly different, due to heterogeneity induced by slightly different dynamic compaction procedures. All samples show volumetric collapse upon wetting, but the amount of collapse does not seem to be influenced by the treatment, neither by the dose of bacteria used to stabilise the soil. The data suggest that the initial void ratio at the time of wetting is the dominant parameter on the amount of collapse. The treatment apparently stiffens the soil, as the comparison of the compression curves after collapse suggests. The compression curves for the samples flooded at v 100 kPa are shown in Figure 7. In this case the sample showing the most important collapse is the natural one. Nevertheless, also in this case it is the sample having the highest void ratio upon wetting, due to poor compaction operations. The two samples, which were re-compacted statically to a much lower void ratio, show very little collapse upon wetting, consistently with the other data. The amount of collapse experienced upon flooding was normalised with the net stress at the time of wetting. The data are reported in Figure 8 as a
0.7
B-5 BT-5
It is interesting to note that calcite production from bacteria does not give an apparent cohesion to the soil. Therefore it should not be interpreted as a cementation effect. Instead, it increases the friction angle, in spite of higher void ratio. This effect is more consistent with the idea that calcite produced by bacteria fills an intermediate pore class, which is compatible with the dimensions of the bacteria. The effect on the shear strength is somehow equivalent to that produced by an increase in the relative density of a granular soil. 5.2 Volumetric stiffness
0.6
BT-5_2D
void ratio, e (---)
0.5
0.4
0.3
10
100
1000
10000
To study the stabilisation effect of bacteria on the volumetric stiffness of the soil, oedometer tests were run on the natural soil, on the soil treated
vertical net stress, v (kPa)
Figure 6. Compression curves for the natural and the treated samples, flooded at v 50 kPa.
576
0.8
B-5 BT-5_R BT-5_2D
0.008
slope of the unloading line, (---)
0.7
void ratio, e (---)
0.6
0.006
0.5
0.004
B-5 BT-5 BT-5_R BT-5_2D
0.4
0.3
10
100
1000
10000
0.002 0 .3 2 0 .3 6 0 .4 0 .4 4 0 .4 8 0 .5 2
final void ratio, ef (---)
Figure 7. Compression curves for the natural and the treated samples, flooded at v 100 kPa.
10
-3
Figure 10. Elastic compression index, , of the natural and treated soil as a function of the void ratio at the end of the tests.
collapse strain / stress level, wv (kPa-1)
10
-4
10
-5
B-5 BT-5 BT-5_R BT-5_2D
10
-6
0 .4
0 .5
0 .6
0 .7
0 .8
void ratio upon wetting, e0 (---)
Figure 8. Collapse strain upon wetting, normalised with the vertical net stress at which the samples were flooded.
0.024
compression index, (---)
0.020
0.016
0.012
B-5 BT-5
0.008
BT-5_R BT-5_2D
0.004
0 .3 2
0 .3 6
0 .4
0 .4 4
0 .4 8
0 .5 2
final void ratio, ef (---)
Figure 9. Elastic plastic compression index, , of the natural and treated soil as a function of the void ratio at the end of the tests.
In fact, the evidence is consistent with the observation that the primary result of treating the soil with bacteria is to reduce the pores in a medium dimension range, of the order of microns to tens of microns as a maximum. The treatment does not affect the largest pores, which are mainly a result of the effectiveness of the compaction procedure. As the largest pores are held responsible of most of the volumetric collapse upon wetting, the higher the initial void ratio, the higher is the amount of collapse. On the contrary, the stabilising procedure seems to influence the elastic and the elasticplastic stiffnesses of the soil. The natural soil has a compression index around 0.02. The slope of its unloading lines is in the range 0.006 0.007. Treating the soil with bacteria reduces the compression index to 0.018 and the elastic compliance to values around 0.005. This reduction is consistent again with the idea that the main effect of bacteria is to fill a class of pores of a certain characteristic dimensions. That pore class will give a smaller contribution to the volumetric deformation of the soil. The result is that the slope of the virgin loading line, as well as that of the unloading lines, will decrease. Re-compacting the soil to reach smaller void ratios, decreases further the slope of the virgin compression line to values around 0.010 0.0012, while it does not have any significant effect on the slope of the unloading lines. 6 CONCLUSIONS
function of the void ratio at the time of flooding. The figure confirms that the amount of collapse experienced by the natural and the treated soils is ruled essentially by the initial void ratio.
A wide investigation is being performed at the Universidad de Almera to analyse the possibility of stabilising soils used in earth constructions by means of a microbiological treatment. Different types of
577
soils are being treated and tested afterwards, to verify the effectiveness of the treatment in terms of hydro-mechanical characteristics. The results of different tests on one of these soils were presented here, including MIP data, direct shear data and oedometer data. The data presented display a consistent behaviour pattern, suggesting that precipitation of calcium carbonate from bacteria takes place in the pores of the soil which are slightly bigger than the characteristic dimension of the bacteria themselves, which is around 1 2 m. As a result, the pore size distribution of the soil changes in the range 3 m 30 m, where the pore mode tends to disappear. The change in the pore density function is reflected in the mechanical behaviour of the treated soil, which presents typical features of a more dense soil with respect to the natural untreated one. The friction angle of the treated soil is slightly higher, and its compressibility is consistently lower, than that of the natural soil. As the bacteria do not seem to produce any cementation effect on the soil skeleton, collapse upon wetting does not seem to be significantly affected by the treatment. On the contrary, comparison of collapse data shows that occurrence and amount of collapse are ruled by the as-compacted dry density. As a whole, the tests performed seem to suggest that the microbiological technique may be effective in improving the mechanical characteristics of the compacted soil, provided more energy were spent in compacting the stabilised soil, so as to give a high initial dry density. From this viewpoint, it seems that higher compaction effort is even more effective than increasing the amount of bacteria introduced to stabilise the soil. The ongoing experimental investigation aims at verifying the preliminary conclusions drawn from the tests presented here on other natural soils, which are equally being adopted in earth constructions. ACKNOWLEDGEMENTS This work was done in the framework of the BIOLIN project financed by ACCIONA Infraestructuras. The results presented herein are those of the authors only and do not, necessarily, reflect the views of the company.
REFERENCES
Ancade, Asociacin Nacional de Fabricantes de Cales y Derivados de Espaa: www.ancade.es Bang, S.S., Galinat, J.K. & Ramakrishnan, V. 2001. Calcite precipitation induced by polyurethaneimmobilized Bacillus pasteurii. Enzyme and Microbial Technology 28 (45): 404409. Boquet, E., Boronat, A. & Ramos-Cormenzana, A. 1973. Production of calcite (calcium carbonate) crystals by soil bacteria is a general phenomenon. Nature 246: 527529. Castanier, S., Le Mtayer-Levrel, G. & Perthuisot, J.P. 1999. Ca-carbonates precipitation and limestone genesisthe microbiologist point of view. Sediment. Geol. 126: 923. De Muynck, W., Verbeken, K., De Belie, N. & Verstraete, W. 2009. Influence of urea and calcium dosage on the effectiveness of bacterially induced carbonate precipitation on limestone. Ecol. Eng., doi:10.1016/ j.ecoleng.2009.03.025. Dick, J., De Windt, W., De Graef, B., Saveyn H., Van der Meeren, P., De Belie, N. & Verstraete, W. 2006. Biodeposition of a calcium carbonate layer on degraded limestone by Bacillus species. Biodegradation 17 (4): 357367. Gray, D. 2001. Microbial Stabilization. University of Michigan, USA. Hkkinen, T. & Vares, S. 1998. Environmental Burdens of concrete and concrete products. Nordic Concrete Research. Publication no. 21 1/1998. Oslo. Norsk Betongforening: 130143. Jimnez-Lpez, C., Jroundi, F., Pascolini, C., RodriguezNavarro, C., Piar-Larrubia, G., Rodriguez-Gallego, M. & Gonzlez-Muoz, M.T. 2008. Consolidation of quarry calcarenite by calcium carbonate precipitation induced by bacteria activated among the microbiota inhabiting the stone. International Biodeterioration & Biodegradation 62: 352363. MFOM 2006. Ministerio de Fomento: www.fomento.es. Stocks-Fischer, S., Galinat, J.K. & Bang, S.S. 1999. Microbiological precipitation of CaCO3. Soil Biology and Biochemistry 31 (11): 15631571. Tiano, P., Biagiotti, L. & Mastromei, G. 1999. Bacterial bio-mediated calcite precipitation for monumental stones conservation: methods of evaluation. Journal of Microbiological Methods 36: 139145. Weiner, S. & Addadi, L. 1997. Design strategies in mineralized biological materials. Journal of Materials Chemistry 7 (5): 689702.
578
Influence analysis of the contaminants fluid in a compacted soil collapse

E.Q. Motta
Universidade Federal de Pernambuco-UFPE, Recife, Brasil
S.R.M. Ferreira
UFPE, Universidade Catlica de Pernambuco-UNICAP, Recife, Brasil
ABSTRACT: The collapse behavior of a compacted soil from Petrolndia-PE in Brazil due to inundation with different percolating liquids is analyzed through lab tests. The samples were statically compacted with a dry unit weight of field and natural moisture. The collapse was induced through inundation of samples with non-processed sewage; solutions prepared with sewage substances; leachate; distilled water, which served as a parameter for comparison among the liquids used. A Principal Component Analysis PCA of different parameters in a collapse of this soil is also present. It has been concluded that the soil collapse depends on the physical-chemical interaction between the soil and the flood liquid and is the result of several factors. The PCA analysis shows the influential parameters quickly and efficiently. 1 INTRODUCTION of a soil is governed by a set of properties including absorption of sodium, exchangeable sodium percentage, pH, soil type and salts dissolved concentration in water (English & Aitchison, 1969). The speed with which the cementing ties lose their strength depends on the fluid that percolates into the soil and the solubility of the cement (Reginatto & Ferrero, 1973, Carvalho et al., 1987). The new Petrolndia County, located in the semiarid region of Pernambuco State, relocated at the time of filling the reservoir of the Itaparica Dam, presented in many of its buildings, serious problems with fissures and cracks, causing in some cases, its complete demolition due to the soil collapse. Aiming to contribute in these problems solution, many researches have been working on the soil problems of this county. The studied soil is a Reddish-Yellow Sand located in a Petrolndia County, in the wilderness of Pernambuco, 520 km away from Recife. In Figure 1 is indicated the location used as the experimental field by Ferreira (1995), where the samples were collected for the current research. Its location has the coordinates: N - 9009250, E - 582,625, 310 m of altitude and plan relief. 2 2.1 MATERIALS AND METHODS Methodology of soil analysis
The collapse of natural porous soil has been the cause of several engineering problems in many places of the world, interfering in the design solutions or affecting works. The collapse has caused extensive damage in engineering works, such as cracks, fissures or ruptures of houses, buildings, reservoirs and canals, depressions in road pavements and landfills, and formation of slope slip-surface. The collapse can be generated by the change of state tension and by the physical-chemical interaction between percolating liquids and soil. Many cases of collapse have been observed associated with leaks from the sewage and water supply, or leaks from storage system for fuel and chemical effluents. However, the soil collapse is usually studied considering only water flow. In the literature there are few reports of the fluid influence in a soil collapse characteristics. Some definitions of collapsible soils are found in the literature. At the Annual Convention of the ASCE (1976), in Philadelphia, collapsible soil was defined as the unsaturated soil that experiments a radical rearrangement of particles and a large volume reduction when flooded with or without additional load (Clemence & Finbarr, 1981). Many collapsible soils present in their structures sand grains coated clay particles, oxides of iron and aluminum, carbonates, among others, which constitute cementing between particles and non-clay particles of macro porous soils. The clay dispersion
Tests on the Reddish-Yellow Sand were performed with dented samples, the depth of 1.0 than 2.2 m.
579
10, 40, 160, 320 and 640 kPa. The duration of each stage of tension was such that the deformation between two consecutive time intervals (t/t 1) was less than 5% of the total ground deformation occurred until the previous time. In the double tests, two loading conditions were tested: in natural moisture (constant), and specimens previously flooded. To flood the samples were used: distilled water, bleach based solution, leachate, kitchen detergent based solution, non-processed sewage, soybean oil and soap powder based solution. The collapse potentials values obtained from the oedometer test singles and doubles were calculated by the following equations, respectively: CP (%) CP (%)
Figure 1. Location of Reddish-Yellow Sand in the Petrolndia CountyPE (Ferreira 1995).
H 100% Hi v 100% 1 v ( nat )
(1) (2)
The physical characterization was carried out according to the Technical Standards Brazilian Association. For sample preparation, the hygroscopic moisture of the air dried sample was determined, and then water was added so that the final moisture content reached 3.5% (natural moisture at a depth of 1.5 m). The sample was placed in a plastic bag and taken to a moist chamber to equalize the moisture. For the sample of soil moisture equalized, the volume of soil required for compression specimens with dry unit weight of field (s 16,50 kN/m3) in natural moisture (wc 3.5%) was calculated. The compaction was performed statically in a compression press type Testop with 10 ton of capacity and 0.008 mm/s of speed. The volume of soil, calculated previously, was placed in the ring, embedded in the compaction mold. The mold shape guarantee that the soil was not more compacted than necessary because it has a security system at the top, where the piston contact with the mold top prevents compression beyond that point. The methodology used for the samples molding (moisture control, compaction mold, specimens weights control and static compaction technique) resulted in samples with reproducibility of expected conditions, with physical constants of a soil with little variation in the values against expected. Oedometer conventional tests single and double, were done using presses type Bishop and oedometer cells type fixed-ring. The vertical stresses were applied incrementally (/ 1). In simple tests, the specimens were flooded in the tensions of
where: CP is the collapse potential, H is the specimen height variation due to flooding, Hi is the specimen height before the flood, v v(nat) v(inund), v(nat) is the volumetric strain of the soil loaded in the natural moisture, v(inund) is the volumetric strain of the soil loaded previously flooded. 2.2 Methodology of liquids analysis The surface tension of the liquid was determined by a Tensiometer integrated with Software Sigma 70. The pH was determined by a pH meter for aqueous solutions, except for oil, which was estimated by the tapes method to determine the pH. The electrical conductivity was determined according to the methodology of Standard Methods for Examination of Water and Wastewater (APHA, 1995). 2.3 Principal Components AnalysisPCA
The PCs determination involves two steps: translation of data to the origin and rotation around the origin. X is the matrix of type n x m, where n corresponds to the objects (samples) and m represents the variables (properties) of objects. Before the extraction of PCs, an algorithm for pre-processing the data is typically used, being the most common procedures the centralization of data on the average (translational) and auto-stepping. Centering the data on average ensures that the variables are replaced as the average zero. In the auto-stepping, data are mean-centered and divided by their standard deviations which means that once processed; each variable will be with unit variance. As variance is information synonymous, all the auto-stepping variables have the same amount of information.
580
Stepping is required for variables with different units, as a linear combination of different magnitudes makes no sense. If the variables are autostepping, then the PCs axes are the eigenvectors of the correlation matrix, with the first eigenvector corresponding to the largest eigenvalue. The PCs axes are also known as latent variables. These variables directions are specified by vectors p (loadings), which are the eigenvectors of the correlation matrix. The auto-stepping data matrix projections on these vectors are the scores vectors t, whose elements are objects coordinates in the PCs axes. The PCA results, after the pre-processing of the data matrix and the PCs extraction are: Variancetells how much information is modeled by each PC. The principal component importance can be expressed in two ways: Variance explainedis the variance contained in the PC in question. Residual varianceshows how much information remains in the data set after removal of the information represented by the components included in the model. Loadingsdescribe the data structure in terms of correlation between variables. The loading of each variable on a PC reflects how this variable contributes to the PC. In geometric terms, the loading is the cosine of the angle between the variable and the PC considered, a small angle indicates a high loading (which can vary between 1 and 1). Scoresdescribe the data structure in terms of samples patterns and, generally, show the differences or similarities between samples. 3 3.1 RESULTS AND ANALYSIS Soil physical characterization
Table 1. Flood liquids
Characterization of flood liquids. Weight Surface tension unit (kN/m3) (mN/m) pH 69,94 67,16 57,57 29,36 36,38 31,53 27,12 Electrical conductivity (mS/cm)
Distilled water 9,819 Bleach 9,826 Leachate 10,001 Kitchen det. 9,824 Non-p sewage 9,825 Soybean oil 9,003 Soap powder 9,879
6,18 0,0185 10,08 0,810 8,04 15,68 6,82 0,436 6,64 6,12 4,50 0,0225 10,38 9,25
substances used for that were: distilled water, leachate, non-processed sewage and some components of wastewater isolated: solutions based on bleach, kitchen detergent and soap powder, prepared in the laboratory, concentration 1:120, and soybean oil, used pure. 3.3 Compressibility and collapsibility characterization
The Reddish-Yellow Sand of Petrolndia-PE is very uniform and poorly graded. The test with deflocculant showed 88% sand, 3% silt and 9% clay. The value of the specific weight of grains of 26.4 kN/m3 suggests a mineralogy composed mainly by quartz. The soil is non liquid and not plastic, is classified under the group SM, silty sand, in the SUCS classification. The value of maximum dry density, obtained in the Normal Proctor energy, is 19.25 kN/m3 at optimum moisture content of 8.85%, with degree of saturation of 62.77% in this condition. 3.2 Flood liquids characterization
The different physicochemical properties of the liquid used to flood the samples, characterized in laboratory tests are presented in Table 1. The
The compressibility of the compacted soil is small. In double oedometer tests, the variation of volumetric strain (v) with the consolidation vertical stress (v) of the compacted soil with constant moisture was 4.85%, Figure 2. The volumetric strain of the soil flooded with distilled water was 10.5% at the end of the load, a compressibility of 2.17 times that of natural soil, which represents 6% of collapse potential. The largest volumetric strains studied were 12.1 and 11.3%, related to the collapse potentials 7.7 and 6.8%, checked with the soil flooded with liquid alkaline pH, and the solutions based on Bleach and Soap Powder, respectively. The other strains were close to the soil flooded with distilled water for liquids of pH close to neutral, while soybean oil, of acid pH, virtually did not generate collapse. In single oedometer tests, flooding the soil with distilled water and stress of 640 kPa, led to a collapse potential of 4.8%, while the highest values of collapse potential were 6.2 and 5.7 checked after flooding the soil with leachate and non-processed sewage, respectively. Figure 3 shows the variation of the collapse potential (CP) with consolidation vertical stress (v) of soil flooded with different liquids. The magnitude of collapse potentials of the soil obtained through double and single oedometer tests (CPdouble versus CPsingle) are presented in Figure 4. It can be seen that the most part of experimental results is located above the equal line, indicating the superiority of the collapse strains obtained through double oedometer tests. The collapse potentials obtained in double oedometric tests are 11% bigger than the obtained in single oedometer tests.
581
0 2 4
Natural Moisture
3.4
Influence of liquids properties
v (%)
6 8 10 12 14 1
Distilled Water Bleach Leachate Kitchen Detergent Non-proc. Sewage Soybean Oil Soap Powder
10
100
1000
Consolidation Vertical Stress (kPa)
Figure 2. tests.
0 1 Collapse Potential (%) 2
Curves v versus v (log)double oedometer
The pH of the liquid used to flood single and double Oedometer tests are correlated with collapse potential of the soil, under stress of 160 kPa, as shown in Figure 5. A tendency to higher collapse potentials verified mainly in double oedometer tests is observed in liquids of more alkaline pH. The surface tension influence of the liquid used to flood the soil in the collapse behavior was evaluated during the flooding on single oedometer tests. Deformations of soils flooded with liquids of lower surface tension took longer to stabilize (Fig. 6). The electrical conductivity of liquids used for flood influences the ions mobility in the soil. The soil showed higher collapse potentials when flooded with good conductivity liquids. These results are presented in Figure 7, comparing the collapse potentials of the double oedometer tests, under stress of 640 kPa, with the liquid electrical conductivities.
9 Collapse Potential - Double Oedometric (%) 8 7 6 5 4 3 2 1 0 -1 0 2 4 6 pH 8 10 12 14 CPdouble = 0,71pH - 2,71 R2 = 0,88
3 4 5 6 7 10 100 Consolidation Vertical Stress (kPa) 1000 Distilled Water Bleach Leachate Kitchen Detergent Non-proc. Sewage Soybean Oil Soap Powder
Figure 3. tests.
8 7 6
Curves CP versus v (log)single oedometer
1
CPdouble = 1,106CPsingle R2 = 0,90
Figure 5. Variation of collapse potential of the soil under stress of 160 kPa, with pH of flood liquids in double oedometer tests.
0
CP (%) - Double Oedometer
1
4
Collapse Deformation (%)
3 2 1 0
3 Distilled Water Bleach Leachate Kitchen Detergent Non-proc. Sewage Soybean Oil Soap Powder 0 1 10 time (min) 100 1000 10000
5
-1 -1 0 1 2 3 4 5 6 7 8
CP (%) - Single Oedometer
Figure 4. Comparison between the collapse potential of the soil obtained through double and single oedometric tests.
Figure 6. Deformation of collapse versus t (log) in the v of 160 kPa, with different flood liquids.
582
9 Collapse Potential - Double Oedometric (%) 8 7 6 5 4 3 2 1 0 -1 0,01 CP double = 0,87Log(conductivity) + 5,71 R2 = 0,30 0,1 1 10 100
Table 2. Variance explained and cumulative of the matrixproperties of liquids and soil collapse potential with different stresses of flooding. Explained (%) PC1 PC2 PC3 64,94 15,51 9,63 Cumulative (%) 64,94 80,45 90,08
LOADS
1,0
TS
Electrical Conductivity (mS/cm)
PC2 : 15,51%
Figure 7. Collapse potential versus electrical conductivity of the flood liquids in double oedometer tests.
0,5
PEsp TI640 TI10 TI320

0,0
Cond
3.5
Principal Components AnalysisPCA
pH TI160
To observe the influence of the properties of different liquids used to flood the samples in the soil collapse using PCA, the data matrix formed by the liquid physicochemical properties, Table 1, and soil collapse potential with the different stresses of flooding obtained by single oedometer tests, Figure 3, was used as input data. The algorithm adopted for the pre-processing of the data matrix was the auto-stepping. All calculations were performed using the software Statistica 6.0 (Statsoft, USA). The percentage variance explained and cumulative (Sharaf et al. 1986) for the first two principal components is about 80% to 90% when we include the third major component, Table 2. Figures 89 show the graphs of loads and scores for the first two principal components. The scores graph shows a contrast along the first principal component (PC1), an ordering of tests depending on the differentiation properties of liquids used in flooded soil. It can be seen in Figure 9, the PC1, that the flooded soil with soybean oil differed from the others, mainly from the flooded soil with a solution based on soap powder and bleach, leachate and non-processed sewage due to pH and electrical conductivity. The unit weight value also stood out for these variables, but was not a determinant of the differential collapse of the soil because their values are very close. The PC2 clearly shows a pattern with the formation of two groups, depending on surface tension, highlighting the soil flooded with distilled water, solution based on bleach and leachate, due to high values. It is observed that none of the loads influence the score of the flooded soil with soybean oil. Although in the graph of the loads the same correlation between all variables, highlighting the
-0,5
TI40
-1,0 -1,0 -0,5 0,0 0,5 1,0
PC1 : 64,94%
Figure 8. Graph loads for the two first principal components (PCs) of the matrixproperties of liquids and flooding stress.
SCORES
2,5 2,0 1,5 1,0
Leachate Bleach A.Sanitria
Distilled Water
PC 2 : 15,51%
0,5 0,0 -0,5 -1,0 -1,5 -2,0 -2,5 -3,0 -6 -5 -4 -3 -2 -1 0 1 2 3 4 5 6 7
Detergente Non-proc. Sewage Kitchen Detergent
Soybean Oil
Soap Powder
PC 1: 64,94%
Figure 9. Graph scores for the two first principal components (PCs) of the matrixproperties of liquids and flooding stress.
flooding stress, and the properties: weight unit, pH and electrical conductivity. Associated with the graph of the loads, the scores are showing that the soil flooded with distilled water and solution based on kitchen detergent does not have a significant result, which means that it approached the average. While the
583
soil flooded with a solution based on soap powder, leachate, non-processed sewage and solution based on bleach stood out, and can be seen that the properties of liquids flooding in highlighted, have greater influence on the collapse potential. In PC2, the scores graph shows the contrast between the soil flooded with distilled water, leachate and solution based on bleach, and solution based on soap powder, non-processed sewage, oil soybean and solution based on kitchen detergent, where the surface tension was the property that contributed most to this differentiation. 4 CONCLUSIONS
The Oedometer test results for the compacted soil with s 16.50 kN/m3 and wc 3.50% indicates that the soil collapse is associated not only to suction decrease, but also to the physical and chemical interactions between soil and flood liquid. This fact suggests that the infiltration of different liquids by accidental leaks due to disruptions of tanks and pipes can cause greater collapse of the soil than those caused in the rainy season. The physicochemical properties of liquids used to flood the soil showed influence in interaction with the soil. The lower the surface tension, the greater the potential for wetting the soil, however, the soil-liquid interaction occurs slowly. The liquid alkaline pH showed a tendency to produce higher collapse potential. The liquids with higher electrical conductivities showed a slight tendency to define higher collapse potential of the soil. However, one must examine all factors that influence the behavior of soil collapse, and not assign it only to a single factor. It has been concluded that the collapse of the reddish-yellow sand depends on its structure, imposed changes in states of stress imposed and the physical-chemical interactions of the flood liquid with structural links, which could increase its potential collapse when flooded by liquid contaminants. It was presented a method using Principal Component Analysis in a soil analysis. The PCA technique allowed the visualization of the soil samples efficiently, showing that this data analysis type allows obtaining fast and reliable information about the similarity between the samples through the graphical preview.
The graphic results of scores and loads of the first two principal components of the matrix liquid properties and soil collapse potential with different stresses of flooding indicate that the soil collapse is associated not only to reducing suction, but also the physical and chemical interactions between soil and flood liquid. It was observed that pH was the main parameter responsible for this interaction. This analysis was considered satisfactory when compared to the completion of the analysis without PCA, which determined that liquids with alkaline pH showed a tendency to produce higher soil collapse potential. It is important to highlight the need for greater interaction between disciplines in interpreting the information of problems of the soil. Researchers and professionals in areas such as agronomy, biology, geology, chemistry, mathematics, statistics and computer science must maintain a permanent exchange in the way to permit an updated overview of problems and possible solutions. REFERENCES
APHAAmerican Public Health Association. Standard Methods for Examination of Water and Wastewater 19th ed. USA, AWWA, WEF. 1995. Carvalho, J.C., Nunes, P.M., Berberiam, D. & Ferreira, E.S. Influencia Del pH Del Lquido de Saturacin en la Colapsabilidad. VIII Cong. Panamer. de Mecnica de Suelos e Ingeniera de Fundaciones, Cartagena, Colombia: vol. 2, pp. 18. 1987.` Clemence, S.P. & Finbarr, A.O. Design Considerations for Collapsible Soils. Journal of the Geotechnical Engineering Division, ASCE, vol. 107, n GT3, March, pp. 305317. 1981. Ferreira, S.R.M. Colapso e Expanso de Solos Naturais No Saturados Devido Inundao. Tese de Doutorado. COOPE/UFRJ. Rio de Janeiro: Maro, 379 p. 1995. Ingles, O.G. & Aitchison. Soil-Water Disequilibrium as a Cause of Subsidence in Natural Soils and Earth Embankments. International Symposium on Land Subsidence, Tokio, AIKS Publication, n 89, vol. II, 342 p. 1969. Reginatto, A.R. & Ferrero, J.C. Collapse Potential of Soils and Soil-Water Chemistry. Proceedings, 8TH Internat. Confer. on Soil Mech. and Found. Engineering. Moscow, vol. 2.2, pp. 177183. 1973. Sharaf, M.A., Illman, D.L., Kowalski, B. Chemometrics, New York, Wiley. 1986.
584
Thermal damage in unsaturated geomaterials

M. Mozayan Kharazi C. Arson
Ecole Normale Suprieure de Cachan, LMT Cachan, France U.R. Navier (CERMES, ENPC), University of Paris-Est, France
B. Gatmiri
Direction ScientifiqueMcanique des Fluides et des Solides, ANDRA, France
ABSTRACT: This paper presents the frame of the THHMD damage model, designed for unsaturated porous media and formulated in independent state variables. The damage variable is a second-order tensor. The behaviour law stems from both micromechanical and phenomenological concepts. A crack related intrinsic water permeability is introduced. The THHMD model has been programmed in -Stock Finite Element code (Arson & Gatmiri 2008). A parametric study has been performed to study damage in heated samples of unsaturated bentonite and granite. Various loads and damage parameters have been used. The results show good physical trends. 1 GENERAL INSTRUCTIONS is a second-order tensor expressed in its principal base: ij d k ni k n j k
k 1 3
This study is motivated by the necessity to predict the behaviour of the Excavation Damaged Zone (EDZ) surrounding nuclear waste disposals. The geological barriers, often made of quasibrittle material like granite or clay-rock, undergo damage during the excavation phase. Most of the damage models dedicated to unsaturated porous media are based on the concept of Biots effective stress, which is not satisfactory to represent all the aspects of the behaviour of unsaturated geomaterials (Arson & Gatmiri 2008). Alternatively, the THHMD model is formulated in independent stress state variables (net stress, suction and thermal stress).The formulation of the THHMD damage model proposed here has been exposed in (Arson & Gatmiri 2009a,b, Gatmiri & Arson 2008). This article presents a study of the thermo hydro mechanical damage of granite and bentonite subjected to a heat flux in which the THHMD model is used. 2 2.1 FORMULATION OF THE THHMD DAMAGE MODEL A formulation in independent state variables
(1)
Each meso-crack is characterized by a direction nk (normal to the crack plane) and a volumetric fraction dk. The stress state variables chosen to formulate the model are net stress, ij ij pa ij , suction s pa pw and thermal stress pT (Arson & Gatmiri 2009b). In compliance with the former works of Gatmiri (Gatmiri & Arson 2008), it is assumed that the total strain tensor can be split into three components, each of which being thermodynamicallyassociated to a stress state variable (Arson & Gatmiri 2009a): 1 e d e d d ij d Mij d Mij ij (d SV d SV ) 3 1 e d ij (d TV d TV ) 3
(2)
Following the principle of spectral decomposition (Ortiz 1985), the internal damage variable ij
d d , d Sv and The incremental inelastic strains d M d d Tv are deduced from the increment of damage. This latter is computed by an associative flow rule.
585
2.2
Behaviour laws
The following expression of the free energy is postulated (Halm & Dragon 1998): 1 S ( Mij , Sv , Tv , ij ) Mji Deijkl (ij ) Mlk 2 1 1 SV s (ij ) SV TV T (ij )TV 2 2 gS g gM ji Mji ij ji SV T ij ji TV 3 3
the influence of cracking on vapour flow (Arson & Gatmiri 2009b). Air and heat flows are modelled by using the same flow rules as an intact porous medium (Gatmiri & Arson 2008). Damage affects flows indirectly and isotropically, through the damaged porosity (Arson & Gatmiri 2009b). 3 3.1 PARAMETRIC STUDY Comparison of the thermal response of two brittle geomaterials
(3)
The derivation of the free energy s (Mij, Sv, Tv, ij) provides the whole stress/strain relations. The operator of Cordebois and Sidoroff (Cordebois & Sidoroff 1982) is used in order to defined damaged stress state variables (Arson & Gatmiri 2009a,b). The damaged mechanical, capillary and thermal rigidities are then computed by applying the Principle of Equivalent Elastic Energy. 2.3 Flow rules
Moisture flows are assumed to be diffusive (Gatmiri & Arson 2008): Vw
R (W ) d (T ) K w .(T ) (Tref ) dT 1 (T ) K w .( s ) K w .( z ) w (Tref )
(4)
Vvap
vap w
Vvap DTvap(T ) DPvap( s )
(5)
Vw and Vvap refer to liquid water and vapour relative velocities, respectively. R is the relative hydraulic potential:
A numerical simulation of the laboratory test of Pintado (2001) has been performed with -Stock Finite Element code (Gatmiri & Arson 2008). The objective of the test is to apply a controlled flux of heat at the top of a cylindrical bentonite sample, the bottom being maintained at a constant temperature (Figure 2). The initial pore water pressure (pw0) is calculated by inversion of the function of the degree of saturation state surface. The initial saturation degree Sw0 is equal to 0.63 like in the experiment conditions. After a heating period of one week, a relaxation period of seven weeks is observed. All of the imposed boundary conditions are given on figure 3. Thermal and mechanical (Table 1) parameters have been varied in order to simulate the experimental test on two geomaterials (bentonite and granite). C0 is the initial damage-stress rate which is necessary to trigger damage. C1 is a yield function parameter, quantifying the influence of damage growth on the evolution of thermal stress (Gatmiri & Arson 2008). A high value of the damage rigidity gM is related to a high influence of tension strain on damage growth. According the adopted strain split
R ( w ) ( Pw Pa ) / w
(6)
(T) is the superficial energy of pore water (in J.m2). DTvap and DPvap are the thermal and capillary vapour conductivities, respectively. The water permeability tensor is split in an intrinsic part and in relative components: Kw kT (T ) kr (Sw ) K int ( n, ij ) (7)
Only the intrinsic water permeability, which depends on the behaviour of the solid skeleton, is influenced by damage (Arson & Gatmiri 2009a): K int ij ( n, rs ) kw 0 10 w e ij k2ij ( n frac , ij )
rev
(8)
The crack related component k2ij ( n frac , ij ) is computed by introducing an internal length parameter. The same approach is used to model
Figure 1. Mesh.
Bentonite heating test (Pintado et al, 2002).
586
Table 1. Mechanical parameters of the two geo materials. Material E Pa 0s Pa GM Pa C0 Pa C1 Pa 5.2e-3 2.2e6
Clay Rock 1.22e10 0.16 5.98e10 1.41 2.3e-4 Granite 3.51e10 0.30 6.07e11 3.3e8 1.1e5
Figure 2. Bentonite heating test (Pintado et al, 2002). Boundary conditions
Figure 4. Space evolution of damage in granite with gM 330 Pa and qt 26042 W/m2.
Figure 3. Space evolution of damage in bentonite with gM 1.414 Pa and qt 116.9 W/m2.
(Eq. 2), thermal stress is conjugated to isotropic thermal strains. Provided that damage evolves with tension strains (Gatmiri & Arson 2008), damage is thus expected to be isotropic. As shown in figures 4 and 5, damage increases with time for both materials. In bentonite, damage appears as soon as the sample is heated. It reaches
587
of the heating period, which means that cracks do not develop any longer when the source sis turned off. As expected, increasing the damage parameter gM for a fixed heat flux results in increasing the maximal value reached by damage (Fig. 6). 4 CONCLUSIONS
Figure 5. Damage gM 1.414 Pa.
evolution
in
bentonite
with
Fractured unsaturated porous media are studied with the THHMD damage model, formulated in independent state variables. Damage is a secondorder tensor. The behaviour law is deduced from a postulated expression of the free energy, and the degraded rigidities are computed by applying the Principle of Equivalent Elastic Energy. Internal length parameters are introduced in the moisture flow rules in order to represent the influence of damage on fluid transfers. The THHMD model has been implemented in -Stock Finite Element code. Parametric studies have been performed on two unsaturated geomaterials. Damage follows the expected trends with satisfactory orders of magnitude. In the mid-term, the THHMD is expected to model the EDZ. REFERENCES
Arson, C. & B. Gatmiri. 2008. On damage modelling in unsaturated clay rocks. Physics and Chemistry of the Earth. 33. S407S415. Arson, C. & B. Gatmiri. 2009a. A mixed damage model for unsaturated porous media. Comptes-Rendus de lAcadmie des Sciences de Paris, section Mcanique.337: 6874. Arson, C. & B. Gatmiri. 2009b. Parametric study on the performance of a THM damage model for unsaturated porous media, 1st International Symposium on Computational Geomechanics. Juan-les-Pins. France. Cordebois J.P. & F. Sidoroff. 1982. Endommagement anisotrope en lasticit et plasticit. Journal de Mcanique thorique et applique. Special issue: 4560. Gatmiri, B. & C. Arson. 2008. -STOCK. a powerful tool of thermohydromechanical behaviour and damage modeling of unsaturated porous media. Computers and Geotechnics, 35(6): 890915. Halm D. & A. Dragon. 1998. An anisotropic model of damage and frictional sliding for brittle materials, European Journal of Mechanics and Solids. 17(3): 439460. Ortiz, M. 1985. A constitutive theory for the inelastic behavior of concrete, Mechanics of Materials. (4): 6793. Pintado X., Ledesma A. & Lloret A. 2002. Backanalysis of thermohydraulic bentonite properties. Engineering Geology. 64(2): 91115(25).
Figure 6. Damage qt 116.9 W/m2.
evolution
in
bentonite
with
its maximum value at the end of the heating period and remains constant during the relaxation period. In granite, damage starts suddenly when the heat flux is stopped (around the eighth day) and continues to increase during the relaxation period. This can be explained by the difference of thermal conductivities and capacities, or by the difference of heating power which is necessary to damage each material. Another possible explanation is that C0 is larger in granite than in bentonite. Figures 4 and 5 also show that the part of the sample which is the closest to the heater is the most damaged. In bentonite, the damaged length is 16 mm, which is about of 20% of the length of the sample, whereas in granite, the damaged length rises to 23 mm, which is almost 30% of the length of the sample. 3.2 Parametric study of thermal damage in bentonite
Damage is initiated as soon as the thermal solicitation is applied, and reaches a maximum at the end
588
Thermo-hydro-mechanical behaviour of an expansive bentonite/silt mixture using different suction controlled techniques
Hossein Nowamooz
Laboratoire Environnement Gomcanique and Ouvrages, Nancy-Universit, Vandoeuvre-ls-Nancy Cedex, France Laboratoire de Gnie de la Conception, INSA de Strasbourg, Strasbourg Cedex, France
Roberto Gmez-Espina, Mara Victoria Villar

CIEMAT, Avd. Complutense, Madrid, Spain
Farimah Masrouri
Laboratoire Environnement Gomcanique and Ouvrages, Marcel Roubault, Vandoeuvre-ls-Nancy Cedex, France
ABSTRACT: This paper presents an experimental study on a bentonite/silt mixture using odometers with suction controlled either by the osmotic technique or by air overpressure at two different temperatures of 20 and 60C in a suction range between 0 and 8 MPa. The compression curves permitted us to interpret the influence of the temperature on the different yielding surfaces. The temperature increase rigidified the samples producing a significant increase of the preconsolidation stress. We observed that the compression curves at different imposed suctions converged toward the saturated state for the high applied vertical stresses. The stress from which the compression curves follow the saturated state is called the saturation pressure (Psat). The Psat values increased with the temperature increase. 1 INTRODUCTION experimental information is limited to relatively low suctions (Villar 1999; Wiebe et al., 1998; Romero 1999; Saix et al., 2000; Romero et al., 2003, Villar & LLoret 2004, Romero et al., 2005). In this context, this article presents the thermohydromechanical behaviour of a silt/bentonite mixture using two different suction imposition techniques: osmotic technique and air overpressure technique at two different temperatures 20 and 60C. 2 STUDIED MATERIAL AND SAMPLE PREPARATION
Temperature changes affect important hydraulic characteristics of compacted clays such as the water retention and the water permeability. In addition most mechanical responses such as swelling pressure, swelling and collapse behaviour, compressibility, yielding, effects on time dependent behaviour are also modified. During the last years a number of laboratory results referring to thermal effects on saturated soils have been presented (Demars and Charles, 1982, Hueckel and Baldi, 1990, Baldi et al., 1991, Towhata et al., 1993, Kuntiwattanakul et al., 1995, Komine and Ogata, 1994; Delage et al., 2000; Burghignoli et al., 2000; Graham et al., 2001, Sultan et al., 2002, Shimizu, 2003). It is known that saturated clays under heavily overconsolidated conditions exhibit quasi-reversible volume expansion when heated (Baldi et al., 1988; Sultan et al., 2002). On the other hand, the thermal contraction increases when the overconsolidation ratio (OCR) is decreased, leading to pure irreversible contraction at OCR 1 (Demars and Charles, 1982; Sultan et al., 2002). However today, little is known about the influence of temperature on the volume change behaviour of unsaturated clays and the
The studied material was made of 40% silt and 60% of a commercially available bentonite. The mineralogical composition of these materials was determined by X ray diffractometry. The silt contains 60% quartz, 20% montmorillonite, 11% feldspar, with the remaining part containing kaolinite and mica. The bentonite contains more than 80% calcium montmorillonite. The liquid limit of the mixture was 87% and its plastic index was 21%. The two materials were mixed together and wetted to a gravimetric water content of 15% which is
589
a value very close to their shrinkage limit. The wet mixture was then sealed in an airtight container and left for at least ten days to reach moisture equilibration inside the mixture. After this period, the mixture was statically compacted under a vertical stress of 1 MPa, directly inside the oedometer ring. It was not possible to prepare samples with a mass water content under 15% because in this case the samples crumbled. The initial dry density was 12.7 0.1 kN m3 for all tested samples. The initial height of the samples was between 10 and 12 mm and their diameter was 70 mm in the odometers. The initial matric suction, measured in accordance with ASTM Test Method for Measurement of Soil Potential (Suction) using Filter Paper (D 5298), was comprised between 20 and 25 MPa. Under these conditions, the swelling potential and the swelling pressure, measured with the free swelling technique in accordance with ASTM Test Method for One-dimensional swell or settlement potential of cohesive soils (D 5298), were respectively 19% and 250 kPa. 3 EXPERIMENTAL TECHNIQUES
under a vertical stress of 0.1 MPa under different suction values. After equilibration of deformation (or after a given period of time), the vertical stress was increased stepwise 4 COMPRESSION CURVES AT THE TEMPERATURE OF 60C
The study of the hydromechanical behaviour of swelling soils in this work required experimental devices able to impose a suction range between zero and several MPa. Therefore, the osmotic technique (Kassif & Ben Shalom 1971) as modified by Delage et al., (1992) and air overpressure method (Escario and Saez, 1973) were selected. Each technique was applied in a different laboratory. In the osmotic method, the soil sample and a solution of macromolecules are placed in contact with a semi-permeable membrane between them (Zur, 1966). This membrane prevents the PEG (polyethylene glycol) macromolecules from moving towards the sample, but it allows for water exchange. Water movements, and thus suction variations, are controlled by the osmotic phenomenon. The higher the concentration of the solution, the higher the imposed suction. The base of the cell used for the air overpressure technique has an embedded porous stone, below which there are two inlet and outlet holes, connected to a deposit with water at atmospheric pressure. A cellulose membrane is placed over this stone, with the sample resting directly on it. This membrane allows water and ions to pass, but not gas. A peristaltic pump, installed between the deposit and the cell inlets, facilitates the removal of the gas that could diffuse through the membrane. Nitrogen is used to apply the gas pressure in excess of the water pressure on top of the sample. Also describe briefly the paths followed: Once in the oedometer frame, the samples were initially stabilized
This section reports the results on the influence of temperature 60C on some hydro-mechanical properties of the compacted bentonite/silt mixture. Figure 1-a presents all the compression curves at the constant temperature of 60C using the air overpressure technique (M6, M7 and M8 tests). These experimental results were compared with the five tests performed at the temperature of 20C: M1, M2, M4 and M5 tests (Figure 1-b). The maximum applied vertical stress for all the air overpressure tests was about 2000 kPa and for the osmotic tests about 1000 kPa. It should be mentioned that the M3 test was performed with the air overpressure technique at the temperature of 20C. The coherence between the results obtained with both methods is remarkable, as can be observed when comparing tests M2 and M3 (Figure 2-b). Figure 2 compares also the compression curves at two different temperatures at different constant suctions: 0, 2 and 4 MPa.
a) T=60C
1.4 1.3 1.2 1.1
M6 (s=0 MPa) M7 (s=2 MPa) M8 (s=4 MPa)
Void ratio (-)
0.9 0.8 0.7
Normally Consolidated Curve (NCC)
0.6 0.5 10 100 1000 10000
Vertical net stress (kPa)
b) T=20C
1.4 1.3
M1 (s=0 MPa) M2 (s=2 MPa)
1.2
M4 (s=4 MPa)
1.1
M5 (s=8 MPa)
Void ratio (-)
0.9
0.8
0.7
0.6 10 100 1000 10000
Figure 1. Compression curves at different suctions a) (T 60C) b) (T 20C).
590
a)
1.4 1.3
Table 1. Mechanical parameters at different applied suctions and temperatures.

M1 (Free swelling technique, T=20) M6 (Air overpressure technique, T=60)
1.2
1.1
Tempera- Suction Test ture (C) (MPa) (s) M1 M2 M3 M4 M5 M6 M7 M8 20 20 20 20 20 60 60 60 0 2 2 4 8 0 2 4 0.31 0.31 0.30 0.30 0.27 0.31 0.30 0.29 0.045 0.015 0.015 0.015 0.015 0.015 0.015 0.015
r 0.045 0.045 0.045 0.015 0.015 0.04 0.035 0.015
P0 Psat (kPa) (kPa) 60 110 150 200 350 110 250 350 150 300 400 800 1500 300 1200 1600
Void ratio(-)
0.9
0.8
0.7
0.6
0.5
10
100
1000
10000
b)
1.4 1.3
M2 (Osmotic technique, T=20) M3 (Air overpressure technique, T=20)
1.2
M7 (Air overpressure technique, T=60)
100
1.1
(LC)at 20
(LC)at 60 (SC)at 20 (SC)at 60
Void ratio(-)
10
0.9
0.8
Suction(MPa)
0.7
0.6
0.1
0.5 10 100 1000 10000
Initial state Yielding surfaces at T= 20 Yielding surfaces at T= 60
c)
1.4 1.3

0.01 10 100 1000 10000
M4 (Osmotic technique, T=20) M8 (Air overpressure technique, T=60)
1.2
1.1
Void ratio(-)
Figure 3. surfaces.
Influence of temperature on the yielding
0.9
0.8
0.7
0.6
0.5 10 100 1000 10000
Figure 2. Compression curves at different temperatures at constant suctions of (a) 0, (b) 2 and (c) 4 MPa.
The compression curves made it possible to estimate the soil mechanical behaviour, i.e., the virgin compression index (0), the preconsolidation mean net stress p0(s),, the elastic compression index and the elastic recompression or unloading index r, summarized in Table 1. Nowamooz and Masrouri (2008, 2009) firstly defined a normally consolidated curve (NCC) for the mixtures compacted under different initial compaction pressures at the constant temperature of 20. They also observed that the compression curves at the unsaturated states converge to the normally consolidated curve (NCC) for the higher stresses defined as the saturation stress psat presented also in Table 1. The following comments can be made: It is important to mention that the normally consolidated curve (NCC) is completely independent
to the temperature. The compression curve at the saturated state followed the curve at the temperature of 60 (Fig. 1 & 2). The preconsolidation stress p0 and the saturation stress psat increased when the temperature increased. The virgin compression index and the elastic compression index values are not influenced by the temperature variations. decreased slightly as suction increased for both samples and decreased with suction increase for both samples. The compression curves permitted to illustrate the influence of the temperature variation on yielding surface (Figure 3). The temperature increase rigidified the mechanical behaviour of the studied samples, the same results reported by Romero et al., (2005). However, these results do not confirm the previous observations indicating a decrease of preconsolidation stress with the temperature increase (Laloui & Cekerevac 2003, Cekerevac & Laloui 2004). At high temperatures, the swelling capacity of clay decreases, as shown in the first step of the tests, in which the samples were submitted to different suctions under a low vertical stress. The influence of temperature is less evident when the applied stress is high.
591
CONCLUSION
This paper reports the results on the influence of temperature (20 and 60C) on the hydro-mechanical properties of a compacted bentonite/silt mixture. The temperature increase rigidified the samples producing a significant increase of the preconsolidation stress. We observed that the compression curves at different imposed suctions converged toward the saturated state for the high applied vertical stresses. The stress from which the compression curves follow the saturated state is called the saturation pressure (Psat). The Psat values increased with the temperature increase. It is important to mention that the normally consolidated curve (NCC) is completely independent to the temperature. The compression curve at the saturated state followed the curve at the temperature of 60. REFERENCES
ASTM D 5298-94, (1995). Standard test method for measurement of soil potential (suction) using filter paper, 4.09, 154159. ASTM D 4546-90, (1995). One-dimensional swell or settlement potential of cohesive soils, 4.08, 693699. Baldi, G., Hueckel, T. & Pellegrini, R. (1988). Thermal volume change of the mineral-water system in lowporosity clay soils. Can. Geotech. J. 25, 807825. hal-00250248, version 111 Feb 2008. Baldi, G., Hueckel, T., Peano, A. & Pellegrini, R. (1991). Developments in modelling of thermo-hydrogeomechanical behaviour of boom clay and claybased buffer marerials. Commission of the European Communities, Nuclear Science and Technology, EUR 13365/1 and EUR 13365/2. Burghignoli, A., Desideri, A. & Miliziano, S. (2000). A laboratory study on the thermomechanical behaviour of clayey soils. Can. Geotech. J. 37, 764780. Cekerevac, C. & Laloui, L. (2004). Experimental study of thermal effects on the mechanical behaviour of a clay. Int. J. Numer. Anal. Methods Geomech. 28, No. 3, 209228. Delage, P., Suraj Da Silva, G.P.R. & De Laure, E., (1992). Suction controlled testing of non-saturated soils with an osmotic consolidometer. Proc. of the 7th Int. Conf. on Expansive Soils, Dallas, pp. 206211. Delage, P, Sultan, N. & Cui, Y.J. (2000). On the thermal consolidation of Boom clay. Can. Geotech. J. 37, 343354. Demars, K.R. & Charles, R.D. (1982). Soil volume changes induced by temperature cycling. Can. Geotech. J. 19, 188194. Escario, V. & Saez, J. (1973). Measurement of the properties of swelling and collapsing soils under controlled suction, Proc. 3rd Int. Conf. Expansive Soils, Haifa, Israel, 195200. Graham, J., Tanaka, N., Crilly, T. & Alfaro, M. (2001). Modified camclay modelling of temperature effects in clays. Can. Geotech. J. 38, 608621.
Hueckel, T. & Baldi, G. (1990). Thermoplasticity of saturated clays: Experimental constitutive study. Journal of Geotechnical Engineering 116, No. 12, 17781796. Kassif, G. & Ben Shalom, A., (1971). Experimental relationship between swell pressure and suction. Gotechnique, vol. 21, pp. 245255. Komine, H. & Ogata, N. (1994). Experimental study on swelling characteristics of compacted bentonite. Can. Geotech. J. 31, 487490. Kuntiwattanakul, P., Towhata, I., Ohishi, K. & Seko, I. (1995). Temperature effects on undrained shear characteristics of clay. Soil and Foundations 35, No. 1, 147162. Laloui, L. & et Cekerevac, C., 2003. Thermo-plasticity of clays: An isotropic yield mechanism. Computers and Geotechnics 30, No. 8, 649660. Nowamooz, H. & Masrouri, F., (2008). Hydromechanical behaviour of an expansive bentonite/silt mixture in cyclic suction-controlled drying and wetting tests, International Journal of Engineering Geology, n101 (34), 154164. Nowamooz, H., Masrouri, F., (2009). Density-dependent hydromechanical behaviour of a compacted expansive soil: experimental and analytical aspects, Engineering Geology, n106 (34), 105115. Romero, E., 1999. Characterisation and thermo-hydromechanical behaviour of unsaturated Boom-clay: an experimental study. Ph. D. Thesis, Universidad Politcnica de Cataluna, Barcelona. 405 pp. Romero, E., Gens, A. & Lloret, A. (2003). Suction effects on a compacted clay under nonisothermal conditions. Gotechnique 53, No. 1, 6581. Romero, E., Villar, M.V. & Lloret, A. (2005). Thermohydro-mechanical behaviour of heavily overconsolidated clays. Engineering Geology 81, 255268. Saix, C., Devillers, P. & El Youssoufi, M.S. (2000). Elment de couplage thermomcanique dans la consolidation de sols non saturs. Can. Geotech. J. 37, 308317. Shimizu, M. (2003). Quantitative assessment of thermal acceleration of time effects in onedimensional compression of clays. Deformation Characteristics of Geomaterials, pp. 479487. Sultan, N., Delage, P. & Cui, Y.J. (2002). Temperature effects on the volume change behaviour of Boom clay. Engineering Geology 64, 135145. Towhata, I., Kuntiwattanakul, P., Seko, I. & Ohishi, K. (1993). Volume change of clays induced by heating as observed in consolidation tests. Soils and Foundations 33, No. 4, 170183. Villar, M.V. (1999). Investigation of the behaviour of bentonite by means of suction controlled oedometer tests. Engineering Geology 54, 6773. Villar, M.V. & Lloret, A. (2004). Influence of temperature on the hydro-mechanical behaviour of a compacted bentonite. Applied Clay Science 26, 337350. Wiebe, N., Graham, J., Tang, G.X. & Dixon, D., (1998). Influence of pressure, saturation and temperature on the behaviour of unsaturated sandbentonite. Can. Geotech. J. 35, 194205. Zur, B., 1966. Osmotic control of the matric soil water potential: I. Soil water system. Soil Science 102, 394398.
592
Direct measurement of thermal expansion in unsaturated soils

X. Pintado A. Lloret
BTECH Oy, Helsinki, Finland Geotechnical Engineering and Geo-sciences Department, Universitat Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: A method designed to measure the thermal dilatation coefficient of unsaturated soils is presented. It is based on the ASTM 4535-85 standard with some important considerations taken into account. A number of tests following this methodology were performed on unsaturated swelling clay. Thermal dilatation coefficients were measured over a temperature range from 25 to 65C for material dry densities and saturation degrees varying between 1617 kN/m3 and 6095%, respectively. The results are somewhat disperse, nevertheless, a clear trend with regard to temperature can be observed and thermal dilatation coefficient values can be extracted. 1 INTRODUCTION
The thermal behaviour of swelling clays has taken added importance due the development of high level nuclear waste repositories. Although clay deformations will result mainly from hydration, thermal deformations will play an important role in the early stages of repository lifetimes because heat transport, mainly by conduction, is faster than the water transport. ASTM 4535-85 describes a method for measuring the thermal coefficient of dilatation in rocks. The test is conducted by measuring the axial strains of a cylindrical soil sample in a temperature controlled bath. Due to the sensitivity of clay materials to water contact, the samples must be properly sealed. 2 TEST DESCRIPTION
Figure 1. Test device.
A stainless steel platform is used to support the sample and measuring system in the water bath (Figure 1). A sample with a diameter of 38 mm and a length of 78 mm sits on the bottom level of the platform. The sample is sandwiched between a Pyrex pedestal on the bottom and a Pyrex cap on the top. These Pyrex pieces are of equal size and shape. A latex membrane is pulled over the sample and Pyrex sandwich pieces. A long Pyrex rod is placed vertically onto the bottom platform level through holes in the top and middle levels. A shorter Pyrex rod is placed vertically onto the Pyrex cap piece through holes in the top and middle
platform levels as well. A micrometer is situated on top of the shorter Pyrex rod in order to measure displacement of the soil sample plus shorter Pyrex rod against the longer Pyrex rod. The entire platform system was submerged between the top and middle levels in a 15 l water bath controlled with resistive heating elements. The maximum bath temperature was 65C. As the bath temperature increases, the strains in both Pyrex rods should be equivalent. Any displacements measured by the micrometer are due to the difference between the soil strains and the portion of the long Pyrex rod with the same length as the sample. The thermal coefficient of Pyrex is approximately 3.75 106 1/K (one order of magnitude less than the thermal coefficient of the soil). The total sample dilatation will be the sum of the displacement measured by the micrometer plus
593
the thermal dilatation of the Pyrex. The thermal coefficient of Pyrex was determined against copper, which is a material with well-known thermal dilatation coefficient. As there is no confining pressure on the sample during the test, it is possible that a small amount of water may seep into the sample at the top or bottom. Saturation of the membrane may also influence the results. Therefore, it is important to perform displacement measurements over periods of constant temperature in order to check for these effects. The problem of partial sample hydration can be avoided through the use of non-hydrating bath liquids (e.g., silicone oil). Temperature was measured at three points on the sample with thermocouples. One such measurement was also used to control the bath temperature. In order to avoid temperature gradients, a pump was used to slowly circulate the bath water. Induced strains were measured continuously with an LVDT transducer. Additionally, a dial gauge was connected to the LVDT for continuous calibration. The latex membrane was sealed around the sample with a set of O-rings and silicone sealant. The continuous measures of the strains allow controlling the displacements when the temperature is constant. 3 RESULTS
Figure 2. Table 1.
Particle size distribution. Identification properties. 93 1 47 2 46 2 2.70 0.04 87 45 725 47 32 3
Liquid limit, in % Plastic limit, in % Plastic index Specific weight (g/cm3) Grain size distribution, in % Fraction less than 74 mm Fraction less than 2 mm Specific surface, in m2/g: Total External, BET
Table 2. Properties of sample set 1. d (kN/m3) 15.7 15.8 16.9 17.2 Table 3. d (kN/m3) 16.0 16.1 16.2 (%) 17.3 16.4 14.4 12.5 Sr (%) 64.2 62.7 61.8 56.2
The soil sample was a bentonite from the Cortijo de Archidona deposit (Almera, Spain) called FEBEX bentonite. This material was used in the FEBEX project (Enresa, 2000), which aimed to demonstrate the feasibility of the engineered barrier system (EBS) in nuclear waste repositories. FEBEX bentonite has a very high content of montmorillonitetype dioctahedric smectite (92% by XRD analysis). This material also contains numerous accessory minerals such as neoformed minerals and unaltered volcanic rock (Enresa, 2000). The particle size distribution of the tested material is shown in Figure 2.The geotechnical engineering identification properties are listed in Table 1. Three sets of samples were prepared to approximately three different levels of saturation (see Tables 2, 3, and 4). The thermal displacements were quite small and difficult to measure and displayed significant scatter; nevertheless some clear trends were observed. A typical series of results are presented in Figures 3, 4 and 5. The thermal expansion increases with increasing temperature for all sample sets. Furthermore, after undergoing an initial heating and cooling cycle, all of the samples display a distinct, unrecoverable expansion strain.
Properties of sample set 2. (%) 21.0 20.9 20.1 Sr (%) 82.2 82.7 80.7
Table 4. Properties of sample set 3. d (kN/m3) 16.0 16.7 16.8 (%) 25.1 21.6 20.8 Sr (%) 97.1 91.2 92.2
594
Figure 3. Vertical strains as a function of temperature for sample set 1.
Figure 6. Vertical strain relative to the change in temperature as a function of specific weight from 20 to 50C.
Figure 4. Vertical strains as a function of temperature for sample set 2.
Figure 7. Increment of linear strain and temperature relationship measured in tests with different dry densities and saturation degrees.
Figure 5. Vertical strain as a function of temperature for sample set 3.
It is rather difficult to clearly establish the role of the initial dry density and the water content. The change in vertical strain relative to the change in temperature is plotted as a function of specific weight in figure 6 from 20 to 50C. There may be a slight increase in this relationship with increasing specific weight, but the scatter in the data make any efforts
at quantification suspect. It has also not been possible to form conclusions about the role of the degree of saturation of the samples because, although the strains in samples with initial saturation degree of 0.8 have been larger than in samples with a saturation degree of 0.6, they have been also larger than samples with a saturation degree of 0.95. The change in vertical strain relative to the change in temperature is plotted as a function of temperature in Figure 7. The scatter in the data clearly increases with temperature. It can also be observed that the relationship values are larger in the initial heating and cooling cycle than during subsequent cycles. This response is coherent with the strain accumulation in the cycles of heatingcooling as can be seen in figures 2, 3 and 4. Linear fits to the data provide the following empirical relationships between thermal strain increases (z) and temperature (T): z/T 0.118 104 6.5 106 T (initial heating) z/T 1.265 104 6.5 106 T (additional heating)
595
z/T 1.538 104 6.5 106 T (cooling), (T in C) The effect of temperature on changes in volume due to thermal effects is similar in the three situations presented. If it is assumed that the changes of suction and confinement stress are small, and that after the first heating the sample is in an elastic regime and the thermal behaviour is isotropic, the increase in volume strain with change in temperature can be written as follows: vol . T
elas.
It is possible that the water from the bath could have an influence in the results due to water intrusion at the sample ends and to membrane saturation, possibly adding some swelling strains to the thermal strains. Replacing water for a non-hydrating bath liquid may be a future option. This test can also be used for measuring the thermal dilatation coefficient of rock. However, it may be preferable to use a different material than Pyrex (e.g., Invar alloy) as the thermal dilatation coefficient of rock is similar to Pyrex. REFERENCES
ASTM D4535-85. Methods for Measurement of Thermal Expansion of Stone using a Dilatometer. ASTM. Enresa 2000. FEBEX project. Full-scale engineered barriers for a deep geological repository for high level radioactive waste in crystalline host rock. Final Report. Technical Report 1/2000. Llore, A., Villar, M.V. & Pintado, X. 2002. Ensayos THM: Informe de sntesis. FEBEX Project. 70-UPCM-0-04. CIEMAT/DIAE/54520/1/02 (in Spanish). Muoz, J.J. 2006. Thermo-hydro-mechanical analysis of soft rock. Application to a large scale heating test and large scale ventilation test. PhD dissertation. Technical University of Catalonia. Barcelona. Pintado, X. 2002. Caracterizacin del comportamiento termo-hidro-mecnico de arcillas expansivas. PhD dissertation. Technical University of Catalonia. Barcelona (in Spanish).
6.75 10 5 1.95 10 5 (T 25), (T in o C )
For a mean temperature of 45C (mean value in the range of 30 and 60C in the tests), the volumetric thermal coefficient is calculated as: vol . T
elas.
4.57 10 4
CONCLUSIONS
A methodology for measuring the relation between the thermal expansion and the increase of temperature has been presented. The test results show significant scatter, but some empirical relationships have been isolated.
596
Determining the thermal conductivity as a function of water content for a soil with different percentages of gravels
C.M. Rubio & R. Josa
Department Agri-Food Engineering and Biotechnology, Technical University of Catalonia, Soils and Environmental Consulting Center, Cervera, Spain
D.R. Cobos & C.S. Campbell F. Ferrer
Decagon Devices, Inc.; Pullman, WA, USA Lab-Ferrer, Soils and Environmental Consulting Center, Cervera, Spain
ABSTRACT: The work presents the study of the relationship between the thermal and water dynamics for stony silt loam soils under laboratory conditions. The samples were obtained from Cam de Can Sol (NE of Spain). The determination of the thermal properties was performed with a thermal sensor of single needle, which uses the infinite line heat source method. The experiments were carried out on monitorized soil columns where their water content and thermal properties were continuously recorded. Different percentages of gravels were added to the soil samples. The variability of the information gathered depending on the coarse fragments was observed. The variability of the thermal and water content data relationship shown higher differences when the coarse elements were about 10% and 50%, being especially relevant the position of the thermal sensor inside the soil samples. 1 INTRODUCTION surface between fine and coarse particle size of the soil, as well. Thus, the macroporosity will be a variable affecting both soil phases, liquid and gas (Bouma, 1983; White, 1985), therefore at the soil thermal properties. Another aspect to take into account will be the influence of these fluxes about a correct development of the root system (Dexter, 19869). The aim of this research is to evaluate the influence of the coarse elements on the relationship between thermal and hydric soil properties. 2 MATERIALS AND METHODS
A soil coarse element refers to all type of particle size that it is not able to be sieved by a 2000 m of diameter. Usually, due to the soil genesis condicitons we can find an abundance coarse materials in the soil profile. There are different works that make reference to the effects caused by these materials about the soil thermal properties (e.g. Elbersen, 1982; Villani and Wright, 1990). Some of these research studies make reference to the quantity and its relationship with the mass volume, to the size, and the depth where these coarse elements are located (Garca Rodrguez, 1985). Other effects induced by these elements are directly related with the soil porosity (Flint and Childs, 1984; Espeby, 1989). Some investigations carried out during the last decade, were focused on the dynamic of the thermal properties of the porous media with coarse elements, observing an anisotropy and heterogeneity in the media (Ingelmo et al., 1994). Higher coarse elements in soils use to present some problems in the experimental design, especially for installing sensors and probes to monitorize the scenario, and not altering the soil structure. Beven and Germann (1982) suggest that the macroporosity in a consequence of the physical and chemical processes, and changes in the contact
Samples were obtained from the top soil horizon (030 cm) of a plot located at Can Sol Road, in the delta plain of Llobregat River (Northeast of Spain). To characterize the soil the physical variables, particle size distribution, bulk density, total organic carbon content, calcium carbonate content were measured. In addition, the residual water content (hygroscopic water) was determined. Particle-size distribution was determined using the wetting sieve method for 2000 to 500 m, and a device by dispersion laser beams (Malvern Mastersizer/E) for particles smaller than 500 m. Bulk density and total porosity were determined from undisturbed sample volumes. Total carbon
597
content was analyzed by loss on ignition at 900C, and inorganic carbon content by loss on ignition at 200C, both using a Shimadzu SSM-5000A and solid sample module. These results allowed to calculate both, total organic carbon content and calcium carbonate content. The residual water content was determined by loss in weight after drying the samples at 105C during 24 h. Synthetic samples were developed using different percentages (10%, 20%, 30%, 40%, 50% and 80% kg-gravelskg-soil1 air dried) of coarse elements (25% kg-gravelskg-soil1 air dried of 24 mm and 75% kg-gravelskg-soil1 air dried of 48 mm). The samples were repacked into a specific soil column device. Measurements of thermal-hydrodynamic properties were made on one soil column, constructed specifically for this experiment. Figure 1 shows the column one, which was developed in polyvinyl chloride component. Soil sample was repacket inside the cylinder and compacted to a target bulk density. To determine the thermal properties a thermal sensor, several single needle sensor (KS-1) (Decagon Devices Inc.) were employed. These kind of sensors use the heat pulse methodology and yield reliable soil thermal resistivity (R) and the inverse thermal conductivity () estimations, obtained by a non-linear least squares procedure during the process. The sensor was inserted in the middle of the soil sample. The thermal data were collected using a KD2Pro reader-logger. To determine the volumetric water content (), the soil column was monitored with ECH2O EC-5 frequency domain probe (Decagon Devices Inc.). A Decagon Devices Em-5b datalogger was required to collect the water content and room temperature data. The column device with repacked synthetic samples were placed during the experiment inside of a isothermal chamber to avoid the thermal drift.
The studied soil from Can Sol Road was classified as silt loam textural class (USDA), with a particle size distribution for silt content always higher than 60%, mean sand content about 34%, and mean clay content about 4%. Mean bulk density was 1.47 gcm3 and total porosity 45%. Mean total organic carbon content was about 3.1%, and mean calcium carbonate content was 40.3%. Figure 2, shows the thermal resistivity values obtained for a soil with different coarse elements content The soil property thermal resistivity was chosen because has a regular use in several aplications as are civil engineering, heat flux transfer in soils with permafrost, among other. The obtained values we consider that are in the range that obtained by other authors as e.g. Al Nakshabandi and Kohnke (1964) for this textural class. In detail, for the same water content (about the half average value in the field capacity water content). The data shows divergences on the thermal resistivity values when the gravel percentage is about 10% weightgravelsweight1sample. Obtained values for a same soil without gravels presented a mean thermal resistivity value close to 1.7 mKW1. When the gravel content increased, the thermal resistivity values increased, as well, presenting a higher value about 3.1 mKW1 for a gravel content of 10% weightgravelsweight1sample. Following the experiment, i.e. increasing the gravel content in the soil sample, the values presented a linear dereasing behaviour, showing a second thermal resistivity increase for 50% weightgravelsweight1sample. The lowest thermal
3.2
R Tsample Tchamber
32
2.8
28
R (mKW-1)
2.4
24
20
1.6
16
1.2
12
20
40
Gravels (%)
60
80
Figure 1. Top view of the soil column design with thermal sensors and moisture probes located.
Figure 2. Relationship between termal resistivity (R) and different gravel content for a silt loam soil. Tsample temperature of the sample; Tchamber temperature of the inner isothermal chamber.
598
Temperatura (C)
resistivity value was 1.3 mKW1 obtained for 80% weightgravelsweight1sample. The non-linearity in the thermal resistivity curve and its relationship with the several gravel content would related with the different variables related in the experiment. The large values of thermal resistivity for a samples with 10% weightgravelsweight1sample, relate an increase of the porosity of the sample (Johansen, 1975; Vanpelt, 1976; Oschner et al., 2001) due to increase of the coarse elements, and the position of the thermal sensor inside the soil. This fact was particularly special, just that the vertical position increased the anisotropy of the system soil-gravels, which presented variations of the thermal property according to the position of the sensor inside of the sample. The porosity less than 40% volvol1 (air fraction) presents a critical rol respect to the thermal conductivity or its invers the thermal resistivity (Naidu and Singh, 2004). On the other hand, the temperature of the sample and the isothermal chamber presented a negligible thermal oscillation, about 0.8 degrees Celsius, therefore the hydro-thermal equilibrium maintened in steady-state conditions, inside the soil column device. Several studies (De Vries, 1963; Campbell et al., 1994; Campbell and Normann, 1998) shown that in a wet soil aroun 10% to 20% of the heat was transferred in latent heat form cross to porous media. This factor, latent heat, presents high influence by the temperature, increasing a double value when the temperature increase 10 degrees Celsius. Therefore, we can hold on that the variable temperaure did not influence or lowest effects on the thermal resistivity. On the contrary, the minimum thermal resistivity values are related with the higher gravels content (80% weightgravelsweight1 sample). In fact, this phenomena can base it on the thin watewr film around the soil fine particles and coarse elements (Al Nakshabandi and Kohnke, 1964). The liquid phase caused a sensor reading close to thermal resistivity of the film water (Rwater 1.5 mKW1), due to increase of the macroporosity for increasing the coarse fragments in the sample. In this case, the realtionshiop could be an increase of the gas phase of the soil (Koorevaar et al., 1983) involving a rising thermal resistivity values for this gravel content. Figure 3, shows the relationship between the thermal resistivity and water content for a soil with a gravel content about 35% weightgravelsweight1sample. In this experiment two thermal sensors KS-1 was placed inside the soil column device. We chose two different positions respect to soil sample surface: vertical (perpendicular to top sample surface) and horizontal (parallel to top sample surface). The results indicated exactly the thermal sensor placed vertically shown divergences to integrate a
10 perpendicular parallel 8
R (mKW-1)
0 0 0.05 0.1 0.15 0.2 m3m-3 0.25 0.3 0.35
Figure 3. Relationship between thermal resistivity (R) and volumentric water content () for a silt loam soil. Sensor insertion sample in vertical (perpendicular position to the surface of the sample), and in horizontal (parallel position to the surface of the sample).
volume in depth. Monitoring the sample indicated that for the same water content value we obtained different thermal resistivity values as a funciton of the sensor positioning. The values always were higher for the vertical sensor. Both sensors coincide for a same value when the water content was about 20% volvol1. The differences are explained by the spatial interaction between the heat transfer and the sample moisture. Mostly important factors were the thickness and the geometry of the water film around the particles, when the watewr content was higher 15% volvol1. The sensor placed horizontally presented values where the water film was more homogeneous, meanwhile the verticl sensor distribbuted the heat pulse for several microscenarios, where occurred increasing and decreasing interactions of the liquid phase in contrast to the gas phase of the soil, presenting thus an increase of the thermal resistivity value. 4 CONCLUSIONS
As a summary, the soils with high coarse elements content present a variability in the thermal and hydric dynamic, being both well-related. High thermal properties variability was obtained in sample with 10% and 50% of coarse elements. On the other hand, for the same percentage of gravels the variability shown in the thermal and hydric dynamic was involved by the geometry and thickness of the water film around de particles.
599
The position of the thermal sensor inside the sample presented divergences between vertical and horizontal location. The diferences in the results were minimal for high water content and highest for scenarios whose water content was less than half average of field capacity water content for these kind of soils. REFERENCES
Al Nakshabandi G., Kohnke H., 1965. Termal conductivity and diffusivity of soils as related to moisture tension and other physical properties. Agricultural Meteorology, 2: 271279. Beven K., Germann P., 1982. Macropores and water flow in soils. Water Resour. Res., 18(5): 13111325. Bouma J., 1983. Use of soil survey data to select measurement techniques for hydraulic conductivity. Agric. Water Manage., 6: 177190. Campbell G.S., Jungbauer Jr. J.D., Bidlake W.R., Hungerford R.D., 1994. Predicting the effect of temperature on soil thermal conductivity. Soil Sci. 158:307313. Campbell G.S., Norman J.M., 1998. An Introduction to Environmental Biophysics, 2nd Ed. Springer Verlag, New York. DeVries D.A., 1963. Thermal properties of soils. In: W.R. van Wijk (Editor), Physics of plant environment. North-Holland Publishing Co., Amsterdam, pp. 210235. Dexter A., 1986. Model experiments on the behaviour of roots at the interface between a tilled seed-bed and a compacted sub-soil. Plant Soil, 95, 135147. Elbersen G., 1982. Mechanical replacement processes in mobile soft calcic horizons; their role in soil and landscape genesis in an area near Merida, Spain. Agricultural Research Report 919, ITC, Wageningen, 220 pp. Espeby B., 1989. Water flow in a forested till slope: Field studies and physically based modelling. Report
1052, Tritakut Series. Department of Land and water Resources. Royal Institute of Technology, Stockholm, 133. Flint, A. y S. Childs, 1984. Physical properties of rock fragments and their effect on available water in skeletal soils. In: D. Kral (Editor), Erosion and Productivity of Soils Containing Rock Fragments, 13. Soil Science Society America, Madison, WI, 91103. Garca-Rodrguez, A. (Editor), 1985. Estudio edifico de la provincia de Valladolid: mapa de suelos, escala 1: 100.000 de la zona situada al sur del rio Duero, 1. Consejo Superior de lnvestigaciones Cientificas. Salamanca, 206 pp. Ingelmo F., Cuadrado S., Ibafiez A. y Hernandez J., 1994. Hydric properties of some Spanish soils in relation to their rock fragment content: implications for runoff and vegetation. Catena, 23, 7385. Johansen O., 1975. Thermal conductivity of soils, Ph.D. Thesis, Trondheim, Norway. Koorevaar P., G. Menelik y C. Dirksen, 1983. Elements of soil physics. Develop. in soils science 13. Elsevier, New York, 230 p. Naidu A.D.y D.N. Singh, 2004. A generalized procedure for determining thermal resistivity of soils. International Journal of Thermal Sciences 43, 4351. Ochsner T.E., Horton R., Ren T., 2001. A new perspective on soil thermal properties. Soil Sci. Soc. Am. J., 65, 16411647. Singh D.N. y K. Devid, 2000. Generalized relationships for estimating soil thermal resistivity. Experimental Thermal Fluid Sci., 22, 133143. Vanpelt D.J., 1976. Thermal conductivity measurements of crushed stone and gravel aggregate, CRREL Technical Note, Unpublished. Villani M. y R. Wright, 1990. Environmental influences on soil macroarthropod behaviour in agricultural systems. Annu. Rev. Enthomol., 35, 249269. White R., 1985. The influence of macropores on the transport of disolved and suspended matter through soil. Adv. Soil Sci., 3, 89120.
600
Temperature effects on soil-water retention properties of densely compacted GMZ01 bentonite

Ye Wei-min
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University, Shanghai, China United Research Center for Urban Environment and Sustainable Development, The Ministry of Education, Shanghai, China
Wan Min & Chen Bao
Key Laboratory of Geotechnical and Underground Engineering of Ministry of Education, Tongji University, Shanghai, China
Yu-Jun Cui Wang Ju
Ecole des Ponts ParisTech, Paris, France Beijing Research Institute of Uranium Geology, Beijing, China
ABSTRACT: Gaomiaozi (GMZ01) bentonite, which originated from a deposit locates in Xinghe County, Inner Mongolia Autonomous Region, China, has been considered as the first choice for using as buffer/ backfill materials for deep disposal of high radioactive waste in China. For this purpose, densely compacted GMZ01 bentonite should present satisfactory thermal stability. In this paper, laboratory tests were conducted to obtain hydro-thermal behaviour of densely-compacted GMZ01 and its water retention properties were determined under both confined and unconfined conditions at temperature: 20, 40, 60 and 80C. The vapour equilibrium technique was employed for suction control. Results show that, i) the water retention capacity of the highly compacted bentonite under confined/unconfined conditions falls with heating; ii) for high suction (greater than 4 MPa), no difference was observed between the water retention curves of confined and unconfined specimen, while for low suction, the water-retention capacity of confined sample was found to be significantly lower than that of unconfined sample at same temperature; iii) the hysteresis of the unconfined compacted bentonite decreases as temperature increases; for high suctions (greater than 4 MPa), there is an increasing trend in hysteresis of unconfined bentonite when suction decreases. Keywords: Gaomiaozi bentonite, nuclear waste repository, soil-water retention properties, hysteresis, temperature 1 INTRODUCTION expansion of the working compacted bentonite, the highly-compacted bentonite experiences coupling of thermo-hydro-mechanical process. In order to ensure the safety and effective operation of highlevel radioactive waste repository, it is important to investigate the effect of temperature on hydro-mechanical characteristics of the compacted bentonite used as buffer materials[1]. Soil-water characteristic is one of the key indexes of hydro-mechanical property of bentonite. Many studies were conducted on the effect of temperature on soil-water characteristics of the highly-compacted bentonite[25]. ROMERO et al[6], and VILLAR and LLORET[7] experimentally obtained soil-water retention curves of the compacted BOOM clay and
In the design concept of multi-barrier system of deep geological disposal for high-level radioactive waste, the highly-compacted bentonite has been widely used for construction of artificial barrier. Its low hydraulic conductivity, micro-porous structure, good sorption properties and swelling capacity make this material an effective barrierprotecting the canister and restricting the movement of radionuclide released from the waste packages after canister failure (Wersin et al., 2007). Because of the radiogenic heat released from the waste container, the infiltration groundwater from the surrounding rock and the mechanical stress induced by the hygroscopic
601
FEBEX bentonite by different temperature controlled experiments, respectively. JACINTO et al[8] measured the retention curves of different dry densities of MX80 bentonite following the wetting path at different temperatures, and modified the van Genuchtens model by considering the effect of temperature on the water retention capacity. But, under the influence of gas generation and groundwater infiltration in the radioactive waste repository, the compacted bentonite will experience soil-water hysteresis caused by the drying wet circles in a certain range of suction[910]. So, studies on the soil-water hysteresis were focused on developing thermo-dynamic models of hysteresis for soil-water characteristic curves, such as the bounding surface model[1112], and the internal-variable models[13]. Based on the internalvariable theory of capillary hysteresis, XU et al [14] developed a mathematical model to describe the soil-water characteristics of geotechnical media experiencing arbitrary drying-wetting cycles. The effect of air entrapment was fully taken into account in XUs model. Gaomiaozi (GMZ) bentonite was selected as buffer materials for the construction of Chinese geological repository for high level radioactive nuclear waste[15]. CHEN et al[16] obtained the soil-water characteristic curves of highly-compacted GMZ bentonite using the osmotic technique and vapor equilibrium technique for suction control. YE et al[17] tested the swelling pressure characteristics of the GMZ bentonite using the constant volume test technique, and found that there was an exponential relationship between the swelling pressure and the dry density. QIN et al[18] analyzed the effects of the dry density, the vertical stress and the soaking liquid on the swelling characteristics of the GMZ bentonite. The objective of the present study is to investigate the temperature effect on the soil-water characteristics and hysteretic behaviors of highlycompacted GMZ bentonite under unconfined and confined conditions. 2 2.1 EXPERIMENTAL Materials
Table 1. Index
Properties of GMZ bentonite. Description 2.66 8.689.86 276 37 570 0.773 0 Na(0.433 6), Ca2(0.291 4), Mg2(0.123 3), K(0.025 1) Montmorillonite (75.4%), quartz (11.7%), feldspar (4.3%), cristobalite (7.3%)
Specific gravity of soil grain pH Liquid limit/% Plastic limit/% Total specific surface area/(m2g1) Cation exchange capacity/(mmolg1) Main exchanged cation/ (mmolg1) Main minerals
Table 2.
Initial state parameters of specimen. Value 6 20 10.65 1.7
Parameter Height/mm Diameter/mm Water content/% Dry density/(gcm3)
The initial state parameters of the specimen are listed in Table 2. 2.2 Apparatus
The properties of GMZ bentonite tested (Table 1) show that GMZ bentonite has a high cation exchange capacity, great adsorption ability, good dispersion and hydraulic properties. For preparation of the specimen, GMZ bentonite powder was compacted with a cylindrical mould. The compaction process was digitally controlled with a vertical velocity of 0.1 mm/min to ensure the specimen uniformity. Even though, it is difficult to achieve the optimum homogenization[19].
A steel cylindrical mould was employed for compaction of specimen with a universal stress. Selfdesigned cells with small holes in two ends were employed for suction control tests under confined conditions. The holes were designed as channels for moisture exchanges between specimen in the cell and the circulating air surround it. And an aluminous container was used for the unconfined test. The setup used for vapour phase technique suggested by YE et al[20], as shown in Fig. 1, was employed here for suction and temperature control test. Where, specimen was placed in the desiccator, and the saturated salt solution was used for suction control. The air pump was employed for keeping air in circulation. The whole system was placed in a digital oven for temperature control with a resolution of (0.1)C. 2.3 Methods and test procedure
2.3.1 Temperature control In order to investigate the influences of temperature on soil-water characteristics of the compacted GMZ bentonite, 20, 40, 60 and 80C were selected for temperature control in the tests.
602
28 24 20
Water content /%
16
12 8 4
20 40 60 80 40 60 80 (Unconfined) (Unconfined) (Unconfined) (Unconfined) (Confined) (Confined) (Confined)
Figure 1.
Setup of vapour phase technique.
0 1 10
100
1000
Table 3. Salt solutions and their corresponding suction at different temperature (MPa). Salt solution LiCl2 MgCl2 K2CO3 Mg(NO3)2 NaNO2 NaNO3 NaCl (NH4)2SO4 KCl ZnSO4 KNO3 K2SO4 0.2 molNaCl 20C 309.0 150.0 113.0 82.0 57.0 39.0 38.0 24.9 21.0 12.6 9.0 4.2 0.94 40C 162.4 122.0 103.1 49.5 40.6 32.2 27.8 5.1 1.023 60C 187.7 144.8 139.0 61.6 44.2 33.4 5.5 80C 219.5 167.4 69.0 48.0 38.4 7.5
Suction /MPa
Figure 2. Comparison of SWRCs in unconfined/ confined wetting paths at different temperatures.
2.3.2 Suction control The vapour equilibrium method was employed for the suction control. Different saturated salt solutions were used to control the relative humidity of the circulating air in the whole system. Their corresponding imposed suctions at 20, 40, 60 and 80C are shown in Table 3. Where, the high suction (4 MPa) were calibrated by TANG and CUI[21] and the low suctions, which controlled by unsaturated NaCl solutions, were suggested by LYTTON[22]. 3 3.1 RESULTS AND DISCUSSIONS SWRCs of densely-compacted GMZ bentonite
The SWRCs of the densely-compacted GMZ bentonite following wetting path at different temperatures under unconfined and confined conditions are shown in Fig. 2. Fig. 2 indicates that, under both confined and unconfined conditions, water retention capacity of compacted bentonite decreases with temperature increases, and the influence increases with suction decreases.
Mechanism of temperature influence on the water retention capacity of densely-compacted bentonite can be analyzed from the aspect of low suction and high suction, respectively[7]. For high suction (4 MPa), the change of clay fabric and intra-aggregate water plays a significant role in water retention capacity of GMZ bentonite. Much intra-aggregate water moves into the macropores with temperature increases. The excessive water flows out from the macropores, for degree of saturation is invariable in macropores at certain suction. Therefore, the water retention capacity of bentonite decreases as temperature increases. For low suction, capillary effect plays a decisive role in the performance of the water retention capacity. The changes of surface tension caused by the increase of temperature result in the increase of the water content of bentonite with increasing temperature at a certain suction. The effect of the constraint conditions on the water retention capacity of bentonite can also be analyzed in the aspect of the high and low suctions. At high suction, the influence of constraint conditions on the soil-water retention capacity of bentonite is not significant, which agrees with the result obtained by CHEN et al[16]. But at low suction, the water retention capacity of the confined specimen is clearly lower than that of unconfined specimen. The difference between water content of the compacted bentonite under the confined and unconfined conditions becomes greater with decreasing suction, which suggests that water retention capacity has a close relationship with the changes of the microstructure. Under the unconfined conditions, the interaggregate pores of bentonite expand constantly, and water can flow into these pores easily. Therefore, the water retention capacity of the unconfined specimen increases rapidly as the suction decreases. As clay particles can only swell and collapse into the
603
macropores under confined conditions, the increasing rate of water content in the confined specimen is significantly lower than that in the unconfined specimen with the decrease of the controlling suction. In order to analyze the influence of temperature on the water retention capacity, a ratio (kt), which indicates that the water content changes with temperature, is defined as kt wt1 wt2 wt1 100% (1)
wd ww 100% ww
(2)
where wt1 and wt2 are the water contents in the SWRCs at temperature t1 and t2 for a given suction, respectively. The relationship between kt and the suction in temperature range from 40 to 80C is shown in Fig. 3. Fig. 3 indicates that the influence of temperature on the water retention capacity has a close relationship with suction. Fig. 3 also shows that the impact of constraint conditions on kt tends to increase with suction decreases. 3.2 Hysteretic property of SWRC
where, wd and ww are the water contents in the drying and wetting paths at a given matric suction, respectively. From Eq. (2), it is very easy to find that the smaller the , the less the hysteretic behaviour. Using Eq. (2), the hysteretic coefficients of unconfined compacted GMZ bentonite changing with suction at 20, 40 and 60C were calculated respectively. Results (Fig. 4 (b)) show that the hysteretic coefficients are less than 7%, and decrease as the temperature increases, which indicates that the hysteresis of highly-compacted bentonite is not so significant. The hysteretic behavior of the unconfined compacted bentonite in different suction ranges at 40C is shown in Fig. 5. Fig. 5(a) shows that
25
(a)
The hysteretic characteristic curves for unconfined GMZ bentonite following the wetting/drying cycles at high suction at 20, 40 and 60C are shown in Fig. 4 (a). It is quite obvious that the hysteresis loop of SWRC at 20C is larger than that at 60C. That is to say, the hystereticbehaviour of the compacted bentonite decreases obviously as temperature increases. The hysteretic coefficient, , which was introduced to quantitatively describe the hysteretic behavior of bentonite of water retention curves following wetting/drying cycles, can be expressed as,
20
Water content /%
A2 B2 C2
15
10
5 1 10
Suction /MPa
100
1000
(a)
(b)
60 0. 32
Suction /MPa
40 4. 01 4. 76 20 6. 85
A2 B2 C2
0 0 1 2 3 4 5 6 7 8 9 10
/%
(b)
Figure 4. Hysteretic characteristics of unconfined compacted GMZ bentonite at high suction (4 MPa) and different temperatures (A2,B2,C2 stand for hysteretic cycle in suction range 4 to 50 MPa at 20,40 and 60C, respectively).
Figure 3. Relationship between kt and suction (t1 is 40C, and t2 is 80C).
604
25
(a) Water content /%

20
B1 B2
(4 MPa), the hysteretic behaviour of the unconfined bentonite tends to increase with decreasing suction. In summary, the hysteretic behaviour of compacted bentonite is not significant. ACKNOWLEDGEMENTS The study was supported by Projects (40772180, 40572161, 40728003) supported by the National Natural Science Foundation of China; Project ([2007]831) supported by China Atomic Energy Authority; Project (B308) supported by Shanghai Leading Academic Discipline Project of China. REFERENCES
[1] Wersin P., Johnson L.H., McKinley I.G. Performance of the bentonite barrier at temperatures beyond 100C: A critical review [J]. Physics and Chemistry of the Earth, 2007, 32(8/14): 780788. [2] Romero E., Villar M.V., Lloret A. Thermo-hydromechanical behaviour of two heavily overconsolidated clays [J]. Engineering Geology, 2005, 81(3): 255268. [3] Sanchez M. Thermo-hydro-mechanical coupled analysis in low permeability media [D]. Barcelona: The Polytechnic University of Catalunya, 2004. [4] Villar M.V., Lloret A. Dismantling of the first section of the FEBEX in situ test: THM laboratory tests on the bentonite blocks retrieved [J]. Physics and Chemistry of the Earth, 2007, 32(8/14): 716729. [5] Lloret A., Villar M.V. Advances on the knowledge of the thermo-hydro-mechanical behaviour of heavily compacted FEBEX bentonite [J]. Physics and Chemistry of the Earth, 2007, 32(8/14): 701715. [6] Romero E., Gens A., Lloret A. Temperature effects on the hydraulic behaviour of an unsaturated clay [J]. Geotechnical and Geological Engineering, 2001, 19(3/4): 311332. [7] Villar M.V., Lloret A. Influence of temperature on the hydro-mechanical behaviour of a compacted bentonite [J]. Applied Clay Science, 2004, 26(1/4): 337350. [8] Jacinto A.C., Villar M.V., Roberto G.E., Ledesma A. Adaptation of the van Genuchten expression to the effects of temperature and density for compacted bentonites [J]. Applied Clay Science, 2009, 42(3/4): 575582. [9] Alonso E.E. Gas migration through barriers [J]. Chinese Journal of Rock Mechanics and Engineering, 2006, 25(4): 693708. [10] Pham H.Q., Fredlund D.G., Barbour S.L. A study of hysteresis models for soil-water characteristic curves [J]. Canadian Geotechnical Journal, 2005, 42(6): 15481568. [11] Li X.S. Modelling of hysteresis response for arbitrary wetting/drying paths [J]. Computers and Geotechnics, 2005, 32(2): 133137.
15
10
5 1 10
Suction /MPa
100
1000
(a)
160 140 120
(b)
2. 03
B1 B2
Suction/MPa
100 80 60 40 20 0 0 1 2 3 4 4. 62 4. 76
/%
10
(b)
Figure 5. Hysteretic behaviors of unconfined compacted GMZ bentonite in different suction ranges at 40C (B1 stands for higher suction and B2 stands for lower suction).
the hysteretic cycle becomes larger with suction decreases. Fig. 5(b) also indicates that hysteretic coefficients are lower than 5% at high suction (4 MPa), and decrease with suction increases. 4 CONCLUSIONS
The water retention capacity of the highlycompacted GMZ bentonite decreases as the temperature increases. At the same time, the impact of temperature on the water retention capacity has a close relationship with the suction. At a certain temperature, the influence of constraint conditions on the water retention capacity of the compacted bentonite is not obvious at high suction, but the water retention capacity of confined specimen is significantly lower than that of unconfined specimen at low suction. The hysteretic behaviour of the compacted bentonite under unconfined conditions decreases as the temperature increases. At high suction
605
[12] Nuth M., Laloui L. Advances in modeling hysteretic water retention curve in deformable soils [J]. Computers and Geotechnics, 2008, 35(6): 835844. [13] Liu Y, Zhao C.G. Modeling the hysteresis for soil-water characteristic curves [C]. New Frontiers in Chinese and Japanese Geotechniques. Beijing: China Communications Press, 2007: 383392. [14] Xu Yan-bing, Wei Chang-fu, Chen Hui, Luan Mao-tian. A model of soil-water characteristics for unsaturated geotechnical materials under arbitrary drying/wetting paths [J]. Chinese Journal of Rock Mechanics and Engineering, 2008, 27(5): 4245. (in Chinese) [15] Liu Yue-miao, Wen Zhi-jian. Study on clay-based materials for the repository of high level ratioactive waste [J]. Journal of Mineralogy and Petrology, 2001, 28(12): 1921. (in Chinese) [16] Chen Bao, Qian Li-xin, Ye Wei-min, Cui Yu-jun, Wang Ju. Soil-water characteristic curves of Gaomiaozi bentonite [J]. Chinese Journal of Rock Mechanics and Engineering, 2006, 25(4): 788793. (in Chinese) [17] Ye W.M., Schanz T., Qian L.X., et al. Characteristics of swelling pressure of densely compacted
[18]
[19] [20]
[21]
[22]
Gaomiaozi bentonite GMZ01 [J]. Chinese Journal of Rock Mechanics and Engineering, 2007, 26(S2): 28613865. (in Chinese) Qin Bing, Chen Zheng-han, Liu Yue-miao, Wang Ju. Swelling-shrinkage behaviour of Gaomiaozi bentonite [J]. Chinese Journal of Geotechnical Engineering, 2008, 30(7): 10051010. (in Chinese) Cui Y.J., Delage P. Yielding and plastic behavior of an unsaturated compacted silt [J]. Geotechnique, 1996, 46(2): 291311. Ye Wei-min, Tang Yi-qun, Cui Yu-jun. Measurement of soil suction in laboratory and soil-water characteristics of Shanghai soft soil [J]. Chinese Jounal of Geotechnical Engineering, 2005, 27(3): 347349. (in Chinese) Tang A.M., Cui Yu-jun. Controlling suction by the vapour equilibrium technique at different temperatures and its application in determining the water retention properties of MX80 clay [J]. Canadian Geotechnical Journal, 2005, 42: 110. Lytton L. Isothermal water movement in clay soils [D]. Austn: University of Tekas, 1967.
606
Testing techniques: Laboratory, in situ and centrifuge
Evaluation of filter paper calibrations for indirect determination of soil suctions of unsaturated soils
K.V. Bicalho
Federal University of Espirito Santo, ES, Brazil
F.A.M. Marinho
University of So Paulo, SP, Brazil
J.-M. Fleureau & A. Gomes Correia
Ecole Centrale Paris & CNRS, Chtenay-Malabry, France University of Minho, Guimares, Portugal
ABSTRACT: This paper evaluates the use of several published calibration curves (Fawcett & Collis-George 1967; Hamblin 1981; Chandler & Gutierrez 1986; Chandler et al. 1992; ASTM Standard D 5298; Leong et al. 2002, and Oliveira & Marinho 2006) for the Whatman 42 filter paper for indirect laboratory estimation of soil suctions. Evaluation of the filter paper calibrations was carried out using the experimental results given by Fleureau et al. (2002) obtained with other techniques used to measure or control the soil suctions in a compacted silty sand. Significant discrepancy exists among the calibrations that are commonly used for determining suction using the gravimetric water content of the filter paper data. The FPM offers a simple technique for the determination of soil suction, provided that an adequate calibration curve is used for the investigated suction range. 1 INTRODUCTION 2 FILTER PAPER METHOD (FPM)
The experimental techniques commonly used for measuring or controlling soil suctions vary widely in terms of cost, complexity, and measurement range. The soil suctions can be determined from previous calibration or can be measured directly. Because of the various difficulties involved in the direct suction measurements, a simple and economical laboratory method for measuring suctions, even if a degree of approximation is involved, is of considerable value. In this paper, the contact filter paper method is used as an indirect method of estimating matric suctions of an unsaturated compacted silty sand. The matric suction values inferred from filter paper measurements depend on a calibration between the water content of the filter paper and suction. Therefore, various calibration curves proposed at the literature (Fawcett & Collis-George 1967; Hamblin 1981; Chandler & Gutierrez 1986; Chandler et al. 1992; ASTM D 5298; Leong et al. 2002; and Marinho & Oliveira 2006) for the Whatman 42 filter paper are used to estimate the suctions of an unsaturated compacted silty sand of known suctions. A modified calibration function, which gives better estimation of the measured suctions, is suggested.
Gardner (1937) was the first to introduce calibrated filter paper as an indirect means of determining the suction in soils. Since then, many researchers have been involved in the use of filter paper for estimating soil suctions (Fawcett & Collis-George 1967; Al-Khafaf & Hanks 1974; Hamblin 1981; Chandler & Gutierez 1986; Greacen et al. 1989; Chandler et al. 1992; Ridley 1993; Marinho 1994; Houston et al. 1994; Leong et al. 2002; Marinho & Oliveira 2006; Bulut & Leong 2008). The filter paper method calculates the soil suction indirectly from predetermined calibration. Basically, the filter paper comes to equilibrium with the soil either through vapor (total suction measurement) or liquid (matric suction measurement) flow. At equilibrium, the filter paper and the soil will have the same suction value. After equilibrium is established between the filter paper and the soil, the gravimetric water content of the filter paper disc is measured. The gravimetric water content of filter paper is converted to suction using a predetermined calibration curve for the type of paper used. This is the basic approach suggested by the American Society for Testing and Materials (ASTM) standard D5298 for the measurement of either matric suction using the contact filter paper
609
technique or total suction using the non-contact filter paper technique. The ASTM D 5298 employs a single calibration curve that has been used to infer both total and matric suction measurements and recommends the filter papers to be initially ovendried (16 h or overnight) and then allowed to cool to room temperature in a desiccator. The ASTM D 5298 calibration curve is a combination of both wetting and drying curves. However, because of the marked hysteresis on wetting and drying of the filter paper, the calibration curve for initially dry filter paper is different from that of the initially wet filter paper. Some publications presents calibration for the wetting path, with the paper initially air dry (Chandler & Gutierrez 1986; Chandler et al. 1992; Ridley 1993; and Marinho 1994). The contact filter paper technique is used for measuring matric suction of soils. In this technique, water content of an initially dry filter paper increases due to a flow of water in liquid form from the soil to the filter paper until both come into equilibrium. Therefore, a good contact between the filter paper and the soil has to be established. The contact filter paper method becomes inaccurate in high matric suction range since water transport is dominated by vapour transport (Marinho & Chandler, 1993; Fredlund et al., 1995). 2.1 FPM calibration curves
Table 1. Calibrations curves for Whatman 42 filter paper. Suction type w (%) range (suction) Log10 (suction) (kPa) (KPa)
References ASTM D5298
The calibration curve for the filter paper matric suction measurement is commonly established using a pressure plate apparatus (e.g., Al-Khafaf and Hanks 1974; Hamblin 1981; Greacen et al. 1989). It is important to note that only ash-less filter papers should be used in the filter paper technique. Although there are several ash-less filter papers available, only Whatman 42 and Sleicher and Schuell 59 (or SS 59) are commonly used. A number of calibration functions for Whatman. 42 filter papers have been published in the literature. The functions share a number of similarities, allowing them to be written in a general form as: Log10 (suction) (kPa) A B w (%) (1)
Total and w 45.3 5.3270.0779 w Matric (suction 62.8) 2.4120.0135 w w 45.3 (suction 63.2) Chandler Matric 4.8420.0622 w w 47 et al. (suction (1992) 82.9) 6.052.48 Logw w 47 (suction 80) Leong Matric 4.9450.0673 w w 47 et al. (suction (2002) 60.5) 2.9090.0229 w w 47 (suction 68) Total 8.7780.0222 w w 26 (suction 1058) 5.310.0879 w w 26 (suction 1014 ) Marinho & Total and w 33 4.830.0839 w Oliveira Matric (suction (2006) 115) 2.570.0154 w w 33 (suction 115) Bicalho Matric 4.750.048 w 36 w 50 et al. (220 suc(2009) tion 1000) 3.3650.027 w 55 w 50 (80 suction 220)
where w is the gravimetric water content of the filter paper at equilibrium. Chandler and Gutierrez (1986) presented a calibration curve for Whatman No. 42 filter paper for suctions in the range of 80 kPa to 6000 kPa that included their own results and also those from Fawcett and Collis-George (1967) (i.e., A 5.777 and B 0.06) and Hamblin (1981) (i.e., A 6.281 and B 0.0822), therefore, the obtained calibration curves are similar with obtained A 5.85 and B 0.0622. Table 1 lists some calibrations presented in the literature for the filter paper Whatman. 42 with an
inflection point occurring at a filter paper gravimetric water content value somewhere between 33 and 47% (corresponding 115 kPa suction 60 kPa). The calibration curves proposed by Chandler et al. (1992), ASTM Standard D 5298 and Leong et al. (2002) Matric suctions are similar with A in Eq. (1) ranging from 4.842 (Chandler et al 2002) to 5.327 (ASTM D5298) and B ranging from 0.0622 (Chandler et al. 1992) to 0.0779 (ASTM D5298). Figure 1 shows calibrations curves for proposed by Fawcett & Collis-George (1967), Hamblin (1981), Chandler & Gutierez (1986), Chandler et al. (1992), ASTM D 5298, Leong et al. (2002) and Marinho & Oliveira (2006) for w values 50%. A similar agreement can be seen in the suctions derived using the curves proposed by Chandler et al. (1992), ASTM D 5298 and Leong et al. (2002)Matric suctions.
610
range (high potential) and generally requires only a few days. In contrast, vapor equilibration is slow in the wet range because a large amount of water needs to be transferred. Thermal equilibration is also important. Temperature gradients in the sample can result in liquid flow. In addition, temperature gradients can result in large errors when vapor exchange is used for equilibration. 3 3.1
Figure 1. 42 paper. Published calibration curves for Whatman
TYPE MATERIAL AND METHODS Material
Considerable variability is observed between their results and those of Fawcett and Collis-George (1967), Hamblin (1981) and Chandler & Gutierrez (1986) (which seem to overestimate the values of suction). The calibration proposed by Marinho & Oliveira (2006) is for a specific batch and cannot be directly compared. Although Leong et al. (2002) suggested the use of different calibration curves for matric and total suction, caution is recommended when using published total suction calibration curves since such curves are expected to be valid only for the equalization time used during the corresponding calibration. If the equilibrium between the filter paper and the soil has not yet been achieved, the total suction calibration curve might give total suction estimations smaller than corresponding matric suction estimations, yielding an unrealistic negative value of osmotic suctions. Walker et al. (2005) and Marinho & Oliveira (2006) suggest that the filter paper calibration is unique in relation to the type of suction (i.e., total or matric). Bulut & Wray (2005) recommend a single calibration curve based on water vapor measurements for both total and matric suction determinations. Even though, Hamblin (1981) did not observed significant difference between batches of filter paper produced at different times, Likos & Lu (2002) and Marinho & Oliveira (2006) have shown that the filter paper calibration curves can significantly vary among the same type of filter paper from one batch or lot to another. They recommend batch-specific calibrations. The non-contact filter paper technique for estimating total suctions must be performed with extra cares to avoid suction errors induced by temperature gradient, relative humidity error, and equilibrium time. It is recommended to allow the filter papers to equilibrate for a sufficient time period. Liquid phase equilibration is fairly rapid in the wet
The tested material is a residual silty sand, hereafter called Perafita sand, formed by the weathering of granite, which has been used as a building material for a road in the north of Portugal. It contains about 20% of grains smaller than 80 m, with a layered structure similar to that of clay particles. The liquid limit of the Perafita sand is 32.6%, the plastic limit is 25%, clay fraction is 2.5%, specific gravity is 2.66, standard Proctor optimum water content is 17.6% and the corresponding dry density is 16.8 kN/m3, modified Proctor optimum water content is 13.2% and the corresponding dry density is 18.6 kN/m3. 3.2 Test program
The preparation procedure of samples is the same for all the tests: the soil is sieved to avoid the presence of coarse grains (maximum size 4.75 mm), then it is mixed up with the right quantity of water; after that, it is placed in a sealed plastic bag for 24 hours to allow the hydric equilibrium to establish at a zero vertical stress condition (Chandler & Gutierrez 1986; Chandler et al. 1992). The contact filter paper tests were carried out on soil specimens compacted to the Modified Proctor Optimum water content (13.2%) and nearly maximum density (18.6 kN/m3) following the drying path (degree of saturation 85%). The compacted soil specimen sizes were 102 mm in diameter and 23.35 mm high. The test procedure involves placing a piece of initially air dry filter paper against the compacted soil specimen whose matric suction is required and sealing the whole to prevent evaporation. The filter paper then wets up to a water content in equilibrium with the magnitude of the soil matric suction, and careful measurement of the water content of the filter-paper enables the soil matric suction to be obtained from a previously established correlation. This provides a measure of the matric suction, which is assumed to be the same numerically as the capillary pressure (the reference being the atmospheric pressure). The Whatman 42 filter paper was used in all tests. The other techniques used to measure
611
or control the negative pore water pressure in the compacted soil specimens are not discussed in this paper since the purpose herein is to discuss the filter paper technique only. Details of the experimental techniques are given in Fleureau et al. (2002). 4 TESTS RESULTS AND ANALYSIS
The measured suctions of compacted Perafita sand specimens resulting from several methods used by Fleureau et al. (2002) and contact filter paper tests investigated in this paper are plotted versus degree of saturation in Figure 2. The term matric suction is used to indicate the negative pressure of water relative to atmospheric air pressure, i.e.(uw uatm). In order to verify the effect of the filter paper calibration curves on the contact filter paper method for matric suction measurement, the calibration curves proposed by Chandler et al. (1992), and ASTM D5298 are used to interpret the measured contact filter paper gravimetric water contents (w). Although it was observed a general agreement between the FPM test results using the calibration curves ASTM D 5298 and Chandler et al. (1992) and other techniques used to measure or control suctions in the compacted soil specimens for 100 kPa suction 300 KPa, the calibration curves overestimated the suctions for suction 300 kPa. The very wide interval presented in Figure 2 may indicate that more data should be collected before anything very definite can be said about the calibration function. Assuming a linear relationship between suctions (logarithmic scale) and degree of saturation, S, (suction 1537 exp (0.03 S)) based on the correlation coefficient criterion (R2 0.92) using the measured soil suctions (Fleureau et al. 2002) a modified calibration function for the Whatman 42 filter paper is determined by curve fitting to the experimental results (Bicalho et al. 2009). The suggested calibration curve for esti-
Figure 3. A pair of 80% confidence intervals (upper and lower limits) calculated from two calibration functions.
mating of soil suctions in the range of 80 kPa to 1000 KPa for the experimental data presented by Fleureau et al. (2002) is: For 36% w 50% Log10 (suction) (kPa) 4.75 0.048 w For 55% w 50% Log10 (suction) (kPa) 3.365 0.027 w (2a) (2b)
Equation 2 is specific to the tested filter paper, soil and suction ranges and has not been tested in other configurations of measurement system. A confidence interval gives an estimated range of values which is likely to include an unknown population parameter, the estimated range being calculated from a given set of sample data. The level of a confidence interval gives the probability that the interval produced by the method employed includes the true value of the parameter. Figure 3 shows a pair of 80% confidence intervals (upper and lower limits) calculated from each calibration line, but varies from calibration line to calibration line, although obtained under the same experimental conditions. The results presented in Figure 3 are obtained for the calibration functions proposed by Bicalho et al. (2009) and ASTM D 5298 and the measured data (Fleureau et al. 2002). The data suggest that the predicted suctions using calibration curve proposed by ASTM D 5298 increase significantly when measured suctions are greater than 300 kPa. Therefore, the calibration curve used for determining suction using the gravimetric water content of the filter paper data needs to be verified before applying the filter paper method. 5 CONCLUSIONS
Figure 2. Effect of the filter paper (FPM) calibrations on the derived soil suctions for Perafita sand.
Many empirical calibration equations have been proposed to calculates soil suction indirectly by
612
measuring the gravimetric water content of the filter paper at equilibrium, but none of them perform well in a wide range of circumstances and for all soil types. Significant discrepancy exists among the published calibrations that are commonly used for determining suction using the gravimetric water content of the filter paper data. The deviation among the calibration curves proposed by Chandler et al. (1992), ASTM D 5298, Leong et al. (2002)Matric suctions and Marinho & Oliveira (2006) decreased at suctions less than about 60 kPa. Although it was observed a general agreement between the FPM test results using the calibration curves ASTM D 5298 and Chandler et al. (1992) and other techniques used to measure or control suctions in the compacted soil specimens for 100 kPa suction 300 KPa, the calibration curves overestimated the suctions for suction 300 KPa. Calibration curves proposed by Fawcett & Collis-George (1967), Hamblin (1981) and Chandler & Gutierrez (1986) overestimated the values of suction. The FPM offers a simple technique for the determination of soil suction, provided that an adequate calibration curve is used. ACKNOWLEDGMENTS The first author is grateful for sponsorship by the Brazilian government agencies CAPES and CNPq. REFERENCES
Al-Khafaf, S. & Hanks, R.J. 1974. Evaluation of the filter paper method for estimating soil water potential. Soil Sci. 117, 194199. ASTM Standard D5298-03, 2007, Standard Test Method for Measurement of Soil Potential (Suction) Using Filter Paper, Annual Book of ASTM Standards, Soil and Rock (I), Vol. 4, No. 8, ASTM International,West Conshohocken, PA. Bicalho, K.V., Nunes, G.W., Marinho, F.A.M. Fleureau, J.-M., Gomes Correia, A. & Ferreira, S. 2009. Evaluation of filter paper calibrations for laboratory estimating of soil suctions, Proc. 4th Asia Pacific Conference on Unsaturated Soils, Australia, 215220. Bulut, R. & Wray, W.K. 2005. Free energy of water suction in filter papers. Geotech Test. J., 28: 4, 355364. Bulut, R. & Leong, E.-C. 2008. Indirect Measurement of Suction. Geotech. Geol. Eng, 26: 6. Chandler, R.J., Crilly, M.S. & Montgomery-Smith, G. 1992. A low-cost method of assessing clay desiccation for low-rise buildings. Proc. Instn. Civ. Engrs Civ. Engng. 92, May, 8289. Chandler, R.J. & Gutierrez, C.I. 1986. The filter paper method of suction measurement. Geotechnique 36, No. 2, 265268. Fawcett, R.G. and Collis-George, N. 1967. A filter paper method for determining the moisture characteristics of soil. Amt. J. Exp. Agric. Anim. Husb. 7, 162167.
Fleureau, J.-M., Hadiwardoyo, S., Dufour-Laridan, E., Langlois, V. & Gomes Correia, A. 2002. Influence of suction on the dynamic properties of a silty sand. Unsaturated Soils, Juc, de Campos and Marinho (eds), Sweets and Zeitlinger, Lisse, 2, 463471. Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics for unsaturated soils. Wiley & Sons, New York. Gardner, R. 1937. A method of measuring the capillary tension of soil moisture over a wide moisture range. Soil Sci. 43, 277283. Greacen, E.L., Walker, G.R. & Cook, P.G. 1989. Procedure for the Filter paper method of measuring soil water suction, Division of soils, Report 108,CSIRO Division of Water Resources, Glen Osmond, Australia. Hamblin, A.P. 1981. Filter paper method for routine measurement of field water potential. J. Hydrol. 53, 355360. Houston, S.L., Houston, W.N., Wagner, A. 1994. Laboratory filter paper suction measurements, Geotechnical Testing Journal, 17, 2, 185194. Lee, H.C. & Wray, W.K. 1995. Techniques to evaluate soil suction a vital unsaturated soil water variable. 1 International conference on unsaturated soils, Paris, 615622. Leong, E.C., He, L. & Rahardjo, H. 2002. Factors Affecting the Filter Paper Method for Total and Matric Suction Measurements, Geotech. Test. J., Vol. 25, No. 3, pp. 322332. Likos, W.J. & Lu, N. 2002. Filter paper technique for measuring total soil suction. Transportation Research Record 1786, Paper No. 022140, 120128. Lu, N. & Likos, W.J. 2004. Unsaturated soil Mechanics, John Wiley and Sons, 556 p. Marinho, F.A.M. & Chandler, R.J. 1993. Aspects of the Behaviour of Clays on Drying In: Unsaturated Soils, Dallas. ASCE National Convention. ASCEAmerican Society of Civil Engineers, 1993. Vol. 39. p. 7790. Marinho, F.A.M. 1994. Shrinkage behavior of some plastic clays. PhD Thesis, Imperial College, University of London. Marinho, F.A.M. & Oliveira, O.M. 2006. The filter paper method revised. ASTM geotechnical testing journal, USA, 29 (3), 250258. Marinho, F.A.M., Take, A. & Tarantino, A. 2008. Measurement of Matric Suction Using Tensiometric and Axis Translation Techniques. Geotech Geol Eng, 26:6. McKeen, R.G. 1980. Field studies of airport pavements on expansive clay. Proc. 4th Int. Coand Expansive Soils 1, 242261. McQueen, I.S. & Miller, R.F. 1968. Calibration and evaluation of a wide-range gravimetric method for measuring moisture stress. Soil Sci. 106, 225231. Ridley, A.M., Dineen, K., Burland, J.B. & Vaughan, P.R. 2003. Soil matrix suction: some examples of its measurement and application in geotechnical engineering. Gotechnique 53, No. 2, 241253. Ridley, A.M. 1993. The measurement of soil moisture suction. PhD thesis, University of London. Ridley, A.M., & Wray, W.K. 1996. Suction measurement A review of current theory and practices. Proc., 1st Int. Conf. on Unsaturated Soils, 12931322. Walker, S.C., Gallipoli, D. & Toll, D.G. 2005. The effect of structure on the water retention of soil tested using different methods of suction measurement. Proc., Int. Sym. on Adv. Exp. Unsaturated Soil Mechanics, Trento, 3339.
613
Useful laboratory guidelines for unsaturated soils research

J.L. Boyd
University of Ulster, UK
V. Sivakumar
Queens University Belfast, UK
ABSTRACT: Many geotechnical research studies are conducted in a progressive manner, with previous research aiding the present researcher and the present researcher guiding future research. No matter how much literature exists in a research area, there will still be challenges for new researchers to deal with. This is true in any field, but especially in the case of unsaturated soils research, which is intrinsically complex and it demands high levels of accuracy. This paper presents challenges faced when developing a laboratory research programme into the behaviour of unsaturated compacted clays under K0 conditions. The paper reports on the difficulties encountered in the research and how they were rectified or resolved. Specific attention is given to the fabrication of suction control mechanism using axis translation technique and the importance of basic material characterization of soil prior to any research. 1 INTRODUCTION research programme into the stress regime in unsaturated soils when laterally confined. 2 MATERIAL SELECTION
If we knew what we were doing, it would not be called research would it? (Albert Einstein). Einstein emphasises the fact that the process of research involves facing many challenges and overcoming obstacles. Many if not all research projects are carried out in a progressive manner. Past experience helps the present researcher, and present research guides the future researcher. However, it is often the case that regardless of the information collected by previous researchers, a new researcher will have to tackle a renewed set of problems. On this note, the research reported in this paper is not an exception. This paper focuses on experimental problems encountered and the resolution of these problems. The problems faced are associated with suction control, and the physical characteristics of the testing material. The successful resolution of these issues permitted the execution of an experimental programme into the stress regime of unsaturated compacted clay under laterally confined conditions. This paper will be of benefit to new researchers in the area of unsaturated soils and also those established researchers who have suffered similar problems. Compacted clay fills are placed in an unsaturated condition and over time the fill will become saturated if the water table rises, which can considerably change the stress regime in the fill. The evaluation of this stress regime in unsaturated soils requires careful experimental procedures to be followed in the laboratory. This paper presents some of the difficulties encountered during the current laboratory
The selection of materials is very important in any research, as this is a critical factor when producing repeatable and meaningful data. Kaolin has been used by many researchers in unsaturated soils research. Kaolin exhibits high consolidation rates compared to many other clays and the material is generally uniform and readily available from commercial suppliers. Part of the research involved comparing the present data with data previously produced by the research team at Queens University Belfast (Sivakumar et al, 2010). However, this objective turned out to be unattainable, when a careful characterisation of the material was conducted. Sivakumar et al. (2010) produced specimens with an initial specific volume of 2.193 when isotropically compressed under 400 kPa at an initial water content of 25%, whilst the current researcher using the same initial conditions produced specimens with an initial specific volume of 2.125. This difference was unexpected, as the speswhite kaolin used in Sivakumar et al. (2010) and this research, were supplied by the same manufacturer and the clays were supplied to the same specifications. The different specific volumes highlighted the fact that something was different between the two batches of kaolin. Very similar responses were observed among the research carried out by Sivakumar (1993), Sivakumar and Wheeler
615
(2000), Tan (2004), Thom (2007), and Sivakumar et al. (2010). These researchers used kaolin supplied at various times in the past 15 years. The specimens were tested in a similar fashion to that reported by Sivakumar (1993), Sivakumar and Wheeler (2000), Tan (2004), Thom (2007), and Sivakumar et al. (2010). The crucial difference in the response of the material, apart from the initial specific volume was the water volume change behaviour when the suction was altered in the specimen. This implied that the material used in the present research had different soil water characteristics when compared to the materials used in the previous research. Figure 1 shows data reported by Sivakumar et al. (2010) and that obtained in the present research respectively, when the suction was reduced from the initial value (just after compaction) to 100 kPa under 50 kPa of mean net stress. The change in specific water volume reported by Sivakumar et al. (2010) was 0.275, which is substantially greater than the increase of 0.126 recorded in the present research, yet the two samples were prepared at identical water contents and compression pressures. Further examinations were performed in order to examine the particle shape and structure (SEM), the particle size distribution (PSD) and the mineralogy composition. The PSD was established using the laser diffraction method and the results with reference to the material used in the present research and the previous research are shown in Figure 2. The PSD confirms that the present research material contained only 40% clay sized particles, and it is significantly lower than the percentage of clay fraction, 80%, quoted by the supplier. The PSD was repeated on several samples from the same batch, although no significant variation was found. Figure 2 shows a PSD performed on kaolin used by previous researchers at QUB, which confirmed that the material contained 85% clay sized particles, closely agreeing with the manufacturers specification. Scanning Electron Microscopy (SEM) images were taken on specimens of material used by Sivakumar et al. (2010) and specimens used in the present research, which are illustrated in Figures 3(a)
0.30 Increase in Specific Water Volume vw 0.25 0.20 0.15 0.10 0.05 0.00 0 Sivakumar et al. (2010) Current Research 5 10 15 20 Square root of time (Hours)1/2 25
Figure 2.
Particle size distribution for materials.
Figure 3. SEM imagery of material (a) Sivakumar et al. (2010) and (b) present research.
Figure 1. time.
Increase in specific water volume against root
and (b) respectively. The figures show that both specimens are composed of platey particles ranging in size from 0.5 m to 3 m. In general the material used by Sivakumar et al. (2010) appears to have a greater proportion of small particles (those less than 1 m) when compared to the material used in the present research. In general the studies have shown that the material used in the present study was significantly different in terms of particle size distribution from that used by previous researchers. Therefore any
616
reliable comparison of data obtained from the present research with Sivakumar et al. (2010), or any of the other previously published research (Sivakumar, 1993; Sivakumar & Wheeler, 2000; Tan, 2004; Thom, 2007) became unattainable, however it highlighted the importance of soil characterisation before starting any major research programmes. 3 HIGH AIR ENTRY FILTER PERFORMANCE
The axis translation technique (Hilf, 1956) was used for controlling suction. The high air entry filter (HAEF) and the arrangement by which it was attached to the pedestal seriously affected the water flow in and out of the specimen. Several tests were affected by either cavitation of the drainage lines or an unusually small water flow into the specimen during wetting. In addition several HAEFs fractured after a single use. 3.1 Best practice for pedestal arrangement
The layout of the pedestal is shown in Figure 4. The specimen was placed on top of the HAEF which was contained within a stainless steel ring. The ring was then secured to the pedestal using counter-sunk screws. The HAEF was cut to the required size to fit into the ring. The filter was cut using a standard lathe, and coolant was used to prevent the cutting tool overheating. This coolant contained a percentage of oil based materials, which clogged the pores of the HAEF, in turn affecting the flow characteristics of the filter, therefore the use of dry cutting was recommended and it significantly improved the performance of the HAEF. The HAEF disk was glued inside the ring using slow hardening araldite. When the araldite had
High Air Entry Filter Disc Counter sunk screws Possible cavity location Drainage lines (a) O-Ring (i) (ii) (b) (c) Drainage cross Slow hardening araldite O-Ring Drainage cross
hardened, it was essential to level the surface of the stone and the glue as shown in Figure 4(b)(i). Failure to do so (see Figure 4(b)(ii)) could lead to the specimen not being in contact with the HAEF. This could create a cavity between the specimen and the HAEF, thus making the water phase discontinuous. The sealing arrangement of the HAEF disk was also very important. Figure 4(a) shows a cross-section of the pedestal arrangement. The initial suction in the sample could be as high as several hundred kilopascals. Therefore when the specimen was in contact with the HAEF disk, it could pull water out of the disk and subsequently from the drainage lines. If the drainage lines were closed, as during setting-up, this process could lead to a drop in water pressure and eventually cavitation of the drainage lines and the pores of the HAEF. This problem could be mitigated by placing a small loop of thin wire between the specimen and the HAEF (Sivakumar, 1993). This wire was pushed into the soil when the initial confining pressures were applied to the specimen. A possible danger of using this method was that when the sample was very stiff a significant confining pressure would be required before the wire was pushed into the specimen. Another technique that may be useful for preventing cavitation of the drainage lines during setting up was that of opening the drainage lines to a source of water supply under a small amount of pressure (approximately 20 kPa) for a period of five seconds and repeating this every 20 minutes over the setting-up period. The disadvantage of this approach was that a small but unknown amount of water flowed into the specimen during setting-up. The ring containing the HAEF was sealed to the pedestal using an o-ring as shown in Figure 4(a). The o-ring had to be large enough to create an adequate seal against leakage. At the same time if the o-ring was too thick, it led to the formation of a thin film of air below the HAEF (this has been exaggerated in Figure 4(a)). When air became trapped below the HAEF it could not be flushed out using standard procedures. In several tests samples were set up and subjected to 400 kPa of initial suction. However the movement of water into the sample during the equalisation process was limited or significantly lower than anticipated. In an attempt to save these tests the externally applied suction was reduced (by elevating the pore water pressure) to significantly lower values (less than 100 kPa). When the lower suction values were used water began to flow into the sample. The substantial increase in water pressure compressed or dissolved the thin film of air possibly trapped below the HAEF. 3.2 High air entry filter properties Despite careful experimental procedures to mitigate the cavitation of the filter and the drainage lines, the problem of reduced water flow into the sample during the initial equalisation stage continued to be a
Figure 4. Pedestal arrangement (a) cross-section (b) (i) well levelled disk surface (ii) poorly levelled disk surface (c) pedestal plan view.
617
problem. In some cases the HAEF also fissured or fractured after a single usage. It was suggested that the HAEF used in the present study was somewhat different to the filters used in previous research programmes. A new batch of HAEFs was purchased, which enabled a comparison to be made between the original and new HAEFs in terms of water flow behaviour. Figure 5 shows the increase in specific water volume plotted against time during the wetting process for two tests, where one filter disk was manufactured from the HAEF purchased at the beginning of the research and the other from the new supply. The specimens were prepared under the same initial conditions and were wetted under a suction of 200 kPa. The figure shows a substantial difference exists in the water flow behaviour. The increase in specific water volume of the specimen when using the original HAEF disk was 0.03, which increased to 0.12 when the new filter was employed. In addition, the rate of water flow into the specimen using the original HAEF was substantially lower than that using the new HAEF. These observations confirm that the original supply of HAEFs did not have air entry values of 5bar, as claimed by the supplier, and could be as low as 2bar. The quality of the saturation chamber was suspected to influence the observed fissuring and fracturing of the HAEFs (see Figure 6). The original saturation chamber was made of aluminium, which was not anodised. The surface of this chamber became heavily corroded over time. It was suspected that the aluminium oxide, which was forming on the surface of the cell, slowly permeated the HAEFs and changed the HAEFs chemical composition. Samples were taken from the fissured HAEFs and the residue collecting on the surface of the chamber. The samples were analysed for evidence of a chemical reaction between the HAEFs and aluminium oxide using X-ray Fluorescence (XRF). The data from the testing showed that aluminium was present in the residue coating the cell; however no aluminium traces were found in the HAEF. The reason for the observed fissuring remains unresolved, though the problem never reoccurred when the HAEFs
Figure 6. disk.
Fissured and fractured high air entry filter
were saturated in a new saturation chamber, which was anodised before use. 4 RECOMMENDATIONS
This paper reported some of the difficulties faced when executing a laboratory testing programme that examined the stress regime in compacted soil during wetting under laterally restrained conditions. The main difficulties related to the proper functioning of the high air entry filter and the material characterisation. The following conclusions are made: Material characterisation is very important prior to any research if the proposed research aims to corroborate or use findings previously reported on the subject. The construction of the pedestal arrangement that incorporates the high air entry filter and the proper saturation of the filter are important. Points to consider include; the method of cutting the filter to size; the sealing arrangement; the attachment of the filter to the metal ring; and saturation of the filter, including the quality of the saturation chamber. ACKNOWLEDGEMENTS The research presented in this paper was supported by the European Social Fund. REFERENCES
Adams, B.A., Wulfson, D., and Fredlund, D.G. (1996) Air volume change measurement in unsaturated soil testing using a digital pressure-volume controller. Geotechnical Testing Journal, 19(1), 1221. Blatz, J.A., and Graham, J. (2000) Technical Note: A system for controlled suction in triaxial tests. Gotechnique, 50(4), 465469.
Figure 5. Specific water volume against time for original and new HAEF.
618
Bishop, A.W., and Donald, I.B. (1961) The experimental study of partly saturated soil in the triaxial apparatus. Proc. of the 5th Int. Conf. on Soil Mechanics and Foundation Engineering, 1, 1321. Dineen, K., and Burland, J.B. (1995) A new approach to osmotically controlled oedometer testing. Proc. of the 1st Int. Conf. on Unsaturated Soils, Paris, France, 2, 459465. Fredlund, D.G., and Rahardjo, H. (1993) Soil mechanics for unsaturated soils. John Wiley and Sons, Inc., United States of America. Hilf, J.N. (1956) An investigation of pore water pressure in compacted cohesive soils. Technical Memorandum 654, U.S. Department of Interior Bureau of Reclamation, Denver. Matouk, A., Leroueil, S., and La Rochelle, P. (1995) Yielding and critical state of a collapsible unsaturated silty soil. Gotechnique, 45(3), 465477. Sivakumar, R., Sivakumar, V ., Blatz, J., and Vimalan, J. (2006) Twin-cell stress path apparatus for testing
unsaturated soils. Geotechnical Testing Journal, ASTM, 29(2), 15. Sivakumar, V. (1993) A critical state framework for unsaturated soil. PhD Thesis, University of Sheffield, United Kingdom. Sivakumar, V., and Wheeler, S.J. (2000) Influence of compaction procedure on the mechanical behaviour of an unsaturated compacted clay, Part 1: Wetting and isotropic compression. Gotechnique, 50(4), 359368. Sivakumar, V., Sivakumar, R., Brown, J., Murray, E.J., and MacKinnnon, P. (2010) Constitutive modelling of unsaturated kaolin (with isotropic and anisotropic stress history). Part 1: Wetting and compression behaviour. Accepted for publication by Gotechnique. Tan, W.C. (2004) Wetting, drying, compression and shear strength characteristics of compacted clay. PhD Thesis, Queens University of Belfast, United Kingdom. Thom, R. (2007) Performance of unsaturated soil under monotonic and repeated loading. PhD Thesis, Queens University of Belfast, United Kingdom.
619
Centrifuge modeling applied to swelling soils

B. Caicedo & C. Segura
Universidad de Los Andes, Bogot, Colombia
ABSTRACT: In most tropical regions around the world soils are in an unsaturated state. One of the most important problems affecting shallow foundations of light structures in these regions is the volumetric changes taking place when unsaturated soils are affected by changes in water content. Volumetric changes in unsaturated soils are the result of the interaction between the water migration and stresses. As centrifuge modeling reproduce soil stresses through scaling laws, it could be an important tool to study the expansive behavior of unsaturated soils. This paper presents the validation of the scaling laws corresponding to swelling using the principle of modeling of models. The curves of heave against time for four models are measured. In these curves the expansion at different elapsed times is analyzed using different scale relationship (n 400, 200, 100 and 50 g). The results of the study show that scaling laws corresponding to diffusion process (t* 1/n2t) and to length (L* 1/n L) can be applied successfully for the rate and the absolute value of heave. Finally a centrifuge model of a sallow foundation on expansive soil is presented. 1 INTRODUCTION
Several types of soils show signs of volumetric changes depending on their changes in water content. Particularly the expansive soils affect wide area in the earth where it appears as the main foundation problem. As an example, Jones & Holtz in 1973 estimates that the economic losses per year produced by expansive soils exceeds the sum of economic losses produced by floods, hurricanes and earthquakes. The expansion of a soil is the consequence of the combination of an expansive susceptible soil with climate conditions that allow significant changes in water content. Concerning the climate, Jimnez Salas 1980 made a distinction between two types of weather that generate two types of movement: the monsoonal climate characterized by a cyclic expansion and contraction synchronized with the raining and drying period of the year, and the Saharan climate where the soil starts in a dry condition because during all the year the evaporation exceeds the precipitation. As a consequence, once a construction covers the terrain and forbid the evaporation, the water content under the building increase and generates monotonic heave (figure 1). 1.1 Benefits of centrifuge modeling of expansive soils
Figure 1. Soil expansion and shrinkage on the Bogot clay deposit.
According to CulliganHensley et al 1995, centrifuge modeling offers two major advantages in the physical modeling of geo environmental problems: (i) the technique provides a means of accelerating
transport process in soil, thus enabling observation of phenomena that occur over long prototype times in short model times; and (ii) the technique simulates all self weight effects in the prototype. These two benefits are fundamental in the study of building problems due to expansion. In fact, concerning the two types of climate distinguished by Jimenez Salas, soil deformations take place under long periods of time. In the case of monsoonal climate, the cycles have a period of 6 months or one year depending on the region of the planet, furthermore the expansion behavior is irreversible and therefore it is necessary to observe a significant number of cycles to appreciate the alteration of buildings. Also, in the case of monotonic expansion, the damage on buildings appears after 3 to 5 years. In addition, the changes in stresses and suction have a clear interaction with the self weight of the building, the soil, and with the rigidity of the building.
621
SCALING LAWS FOR EXPANSION PROCESS
by an n2 factor, as well as all diffusion problems in centrifuge modeling (Lord 1999): w n2 w t p t m (4)
Numerous studies about scaling laws for infiltration and drainage have been conducted during the NECER project, Garnier et al. 2000. Most of these studies have been carried out using sandy soils having minor volumetric changes. According to these studies, the infiltration velocity and the capillary rise follow the scaling laws corresponding to time (diffusion process) and length. These scaling laws were validated up to 32 g. In this paper we use higher acceleration values (100 g, 200 g and 400 g) in order to define a larger operation domain for centrifuge modeling of unsaturated expansive soils. These acceleration values are justified by the following facts: expansion takes place in clayed soils, these soils have very small pores and capillary meniscus that are unaffected by the gradient in the acceleration field that appears in the centrifuge. Furthermore, due to the very low permeability of clayed soils, water migration occurs as a laminar flow. To analyze the scaling laws for expansive soils the simplest case of water infiltration coming from the surface of the soil could be considered. In this case, the flow of water in the unsaturated soil is controlled by the generalized Darcys Law (Lloret & Alonso 1980). In one direction this equation is: q Kw ( w ) H z (1)
Equation 4 shows that the relationship between the time in the model t*, and the time in prototype is 1/n2. Furthermore, the model follows the same stress and suction path than the prototype; as a result the strain in the soil is the same than in the prototype. Therefore the heave H that is the integral over length of the strain is scaled in same proportion than the length (H* 1/nH). The suitable technique for checking the scale effects in centrifuge is the modeling of models. It is particularly useful to validate scaling laws. To perform this technique a set of centrifuge models of different scale are tested at appropriate acceleration such that they then correspond to the same prototype. Using the scaling laws the models should predict the same behavior of the prototype. 3 VERIFYING SCALING LAWS FOR EXPANSIVE SOILS USING MODELING OF MODELS Prototype
3.1
In this equation H is the hydraulic potential, Kw is the unsaturated permeability, w is the volumetric water content, and z the direction of flow. As the soil is the same for the model and the prototype, the unsaturated permeability is conserved. Considering a reduced scale model with n as a length scale factor, the discharge of water in the model is n times larger than the discharge in the prototype. This appears since the hydraulic gradient grows in this proportion: the hydraulic potential is conserved and the length is reduced n times. Therefore the flow on the prototype qp and the flow on the model qm are related by the following relationship: qm nq p (2)
In this part of the paper, we attempt to verify the benefits of centrifuge modeling to study the expansion behavior of soil focused in particular to the Bogot soils. The Bogot formation is a very deep deposit of lacustrine clay (the bottom of the deposit is 600 m deep in some places of the city). This deposit is over consolidated in the first 7 to 10 m and represents one of the major geotechnical problems for the shallow foundations. Table 1 shows the main geotechnical characteristics of the first 10 meters deep of the Bogot lacustrine deposit. 3.2 Expansive soil preparation
The samples of expansive soil were obtained by remolding the Bogot clay and using compaction
Table 1. Geotechnical characteristics of Bogot lacustrine clay. Characteristic Liquidity limit Plastic limit Plasticity index Specific gravity (Gs) Organic mater content Unit % % % % Value 147.67 51.38 96.29 2.43 10.75
The evolution of volumetric water content, w, in any point of the soil mass is controlled by the Richards equation (Richards 1931): w H Kw ( w ) t z z (3)
As the scale relationship for length is n, the relationship for the evolution of water content between the prototype and the model is related
622
under vertical stress. The following process was carried out to obtain the expansive samples: 1. Drying: because of the high organic matter content, the soil was air dried to avoid any substantial alteration of the soil, mainly in the particular aspect of expansion in contact with water. Once dried, the soil was reduced to fine particles by grinding (figure 2). 2. Mixture and homogenization: the material was sieved using the sieve #100 in order to avoid the presence of ungrinded particles and obtaining expansion in a continuum medium. The dried material was mixed with 30% of water, this water content is lower than the plasticity limit (PL 52%). This is a favorable situation to obtain an expansive soil. 3. The material was placed in the different moulds (with walls previously lubricated) and was compacted in 5 layers using vertical stress and controlling the height of each layer to obtain a soil having 1800 Kg/m3 unit mass. To prove the expansion characteristics of soil fabricated using the process previously described, a sample was submitted to an oedometric controlled expansion test using the ASTM standard D 454696. Figure 5 represent the expansion curve, this figure shows that the expansion is 4% for a stress of 100 kPa and have an expansion pressure of 3 MPa (stress corresponding to 0% expansion). These values reflect high expansion susceptibility.
Expansi on %
4.5 4 3.5 3 2.5 2 1.5 1 0.5 0 0.01 0.1 1 Vertical stress (Mpa) 10
Figure 5.
Expansion curve in oedometric test.
4 4.1
CONFIGURATION OF THE CENTRIFUGE TESTS Expansion modeling
Figure 2.
a) Air drying process. b) Manual grinding.
Figure 3.
a) Water addition. b) Soilwater mixing.
Figure 4.
Compaction.
The expansion of soils was modeled using two centrifuge machines at Los Andes University in Bogot Colombia. These machines allow geotechnical modeling using a wide range of acceleration. The prototype is a column of 10 m height having over consolidated lacustrine soil from the Sabana de Bogot. The expansion is measured when applying a total inundation of 1 m height at the start of modeling. The principle of modeling of models is used to validate the scaling laws. The models are prepared using cylindrical moulds trying to obtain a uniaxial expansion field. A relationship 1:1 between the diameter of the mould and the height of the soil was used to reduce the wall effect due to the friction between the soil and the wall of the mould. Using this height/diameter relationship the wall effect could be reduced but this effect is still present in the models (Holtz & Kovacs 1981). For this reason, each model was carried out using different moulds maintaining the height/diameter relationship. With this particular care, the wall affects each model in the same way. Table 2 shows the dimensions for each model. Since one of the purposes of this work is to verify the time scale factor in expansive soils, the flight time for each model was defined using the relationship t* 1/N2t. The measure of expansion was performed using a displacement sensor paced at the center of the soil surface. A circular holed plate was placed at the center of the model to distribute the load of the displacement sensor. This plate had a 7.6 m diameter in prototype dimension and the diameter in the models conserved the 1/N scale relationship. Due to the different sizes of the models, and the centrifuges used for the modeling, the measures of displacement were obtained using a stroboscope for the 400 g and 200 g models. For the 100 g models the measures were obtained using an LVDT connected to a computer.
623
Table 2.
Dimensions of the moulds. Acceleration level (g) Model height (cm) Model diameter (cm)
RESULTS FROM THE CENTRIFUGE TEST
Model
5.1 Expansion curve In all models, the expansion curve doesnt arrive to an asymptotic value since at the final stage of the modeling the water continues in the infiltration phase. When the expansion displacement is scaled by N and the time by N2 as shown in figure 8, the curves for the 200 g and 400 g conditions joined with good accuracy. In the case of the 100 g model it is necessary to compare the curves at different modeling time since this model was done at only 1 year in prototype time. Figure 9 shows the results for the 100 g, 200 g and 400 g models, these curves also joined with good accuracy. 5.2 Water content profile
Prototype height (hp): 10 m 1 2 3 400 200 100 2.5 5.0 10.0 2.5 5.0 10.0
Figure 6. Steel moulds used for the physical models carried out at 400 g, 200 g and 100 g.
The expansion of the models takes place as a result of the water infiltration. As we observe in figures 8 and 9 the expansion of the model is scalable using a factor of N for expansion and N2 for time. However, to validate the scale factor for expansion process it is necessary to prove that the soil conditions along the depth is the same in the different models. The water content was measured taking out soil samples at different depth and at the same prototype time for the 200 g and 400 g models.
70 60 Expansion (mm) 50 40 30 20 10 0 0 200 400 600 800 Time (min)
Scale 400 g 200 g 100 g
1000
1200
Figure 8. Figure 7. Schematic drawing of the tests.

20 15
Prototype expansion for different g levels.
The tests started with period of settlement under self weight, this period represents 2.7 years in prototype time. Once the stabilization of the settlement was achieved, the surface of the model was inundated with water using a solenoid valve. For all models this volume of water corresponds to an inundation of 1 m height and no more water was added to the model during the test.
Expansion (mm)
4.2
Test procedure
10
Scale
5 0 0 50 100 Time (min)
400 g 200 g 100 g
150
200
Figure 9.
Prototype expansion for different g levels.
624
30 0
32
Water content % 34 36 38
40
the study of boundary value problems remains. In fact centrifuge modeling could be useful to study the interaction between foundations and soils during swelling. 7 CENTRIFUGE MODELING OF SHALLOW FOUNDATION ON EXPANSIVE SOIL
Depth (m)
6
Scale
200 g 400 g
10
Figure 10. Prototype water content profile for 200 and 400 g levels.
Concerning expansion, the water content profile is an appropriate parameter that reflects the evolution of expansion in depth. Figure 10 shows the water content profile for the test performed at 200 g and 400 g. This figure shows a reasonably agreement between the curves corresponding to 200 g and 400 g. This result reveals that infiltration process in expansive soil follow the proposed scaling laws. 6 LIMITATIONS OF CENTRIFUGE MODELING OF EXPANSIVE SOILS
A preliminary result of a shallow foundation is presented in this paragraph to illustrate the possibilities of centrifuge modeling in studying boundary value problems with expansive soils. This physical model has similar characteristics than the numerical model presented by Mrad et al. (2007), and Nowamooz et al. (2008), Figure 11. The foundation rest over a swelling soil made of a mixture of 60% of Kaolin Speswhite, and 40% of bentonite. The soil is mixed at a water content of 41% and compacted applying a vertical of 500 kPa. The model that is showed in Figure 12 is tested using n 30 as scale factor. The model has two air cylinders equipped with load cells that allow imposing the vertical load on the foundation and the vertical stress on the soil. A displacement sensor allows measuring the settlement or heave of the foundation, and markers are installed to follow the swelling in the soil.
Expansion of soils takes place as a result of the evolution of water content in the voids of the soil. The void space is composed by a double porosity structure. As a result the swelling takes place in two parts: microstructural and macrostructural expansion, Yang et al. (1998). Flow of water at macrostuctural scale is represented successfully by the Darcys law and Richards equation (equations 1 and 3), on the other hand the flow of water at microstructural scale depends on the gradients of water potential in the soil aggregates. These considerations allow concluding that the scale factor for time concerning the macrostuctural swelling is 1/n2 since its controlled by the Darcys law. But the scale factor for swelling at microstructural scale is 1 since the size of the aggregates in the soil is the same in the model and in the prototype. Similar considerations are valid in centrifuge modeling of primary and secondary consolidation. Although microstrural and macrostructural swelling take place in centrifuge models with different time scale, the interest of centrifuge models is
Figure 11. ing soil.
Prototype of a shallow foundation in swell-
Figure 12. Details of the centrifuge model.
625
Figure 13.
Heave of the soil at the final of the test.
The model undergoes a first cycle of drying circulating dyed air at the surface of the model. This phase takes 6 months in prototype scale. Once the drying phase is achieved, an inundation of 30 cm height is applied to the model on the free soil surface; this phase takes 6 months in prototype scale. Figure 13 shows the heave at the surface of the soil at the final of the test. This figure allows observing the following facts: 1. The infiltration depth of the water is lower than the depth of the foundation (60 cm in prototype scale); as a result the heave takes place only in the soil located above the foundation level. 2. As expected the heave is considerably higher in the free surface of the soil, some heave is noticeable in the impervious surface however this heave is due to the undesirable infiltration along the walls of the model. This preliminary test show the possibilities of centrifuge modeling in problems of expansive soils, however some improvements are needed: 1. Detailed studies on the techniques for soil preparation are needed trying to reproduce expansive soils representatives of the soil in place. 2. The devices used to impose the drying and wetting cycles have to be improved to reproduce more precisely the natural weather. 3. Better instrumentation have to be included in the model to measure the heave in different point of the surface, and including sensors for water content and suction in the soil. 8 CONCLUSIONS
tool to study expansive behavior of unsaturated soils. More precisely, this paper uses the principle of modeling of models to validate the scaling laws of time and length for the expansion process in centrifuge. However additional research needs to be done to find out the role of microstructural and macrostructural swelling in centrifuge models. The preliminary test of shallow foundation on expansive soils shows that centrifuge modeling cold be a useful methodology to study boundary value problems dealing with this kind of soils. However aspects like soil preparation, atmospheric actuators and instrumentation for unsaturated soils in centrifuge need further research. REFERENCES
CulliganHensley P.J., Savvidou, C. 1995. Environmental geomechanics and transport processes. In R.N. Taylor (ed), Geotechnical centrifuge technology: 196261. Blackie Academic & Professional. Garnier J., Thorel L., Haza E. (ed.). 2000. Physical modeling and testing in environmental geotechnics. Network of European Centrifuges for Environmental Geotechnic Research. Laboratoire Central des Ponts et Chausses. Holtz R.D., Kovacs W.D. 1981. An introduction to geotechnical engineering. Prentice Hall Inc. Jones, D.E. Jr., and Holtz, W.G. 1973. Expansive soils the hidden disaster. Civ. Eng., ASCE, Vol. 43. Jimenez Salas J.A. 1980. Geotecnia y Cimientos. Madrid: Ed. Rueda. Lloret A. & Alonso E. 1980. Consolidation of unsaturated soils including swelling and collapse behavior. Gotechnique Vol. 30, No. 4, pp. 449477. Lord A.E. 1999. Capillary flow in the geotechnical centrifuge. Geotechnical Testing J. ASTM, Vol. 22, No. 4, pp. 292300. Mrad M., Abdallah A., Masrouri F. (2007). Modlisation numrique du comportement dun sol gonflant charg soumis des variations hydriques. Revue Francaise de Gotechnique No. 120121. Nowamooz H., Mrad M., Abdallah A., Masrouri F. (2008). Experimental and numerical studies of the hydromechanical behaviour of a natural unsaturated swelling soil. In Unsaturated soils, advances in geo-engineering. Toll et al (eds). Taylor and Francis Group. pp. 847853. Richards L.A. 1931. Capillary conduction of liquids in porous mediums. Physics 1, 318333. Yang D.Q., Alonso E.E., Rahardjo H. (1998). Modelling the volumetric behavior of an unsaturated expansive soil. 2nd International Conference on Unsaturated Soils.
According to the results showed in this paper we can conclude that centrifuge modeling is a powerful
626
Centrifuge modeling of soil atmosphere interaction

B. Caicedo & J. Tristancho L. Thorel
Department of Civil and Environmental Engineering, Universidad de los Andes, Bogot, Colombia Laboratoire Central des Ponts et Chausses, Nantes, France
ABSTRACT: Atmosphere process of infiltration or evaporation affect the behavior of geotechnical structures located near the soil surface. This paper focuses on the drying process of soils due to evaporation. The scaling laws are analyzed and afterwards the results on applying two cycles of heating and cooling on a soil mass are presented. Based on these results, conclusions about the feasibility of reproducing evaporation on centrifuge models are recommended. 1 INTRODUCTION 2 2.1 SCALING LAWS FOR INFILTRATION AND EVAPORATION General considerations
The behavior of geotechnical structures like shallow foundations, retaining walls, embankments, slopes and pavements is highly influenced by the water content and pore pressure of the soil located near the surface, where it is in contact with the atmosphere. The phenomena taking place at the boundary between the soil and the atmosphere are associated with heat and mass transport. In the case of compressible soils these phenomena are coupled with volumetric changes. The main parameters involved in the climate affect this thermo-hydro mechanical process: sun radiation, air temperature, atmospheric pressure, wind velocity, rain intensity and hygrometry. The effects of the weather conditions over shallow geotechnical structures are usually modeled numerically. Only few full scale physical models have been carried out; nevertheless, these models require measurements over long periods of time, in particular when the attention focuses on the cyclic behavior. For this reason few physical models are available to calibrate numerical models. Changes in water content and temperature at the soil surface are controlled by diffusion process. For this kind of process, centrifuge modeling is a powerful tool for accelerating the time and could be a cheap way to study the effects of the multiyear weather conditions on shallow geotechnical structures. This paper focuses on the drying process of soils due to evaporation, the scaling laws are analyzed and afterwards the results on applying two cycles of heating and cooling on a soil mass are presented.
The water content near the soil surface is affected by two processes: infiltration and evaporation. Each of these processes is the result of the equilibrium between the atmospheric variables and the capacity of the soil to accept water coming from the atmosphere during rainfall, or to provide vapor to the atmosphere on evaporation (Fig. 1). The flow of water in unsaturated soils is controlled by the generalized Darcys Law (Lloret & Alonso 1980). In one direction this equation is: q Kw ( w ) H z (1)
In this equation H is the hydraulic potential, Kw is the unsaturated permeability, w is the volumetric water content, and z the direction of flow. As the soil is the same for the model and the prototype, the unsaturated permeability is identical.
Figure 1. Equations controlling flow of water on unsaturated soils.
627
Considering a reduced scale model with n as a length scale factor, the discharge of water in the model is n times larger than the discharge in the prototype. This appears since the hydraulic gradient grows in this proportion: the hydraulic potential is conserved and the length is reduced n times. Therefore the flow on the prototype qp and the flow on the model qm are related by the following relationship: qm nq p (2)
The evolution of volumetric water content, w, in any point of the soil mass is controlled by the Richards equation (Richards 1931): w H Kw ( w ) t z z (3)
of the soil to conduct it through the surface, as a result on this zone the infiltration or the evaporation is controlled by the soil. Figure 3 describes the soilatmosphere equilibrium on the prototype and on the model. As described, the infiltration or evaporation flow through the surface of the model grows n times respect to the flow on the prototype (curve 3 on figure 3). Therefore rainfall or evaporation should increases in the same proportion (curve 4 on figure 3), otherwise the water content at equilibrium will be different in the model and in the prototype. 2.2 Modeling rainfall
As the scale relationship for length is n, the relationship for the evolution of water content between the prototype and the model is related by an n2 factor, as all the diffusion problems in centrifuge modeling (Lord 1999): w n2 w t p t m (4)
Rainfall is characterized by its intensity, R, duration D, and frequency F. Considering the scale relationships for time and water flow, this rainfall characteristics are scaled as follows (Fig. 4): Rm nRp , Dm 1 Dp , Fm n2 Fp n2 (5, 6, 7)
Now considers a volume of soil that is subjected to infiltration or evaporation as a result of an imposed pore pressure at the surface. The solution of the Richards equation indicates that the water flowing through the surface reduces with time as indicated in the curve 1 on figure 2 (Wilson 1990). If the atmosphere imposes a constant rainfall or a constant evaporation (curve 2 on Figure 2), the equilibrium between the flow in the soil and the atmosphere are done by the shady zone on figure 2. In fact on the zone A of figure 2 the capacity of the soil to accept or evacuate water is higher than the capacity of the atmosphere; as a result this zone is controlled by the atmosphere characteristics. On the other hand, on zone B, the atmosphere can supply or evacuate more water than the capacity
Reproducing rainfall in a centrifuge model respecting scaling laws is technologically possible using a set of nozzles (Kimura et al. 1991). However to simulate the whole interaction between soil and atmosphere its necessary to reproduce evaporation respecting scaling laws. 2.3 Modeling evaporation in centrifuge models
Modeling evaporation in centrifuge models respecting scaling laws is rater more challenging than rainfall
Figure 3. Soilatmosphere equilibrium on the prototype and on model.
Figure 2. Equilibrium between rainfall or potential evaporation and water flow at the surface of the soil mass.
Figure 4.
Prototype and model rainfall.
628
simulation. When studying evaporation three major factors need to be considered (Fig. 5). The energy input is crucial to supply necessary power for the latent heat of vaporization; the water availability is also a factor that controls the evaporation, and finally the transport of vapor in the atmosphere is important to sustain the vaporization capacity. Depending on the factors included in the process one can distinguish different types of evaporation (Fig. 6): the potential evaporation occurs if only the atmospheric factors are considered and the availability of water is not limited, the actual evaporation comes out when the water availability is considered; finally the evapotranspiration appears when the effect of plants is considered. Our main interest is reproducing the soilatmosphere interaction in centrifuge without considering the effect of plants. Therefore the challenge is reproducing the actual evaporation in centrifuge. In centrifuge the effect of the soil in controlling the water availability appears directly since the unsaturated permeability and hydraulic potential are conserved. Consequently to reproduce actual evaporation it is necessary to reproduce an atmosphere over the model with enhanced evaporation capacity, in fact an atmosphere having potential evaporation n times larger than the potential evaporation of the prototype. To have an idea of the characteristics of the atmosphere at a particular site it is possible to use a psychrometric chart. This chart represents the physical properties of moist air at a constant pressure, and relates how all the properties relate to each other. The thermo physical properties found on psychrometric charts are: the dew point temperature; the relative humidity and the absolute humidity. Figure 7 shows the daily average of the
characteristics of moist air in different places in the world. Replicating the same humidity conditions of a site over a centrifuge model its not enough to reproduce the potential evaporation of a site respecting the scaling laws. In fact, its necessary to enhance n times the potential evaporation over the model. Dalton Laws allows understanding how it can be achieved. John Dalton expressed a relationship for predicting the rate of evaporation from a body of water (Dalton 1802). This was really just a general relationship that said, in essence, that evaporation rate depends on the difference between the watervapor pressure at the evaporating surface and the water vapor pressure in the air above that surface, and will be further influenced by the rate at which wind carries away the evaporated water molecules (Fig. 8). In other words Dalton concluded that the rate of evaporation is in proportion to the deficit of water pressure and the wind speed as:
* E f (u )(es ea )
(8)
where E is the rate of evaporation as height of water per unit time, e*s the saturation vapor pressure at the temperature of the water surface, ea the vapor pressure in the air and f(u) is a function of the mean wind speed u.
Figure 5. Factors controlling evaporation.
Figure 7. Weather in different places on the psychrometric chart.
Figure 6. Different types of evaporation depending on the factors considered.
Figure 8. Schematic diagram of evaporation over water surface.
629
There are dozens of variations on Daltons Law, each with its own unique set of constants and exponents, empirically calibrated to specific evaporation experiments performed all over the world. Most of these equations are in the general form (Sartori 2000):
* E f ( As Bs u )(es ea )
(9)
Figure 9. Schematic drawing of the climatic chamber.
where As and Bs are empirically calibrated constants. Analyzing the Daltons equation and its modifications it appears that there are the following possibilities to grow the potential evaporation over a model: growing the wind speed; growing the temperature at the surface of the model and as a consequence the saturation vapor pressure, and reducing the vapor pressure on the atmosphere either reducing the temperature or the relative humidity. 3 3.1 A FIRST TEST ON SOIL ATMOSPHERE INTERACTION IN CENTRIFUGE Description of the model
A first test to model the soil atmosphere interaction in centrifuge during drying was carried out at the Laboratoire des Ponts et Chausses in Nantes using the climatic clamber developed at the University of los Andes at Bogot (Tristancho & Caicedo 2008). This first test consists in applying two cycles of heating and cooling over a centrifuge model made of compacted kaolin. Temperature, suction, relative humidity and settlement are measured during the cycles. The applied g-level is 20. 3.2 Preparation of soil
Figure 9 shows the internal structure of the CSC. The air located over the surface of the model circulates towards the chamber driven by three axial discharge fans. The moist air run into a dehumidifier prism, this prism uses the psychrometric process of latent heat elimination. The dew point temperature of the prism is controlled using Peltier plates that reduce the temperature in the prism obtaining condensation. The condensed water generated by the loss of latent heat is canalized by the gravity to a closed deposit and monitored. Afterwards the air is canalized to the heating prism that increases its temperature using again the Peltier effect. Once the humidity and temperature of the air are adjusted, the air is driven by means of three fans towards the surface of the model. The fans installed are capable of generating a wind speed the order of 7.2 m/s (Approx. 26 km/h). The system at full load is able to make the complete air interchange on the model every 2 sec. 3.4 Instrumentation
The soil is placed into a 40 60 cm rectangular adiabatic strongbox made of composite materials. The soil mass is subjected to a compressive vertical stress v 2.5 MPa using a pressure-controlled hydraulic actuator. This value enables obtaining a homogeneous material at the desired dry density (Murillo et al. 2009). The static compaction of the soil is carried out during 10 minutes on three 10 cm-thick layers. For unsaturated soils, a few minutes only are necessary to obtain the required compaction settlement. 3.3 Climatic chamber for centrifuge
The control system set inside the CSC is a close-loop system. The CSC has the following instrumentation: 2 Relative humidity sensors 3 Contact thermometers 1 Pressure sensor. The interaction between the atmosphere and the soil induces water migration and volumetric changes. These volumetric changes generate heave or settlement and cracks could appear at the soil surface. The following sensors are installed over the model to measure the behavior of the surface of the model on heating and cooling: Laser displacement sensor Non contact infrared thermometer (surface temperature) Relative humidity sensor Contact thermometer (air temperature) Wind velocity sensor The total set up of the system is represented in figures 10 and 11.
The Climatic Simulation Chamber (CSC) is located on the upper part of the strongbox, this chamber was designed to control all the weather parameters. The principle of operation is based on heat transfer for convection. The relative humidity is controlled by the extraction of moisture by condensation, dew point.
630
Figure 10. Detail of the IR temperature sensor and laser sensor.
figures 14 and 7, it appears that the climatic chamber allow reproducing driest atmosphere than most of the sites presented in figure 7, however it seems that reproducing a very dry environment, as for example a relative humidity of 20%, needs enhancing the power of the set of Peltier plates. On the other hand, using the convection process as the only mechanism to supply energy for evaporation could be insufficient to model evaporation. In fact, according to the analysis made on paragraph 2.3, modeling evaporation respecting scaling laws needs to increase the potential evaporation in the model in the same proportion than the acceleration. The measurement of the temperature at the surface of the model made by the infra red sensor allows calculating the deficit of vapor pressure between the surface of the soil and the volume of air above
Figure 11. Setup of the model with the climatic chamber.
Figure 12. Temperature in the atmosphere and at the surface of the soil.
Figure 12 show the measures of temperature in the air over the model and at the surface of the soil. On this first tests the heating and cooling of the soil is made by convection between the air over the model and the soil. As a result, to increase the temperature at the soil surface up to 30C its necessary to rise the air temperature up to 38C. The behavior on cooling is similar than in heating, in this case the temperature of the air at the final of each cooling cycle is lower than the temperature on the surface of the soil. These results obtained on heating and cooling show the effectiveness of the Peltier plates placed in the climatic chamber as a reversible heat pump. Figure 13 shows the measurements of the relative humidity in the volume of air above the model. At the final of the heating cycles the relative humidity goes down to 55%, although at the final of the cooling cycles the relative humidity grows up to 88%. In addition two condensation events are measured during the commutation from cooling to heating, these events appear since the commutation was made stopping the centrifuge and therefore the change in air pressure provokes condensation. The measures of temperature and relative humidity allow drawing the heating and cooling cycles in the psychrometric chart, figure 14. Comparing
Figure 13. Relative humidity during cooling and heating.
Figure 14. Heating and cooling cycles represented in the psychrometric chart.
631
the model. This deficit is presented in figure 15. As observed, the maximum value of the deficit of vapor pressure is 3 kPa and is obtained at the final of the heating cycle. If the growing of potential evaporation is made only by growing the vapor pressure deficit (VPD), the value of VPD 3 kPa on the model at 20 g correspond to a VPD 150 Pa on the prototype. This could be insufficient for most sites, for example Choudhury (1998) reported extreme values of VPD up to 6 kPa with the majority of the values in the range of 0 VPD 1 kPa. This result allows concluding that growing the potential evaporation needs to combine all the possibilities of supplying energy: applying radiation, reducing the relative humidity and growing the wind speed. Finally figure 16 shows the settlement of the surface of the model measured using the laser sensor. As described, the soil used for this model is compacted Kaolin, which exhibits few volumetric changes on drying due to the high compaction level and low water content. As a result the measured settlement is the combination of volumetric strains due to drying and thermal strains. These two mechanisms act in the same direction during cooling and in opposite direction during heating, for this reason the slope of the settlement curve is higher during the two cooling cycles. Future tests need to be performed on soils exhibiting more volumetric changes on drying.
CONCLUSIONS
This paper deals with the physical modeling of soil atmosphere interaction in centrifuge. Although two main processes control this interaction: infiltration and evaporation, this paper focus mainly in evaporation. The scaling law for evaporation is analyzed concluding that to preserve the same equilibrium between the soil and the atmosphere, in the prototype and the model, the potential evaporation have to grow in the same proportion than the length scale. The results of a first test on soil atmosphere interaction using the Universidad de los Andes climatic chamber are presented. In this test the energy supply for evaporation is made only by convection. The results show that only convection is insufficient to grow substantially the vapor pressure deficit. Consequently its necessary to supply more energy for evaporation using for example radiation devices, reducing the relative humidity and growing the wind velocity. How these three factors are combined to reproduce the potential evaporation respecting the scaling laws is a topic under study. REFERENCES
Choudhury B.J. 1998. Estimation of vapor pressure deficit over land surfaces from satellite observations. Adv. Space Res. Vol. 22, No. 5, pp. 669672. Dalton J. 1802. Experimental essays on the constitution of mixed gases; on the force of steam or vapor from water and other liquids in different temperatures, both in a Torricellian vacuum and in air; on evaporation and on the expansion of gases by heat. Mem. Manchester Liter. and Phil. Soc. 511, 535602. Kimura T., Takemura J., Suemasa N., Hirooka A. 1991. Failure of fills due to rain fall. Centrifuge 91, International Conference, Boulder Colorado, pp. 509516. Lord A.E. 1999. Capillary flow in the geotechnical centrifuge. Geotechnical Testing J. ASTM, Vol. 22, No. 4, pp. 292300. Lloret A. & Alonso E. 1980. Consolidation of unsaturated soils including swelling and collapse behavior. Gotechnique Vol. 30, No. 4, pp. 449477. Murillo C.A., Thorel L., Caicedo B. 2009. Spectral analysis of surface waves method to assess shear wave velocity within centrifuge models. J. Applied Geophysiscs 68, pp. 135145. Richards L.A. 1931. Capillary conduction of liquids in porous mediums. Physics 1, 318333. Sartori E. 2000. A critical review on equations employed for the calculation of the evaporation rate from free water surfaces. Solar Energy Vol. 68 No. 1, pp. 7789. Tristancho J. & Caicedo B. 2008. Climatic chamber to model soil atmosphere interaction in the centrifuge. In Unsaturated Soils, Advances in Geo-Engineering, Toll et al. (eds), Taylor & Francis Group. Wilson W. 1990. Soil evaporative fluxes for geotechnical engineering problems. Ph.D. dissertation, University of Saskatchewan, Canada, 464 pp.
Figure 15. Vapor pressure at the soil surface and in the air and deficit of vapor pressure during the cycles.
Figure 16.
Settlement measured during the cycles.
632
Physical modelling on unsaturated soils using centrifuge

B. Caicedo & J. Tristancho L. Thorel
Department of Civil and Environmental Engineering, Universidad de los Andes, Bogot, Colombia Laboratoire Central des Ponts et Chausses, Nantes, France
ABSTRACT: The behavior of geotechnical structures located near the soil surface is highly influenced by unsaturated soil conditions. Some researches about the potentialities of centrifuge modeling of geotechnical problems involving unsaturated soils are presented, with a focus on scaling laws, soil preparation, instrumentation and atmospheric actuators. Finally some ideas to improve the practice of centrifuge modeling with unsaturated soils are presented. 1 INTRODUCTION designed to investigate process using the centrifuge field as a tool (centrifuge testing), and those that are performed to reproduce data to predict behavior of geotechnical process (centrifuge modeling). This last category makes use of scaling of the results of centrifuge experiments (the model scale) to the problem under investigation (the prototype scale). A powerful means to analyze the scaling laws, when the mathematical description of the process is well established, is the inspectional analysis (Barry et al. 2001). This analysis involves mapping the equation controlling the process in a non dimensional form. However inspectional analysis relies on the invariance of the physical law under changes of scale (Brikhoff 1960). Modeling of models (Schofield, 1980) can be used when the theoretical model is not available or to validate the scaling laws obtained using inspectional analysis. This technique involves measures at different acceleration levels so that scaling laws can be inferred directly from the measured data. 2.2 Validation of scaling laws for non-deformable unsaturated soils Prototype-model scaling in unsaturated porous media that undergo negligible consolidation in the centrifuge has received much attention previously. Several authors (e.g. Barry et al. 2001, Culligan et al. 1997, Burkhart et al. 2000, Cranon et al. 2000, Khalifa et al. 2000, Rezzoug et al. 2000 (a), Rezzoug et al. 2000 (b), Knight et al. 2000, Soga et al. 2000, Thorel et al. 2000, Depountis et al. 2001) have presented scaling analysis for porous media flow in unsaturated soils. Three processes are relevant on scaling laws of flow in unsaturated soils having negligible volume changes: the flow velocity or the discharge,
Centrifuge technique has been widely used to model geotechnical structures mainly when the free surface or the stress gradient plays an important role on the structures behavior. Most of these models have been carried out using mainly saturated clays or dry sands. However the behavior of geotechnical structures located near the soil surface is highly influenced by the water content of the soil located near the surface where the soil is frequently unsaturated. Experimental complexities related with unsaturated soils are well known in laboratory. Concerning centrifuge modeling these complexities are enhanced, however the potentialities of centrifuge modeling of geotechnical problems involving unsaturated soils may well justify the additional effort. On the contrary of many centrifuge applications, very few data is available on scaling laws for unsaturated soils (Garnier et al. 2007). This paper presents the results of several authors on scaling laws related to unsaturated soils as well as the results of some models carried out on unsaturated soils. The problems related with soil preparation, instrumentation and atmospheric actuators are highlighted, as well as successful experiences found in the literature. Finally some ideas to improve the practice of centrifuge modeling with unsaturated soils are presented. 2 2.1 SCALING LAWS FOR UNSATURATED SOILS IN CENTRIFUGE General considerations
Experiments carried out using geotechnical centrifuge generally fall in two categories: those
633
the dynamics of the evolution of water content or saturation, and the water content profile at equilibrium. The suction curve of a soil depends on its fabric and mineralogy, these two factors controls the pore sizes and the adsorption of water. Furthermore the ions in water affect the osmotic suction. Due to the complexities of unsaturated soils, it is better to use in the model the same soil than the prototype. In other words, to use a soil having the same mineralogy, reproducing the same fabric and uses the same water. Under these conditions the suction curve is conserved and the profile of water content at equilibrium in a soil column is scaled in the same ratio than the length. In fact, considering a soil column subjected to capillary infiltration (Fig. 1), the pore water pressure in two homologous points Ap and Am, located at heights Zp and Zm Zp/n are the same: u p w gZ p um w gnZm w gn Zp n (1) (2)
Figure 2. Capillary height for different g levels: (a) model, (b) prototype (Thorel et al. 2000).
Considerable efforts has been done during the NECER project (Garnier et al. 2000), to validate the scaling law for capillary rise. For example, Figure 2a shows the results reported by Thorel et al. (2000) concerning the profile of water content in capillary rise for different centrifuge levels. Figure 2b shows the good agreement for all the results reported in prototype scale using 1/n as scale factor. The flow of water in unsaturated soil is commonly described using generalized Darcys law
Figure 3. Drainage flow versus g-level (Thorel et al. 2000).
(Darcy 1856, Buckingham 1907, Richards 1931, Childs & Collis George 1950): vw Kw H z (3)
In this equation H is the hydraulic potential, Kw is the unsaturated permeability, and z the direction of flow. The coefficient of permeability, Kw, is a function of any two of three possible volume-mass properties (Lloret & Alonso 1980): K w K w (S , e ) or K w K w (e, w ) or K w K w (w, S ) (4) If the model and the prototype have the same permeability and the pore water pressure is conserved in homothetic points, the flow of water in the prototype and in the model are:
Figure 1. Capillary rise in prototype and in model scales.
vwp K w
H H , and vwm K w z p zm
(5)
634
Then: vwm nvwp (6)
Figure 3 shows the results reported by Thorel et al. (2000) concerning the flow of water during capillary rise tests at different g levels. Good agreement is observed using n as a scaling factor. The evolution of water content in unsaturated soils is done by the Richards equation (1931): w H Kw t z z (7)
volumetric changes on unsaturated soils are very active. However most authors agree with taking separately the role of total stress and suction, for example Fredlund & Rahardjo (1993) propose for the volumetric changes the following equation: v s ( ua ) s ( ua uw ) m2 m1 t t t (9)
In this case the double derivation about z, implies that the evolution of the water content as a function of time is scaled with a factor of n2 (Lord, 1999): w n2 w t p t m (8)
where m1s is the coefficient of volume change with respect to a change in net stress ( uw); and m2s is the coefficient of volume change with respect to a change in matric suction (ua uw). The change of matric suction with time is controlled by equation 7. As a result if the change in net stress is null, the relationship of volumetric strain between the model and the prototype is: v n2 v t p t m (10)
Figure 4 shows the results reported by Rezzoug et al. (2000) concerning the dynamic of the capillary rise as a function of time, and the good agreement of the n2 scale factor. 2.3 Validation of scaling laws for deformable soils
This result is valid too when there is a change in net stress, however if the change in net stress is the result of an external load, the loading velocity must be scaled with a factor of n2. The displacement H of the surface of a soil column is: H v dz (11)
Most of studies relating scaling laws applied for unsaturated soils nave been done on sands having little or negligible volumetric changes on infiltration. Concerning unsaturated deformable soils (Bear et al. 1984) studied the centrifugal filtration in unsaturated deformable soil, and Caicedo et al. (2006) presented a validation of the scaling laws for expansive soils. Volumetric changes in unsaturated soils depend on the variation of total stress and suction (Fredlund & Rahardjo 1993). Discussions about the best way to take in to account suction and total stress in
As the volumetric change is conserved in the model and in the prototype, the scale factor for displacement is: H p nH m (12)
The results reported by Caicedo et al. (2006) confirm these scaling relationships for expansive soils. These results concern the modeling of a 10 m high soil column of over consolidated lacustrine soil from the Sabana de Bogot. The expansion was measured when applying a total inundation of 1 m high at the start of modeling (Fig. 5). The soil
Figure 4. Capillary rise versus time: (a) model, (b) prototype (Rezzoug et al. 2000).
Figure 5.
Schematic draw of expansion tests.
635
Figure 6.
Prototype expansion for different g-levels.
Figure 7. 400 g.
Prototype water content profile for 200 and
column was modeled with accelerations of 100 g, 200 g, and 400 g. Figure 6 shows the expansion displacement in prototype scale using n and n2 as scale factors for heave and time respectively. These curves show the good agreement obtained using these scale factors. Concerning expansion, the water content profile is an appropriate parameter that reflects the evolution of expansion in depth. To verify that the soil conditions along the depth are the same in the different models, the water content was measured taking out soil samples at different depth and at the same prototype time for the 200 g and 400 g models. Figure 7 shows the good agreement of the water content profile for the tests performed at 200 g and 400 g. 3 PREPARATION OF SOILS
suitable to predict the behavior of soils during the test. Some studies have been carried out in order to establish controlled methodologies to reproduce intermediate unsaturated soils. These procedures could be grouped in two main techniques: (i) the inclusion of a cementing material in sandy soils (Abdulla & Goodings 1994, Dupas & Pecker 1979, Ismail et al. 2000), and (ii) mixtures of clay and sand compacted by uniaxial compression (Kimura et al. 1994, Boussaid et al. 2005, Murillo et al. 2009). However the previously mentioned methods suffer from some limitations mainly concerning the possibility of controlling the stress path during compaction. In fact, traditional compaction techniques using blows or kneading made impossible the knowledge of the stress path during compaction. Although uniaxial compression allows soil preparation under controlled vertical stress, the whole stress path remains unknown. However the behavior of compacted unsaturated soils having expansive or collapsible behavior is strongly dependent on their negative pore water pressure and their stress-suction history. An alternative to identify the stress-suction history during compaction is the use of suction monitored apparatus (triaxial or oedometer). These apparatus allows the characterization of unsaturated soils in a fraction of time compared with suction controlled apparatus (Blatz & Graham 2003, Jotisankasa et al. 2007). This kind of apparatus makes possible to assess the stress-suction state and the history of an unsaturated soil on vertical stress compaction. Furthermore its possible to establish the best preparation procedure to obtain soils having different behavior. For example, Figure 8 shows the stresssuction trajectory of kaolin during compac-
Physical modeling of unsaturated soils has heightened the need for soil preparation techniques
Figure 8. Stresssuction history of a Kaolin under vertical stress compaction.
636
tion, this compaction was made applying a first loading stage and then a cycle of unloading and reloading. Finally the sample is saturated to assess its behavior during wetting. Although preparation of unsaturated soils for physical modeling by vertical stress compaction is a useful technique, this method is only valid for modeling compacted soils. In fact this technique doesnt replicate the fabric of natural unsaturated soils. For this reason there is a need of developing new preparation techniques that replicates in a better way the unsaturation process in the field, for example reducing phreatic level or drying from the surface. 4 4.1 INSTRUMENTATION Suction sensors
can be obtained using psychrometers in isothermal conditions under temperature gradients. The offset measured in the psychrometer grows and therefore the measure is less accurate. 4.2 Water content sensors
The measure of suction in unsaturated soils is a difficult task for the following reasons: It doesnt exist sensors covering the broad range of suction values of unsaturated soils (Table 1). Direct measurements of suction have to deal with cavitation. The response time of the sensors varies in a broad range. All these difficulties are enhanced in centrifuge modeling due to the presence of the acceleration field and the need for miniaturization. For physical modeling in centrifuge the use of standard tensiometers have been reported by Chiu et al. (2005), high performance tensiometers have been tested in centrifuge during the MUSE project (Loureno et al. 2006, Muoz et al. 2010). The measurements carried out with tensiometers in centrifuge are consistent with those carried out in laboratory. However, at an elevated g-level the tensiometers cavitate at a lower maximum sustainable suction than in laboratory (Chiu et al. 2005). Psychrometers have been tested in centrifuge by Tristancho & Caicedo (2008) using a CR7 Campbell data acquisition system installed in the centrifuge arm. The results reported by Tristancho & Caicedo (2008) show that consistent measurements
Different devices for measuring water content in centrifuge have been tested (Gunzel et al. 2003), such as capacitive sensors (Dupas et al. 2000), or TDR sensors (Cranon et al. 2000, 2001). The principle of capacitive sensors is based on the variation of the relative dielectric permittivity of soil, , and the contrast between water ( 80), soil ( 4 to 10), and air ( 1). The electronic circuit to measure the dielectric permittivity in water content sensors is a variable frequency oscillator using an inductor and a variable capacitor. Although measures using capacitive sensors are reliable, carefully calibration is needed because the readings depend on the shape of the electrodes and on the type of soil. The technique to measure water content based on the time domain reflectometry, TDR, is based on the propagation velocities of the waves in the porous medium. As the presence of water in the medium affects the speed of the electromagnetic wave, the energy which does not become dissipated in the medium and returns to its source depends on the water content. The accuracy of TDR measurements depends on precise measurement of time and careful calibration with the relative volumetric content of water around the probe. However TDR sensors in centrifuge models have to deal with the problem of its size. In fact the principle treasuring technique of the TDR sensors makes its miniaturization difficult. 5 ACTUATORS
Table 1.
Devices for measuring soil suction. Range (MPa) 1090 0.56.5 02 00.09
Suction sensor Hygrometer Psychrometer High performance tensiometer Standard tensiometer
Mechanical actuators like load actuators or robots designed to work on traditional centrifuge models on dry or saturated soils are useful to work on unsaturated soils. However the most challenging actuators for centrifuge models on unsaturated soils are atmospheric actuators. These actuators try to replicate a number of atmospheric variables to study the soil response under particular environmental conditions. The complexity of the atmospheric actuators depends on the number of environmental variables that are reproduced. For example the following actuators have been used in centrifuge modeling: Rain actuators (Kimura et al. 1991). Rain and relative humidity environmental chamber (Take & Bolton 2002).
637
Rain, relative humidity, wind, temperature and radiation in environmental chamber (Tristancho & Caicedo 2008). Atmospheric actuators offer interesting possibilities of studying the behavior of geotechnical structures under extreme environmental conditions. However the replication of atmospheric variables respecting scaling laws need a complex control system on the actuators, this particular point is under development nowadays. 6 CONCLUSIONS
This paper deals with the physical modeling of unsaturated soils, both considered in the cases of non-compressible and compressible. An overview on soil preparation, instrumentation and actuators used for experiments on unsaturated has been presented. The challenge is to perform new experiments with engineering applications, taking into account water migration in unsaturated soils. REFERENCES
Abdulla W.A. & Goodings D.J. 1994. Study of sinkholes in weakly cemented sand. Centrifuge 94, Leung, Lee and Tan(eds) Balkema, Rotterdam, pp. 797802. Barry D., Lisle I., Prommer H., Parlange J., Sander G. & Griffioen J. 2001. Similitude applied to centrifugal scaling of unsaturated flow. Water Resour. Res. Vol. 37, 10, pp. 24712479. Bear D.A. Corapcioglu M.Y. & Balakrishna J. 1984. Modeling of centrifugal filtration in unsaturated deformable porous media. Adv. Water. Resour. 7, pp. 150167. Birkhoff G. 1960. Hydrodynamics 2nd ed. Princeton Univ. Press, Princeton N.J. Blatz J.A. & Graham J. 2003. Elasticplastic modelling of unsaturated soil using results from a new triaxial tests with controlled suction. Gotechnique 53. No. 1, pp. 113122. Boussaid K., Thorel L., Garnier J., Ferber V. & David J.P. 2005. Comportement mcanique de sols intermediaries reconstitutes: Influence de la teneur en eau et du percentage dargile. Congrs francais de mcanique, Troyes, France. Buckingham E. 1907. Studies of the movement of soil moisture. U:S:D:A: Bur. Of Soils, Bulletin No. 38. Burkhart S., Davies M.C.R, Depountis N., Harris C. & Williams K.P. 2000. Scaling laws for infiltration and drainage tests using a Geotechnical Centrifuge. In Physical Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 191198. Caicedo B., Medina C. & Cacique A. 2006. Validation of time scale factor of expansive soils in centrifuge modeling. In Physical modeling in Geotechnics ICPMG 06, Ng, Zhang & Wang (eds.). pp. 273277.
Childs E.C. &, Collis-George N. 1950. The permeability of porous materials. Proc. Royal Soc. Vol. 201 A, pp. 392405. Chiu C., Cui Y., Delage P., de Laure E. & Haza E. 2005. Lessons learnt from suction monitoring during centrifuge modeling. Advanced Experimental Unsaturated Soil Mechanics. Tarantini, Romero, Cui Eds. Taylor & Francis, pp. 38. Crancon P., Guy C. & Pili E. 2000. Modeling of capillary rise and water retention in centrifuge tests using TDR. In Physical Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 199206. Cranon P., Thorel L. & Garnier J., 2001. Flows in unsaturated soils: experiments and physical modelling in geotechnical centrifuge. XVth ICSMGE. Istambul 2731 aot 01, ISBN 90 2651 853 6, pp. 569574 (in French). Culligan O., Barry D. & Parlange J. 1997. Scaling unstable infiltration in the vadose zone. Can. Geot. J., 34, pp. 466470. Darcy H. 1856. Histoire des fontaines publiques de Dijon. Paris Dalmont pp. 590594. Depountis N., Davies M.C.R., Harris C., Burkhart S., Thorel L., Rezzoug A., Knig D., Merrifield C. & Craig W.H. 2001 Centrifuge modelling of capillary rise, Engineering Geology, ISSN: 00137952 Vol. 60, 14, June 01, pp. 95106. Dupas A., Cottineau L.M., Thorel L. & Garnier J. 2000. Capacitive sensor for water content measurement in centrifuged porous media. In Phys. Modeling & Testing in Environmental Geotech.. Garnier, Thorel, Haza Eds, pp. 1117. Dupas J.M. & Pecker A. 1979. Static and dynamic properties of sandcement. ASCE Journal of Geotechnical Engineering Vol. 105, GT3, pp. 419436. Fredlund D.G. & Rahardjo H. 1993. Soil mechanics for unsaturated soils. John Wiley & Sons (ed.). Garnier J., Thorel L. & Haza E. (ed.). 2000. Physical modeling and testing in environmental geotechnics. Network of European Centrifuges for Environmental Geotechnic Research. Laboratoire Central des Ponts et Chausses. 392 p. Garnier J., Gaudin C., Springman S.M., Culligan P.J., Goodings D., Konig D., Kutter B., Phillips R., Randolph M.F. & Thorel L. 2007 Catalogue of scaling laws and similitude questions in geotechnical centrifuge modelling. Int. J. Physical Modelling in Geotechnics. Vol. 7, n 3, pp. 124. Gnzel F.K., Craig W.H., Cranon P., Cottineau L.-M., Kechavarzi C., Lynch R., Merrifield C.M., Oung O., Schenkeveld F.M., Soga K., Thorel L. & Weststrate F.A. 2003 Evaluation of Probes and Techniques for Water Content Monitoring in Geotechnical Centrifuge Models. Int. J. Physical Modelling in Geotechnics, Vol. 3, n 1, pp. 3143. Ismail M.A., Joer H.A. & Randolph M.F. 2000. Sample preparation technique for artificially cemented soils. ASTM Geotech. Testing J. Vol. 23 No. 2, pp. 171177. Jotisankasa A., Ridley A. & Coop M. 2007. Collapse behavior of compacted silty clay in suction-monitored oedometer apparatus. J. Geot. & Geoenv. Engng. July, pp. 867877. Khalifa A., Garnier J., Thomas P. & Rault G. 2000. Scaling laws of water flow in centrifuge models. In Physical
638
Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 207216. Kimura T., Takemura J., Hiro-Oka A. & Okamura M. 1994. Mechanical behaviour of intermediate soils. Centrifuge 94. Singapore, Leung et al. (Ed), Balkema, pp. 1324. Kimura T., Takemura J., Suemasa N. & Hirooka A. 1991. Failure Of Fills Due To Rain Fall. Centrifuge 91, International Conference, Boulder Colorado, pp. 509516. Knight M.A., Cooke A.B. & Mitchell R.J. 2000. Scaling of movement and fate of contaminant releases in the vadose zone by centrifuge modeling. In Physical Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 233242. Lloret A. & Alonso E. 1980. Consolidation of unsaturated soils including swelling and collapse behavior. Gotechnique Vol. 30, no. 4, pp. 449477. Lord, A.E. 1999. Capillary flow in the geotechnical centrifuge. Geotech. Testing J., ASTM, Vol. 22, No. 4, pp. 292300. Loureno S.D.N, Gallipoli D., Toll D.G. & Evans F.D. 2006. Development of a commercial tensiometer for triaxial testing of unsaturated soils. Geotechnical Special Publication N 147, ASCE, Reston, Vol 2, 18751886. Muoz J.J., Casini F., Loureno S., Vaunat J., Pereira J.-M., Thorel L., Garnier J., Delage P; Alonso E. & Gallipoli D. 2010 Centrifuge modelling of a shallow foundation on an unsaturated compacted silt. Gotechnique. (to be submitted). Murillo C., Thorel L. & Caicedo B. 2009. Spectral analysis of surface waves method to assess shear wave velocity within centrifuge models. J.Applied .Geophysiscs 68 pp. 135145.
Rezzoug A., Konig D. & Trantafylidis Th. 2000 (a). Scaling laws in centrifuge modeling for capillary rise in soils. In Physical Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 217224. Rezzough A., Konig D. & Trantafylidis Th. 2000 (b). Numerical analysis of scaling laws for capillary rise in soils. In Physical Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 217224. Richards L.A., 1931. Capillary conduction of liquids in porous mediums. Physics 1, 318333. Schofield A.N. 1980. Cambridge geotechnical centrifuge operations. Gotechnique 30, pp. 227268. Soga K. & Comoulos H. 2000. Some remarks on water movement in homogeneous unsaturated soils in relation to centrifuge testing. In Phys. Modeling & Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 243250. Take W.A. & Bolton M.D. 2002. The use of centrifuge modelling to investigate progressive failure of over consolidated clay embankments. In Constitutive and Centrifuge Modelling: Two Extremes, Springman (ed), pp. 191197. Thorel L., Noblet S., Garnier J. & Bisson A. 2000. Capillary rise and drainage flow through a centrifuged porous medium. In Physical Modeling and Testing in Environmental Geotechnics. Garnier, Thorel, Haza Eds, pp. 251258. Tristancho J. & Caicedo B. 2008. Climatic chamber to model soil atmosphere interaction in the centrifuge. In Unsaturated Soils, Advances in Geo-Engineering, Toll et al. (eds), Taylor & Francis.
639
Commentaries on the consolidation-swell test

K.C. Chao, J.D. Nelson, D.D. Overton & E.J. Nelson
ABSTRACT: The consolidation-swell test is commonly used for measuring expansion potential and predicting heave for expansive soils in the Front Range area of Colorado, USA. In performing the test and analyzing the results, it is important to consider the specific nature of the dense highly consolidated claystones and clayshales of this area. Such materials have also been observed to exist throughout the Western United States and other parts of the world. This paper presents a discussion of many factors that are considered in the ASTM Standard Test Method but which are inadequately addressed for these materials and are not always observed by commercial testing laboratories. 1 INTRODUCTION 2 COMMENTARIES
Standards for the performance of tests to measure expansion potential of soils are set forth in ASTM Standard Test Method D454608 (ASTM, 2008). This particular standard will be termed simply The Standard in this paper. These test methods were developed for expansive soils in general. The area around Denver, Colorado, USA, is also known more generally as the Front Range area, because it exists along the Eastern or Front range of the Rocky Mountains. In this area, very dense highly over-consolidated claystones and clayshales are encountered. Severe distress to light structures constructed on these soils is common. Consolidation-Swell tests are routinely conducted as part of geotechnical investigations. Because of the nature of these soils and bedrocks, deviations from The Standard have been developed for use in this area. The revised test is commonly called the Denver Consolidation-Swell test. Some of the factors that will influence the results of the Consolidation-Swell test are different for the Front Range area than addressed in the ASTM Standard. Also, some parts of The Standard are inadequately followed by some commercial testing laboratories. This paper will discuss the factors affecting consolidation-swell test data and review the current testing procedures. The paper will present commentaries on the various factors based on observations and research performed by the authors. The commentaries are not intended as criticism of The Standard. Rather they are observations by the authors, and present consideration for practitioners concerned with such soil and bedrock.
2.1 Swell pressure A typical oedometer used in Colorado, USA is shown in Figure 1. This is termed a consolidometer in The Standard. This apparatus is commonly used to measure the percent swell, S%, and the swell pressure of a soil. The Standard defines swell pressure as the minimum stress required to prevent swelling.
DEFORMATION DIAL GAUGE FULCRUM POINT
LOAD APPLIED LOADING BAR
WEIGHT
SAMPLE
P = APPLIED LOAD
CONSOLIDOMETER BASE POROUS STONE SPECIMEN RING
LOAD/ PRESSURE BLOCK WATER LEVEL
FILTER PAPER
SOIL SAMPLE POROUS STONE
Figure 1.
Oedometer apparatus.
641
There are two types of oedometer test that are commonly performed. One is the consolidation-swell (CS) test in which a sample is allowed to swell, and then loaded to compress the sample back to its original height. This corresponds to method C in The Standard. Another type of oedometer test is the constant volume (CV) test. In that test the sample is confined and not allowed to swell while it is inundated or wetted. The stress required to prevent the sample from swelling is termed the constant volume swell pressure. Method A in The Standard consists of performing oedometer tests on four different samples of soil. Each sample is inundated at a different applied stress and the percent swell, S%, is plotted against the inundation stress. Figure 2 shows data for Method A as plotted in The Standard. It is normal to plot oedometer test results in semi logarithmic form. Figure 3 shows the Method A data from The Standard plotted in semi logarithmic form. It is seen that the data plots as a straight line. Figure 4 shows idealized oedometer test results. It is seen that the swell pressure required to prevent expansion of the sample in the constant volume test is less than the swell pressure obtained in the consolidation-swell test in which the sample has imbibed a significant amount of water during swelling. The swell pressure measured in the consolidation-swell test is termed the ConsolidationSwell Swelling Pressure, cs. The swelling pressure measured in the constant volume swelling test is termed the constant volume swelling pressure, cv. This is the swell pressure that would correspond to the vertical strain of zero in Figure 2 taken from The Standard. A relationship between the constant volume swell pressure and consolidation-swell
12 Vertical strain,% Collapse () Swell (+) 10 8 6 4 2 0 -2 -4 -6 -8 0 100 200 300 400 500 600
lb ft 2)
6 5 4 Vertical strain,% 3 2 1 0 S% CH
-1 -2 -3 10 (200 psf)
'i 'vo 100
(2,000 psf)
'cv
1,000 (20,000 psf)
Figure 3. Method A data from The Standard plotted in semi-logarithmic form.
B
PERCENT SWELL S%
CONSOLIDATION-SWELL TEST DATA
S%
CH
D
CONSTANT VOLUME TEST DATA
APPLIED STRESS, ' (LOG SCALE)
'i
'CV
CS '
Figure 4.
Idealized oedometer test results.
swelling pressure was derived by Nelson, et al. (2006). The equation presented in Nelson, et al. (2006) for that relationship is given in Equation 1.
cv i cs i
(1)
where i the applied stress at the time of inundation; a constant; and the other terms have been defined above. The value of was seen to be approximately equal to 0.6 from the data presented in Nelson, et al. (2006). 2.2 Competency of laboratory personnel
700
Vertical Stress, kPa (1 kPa = 20.9
Figure 2. Data from Method A (Fig. 3 from The Standard).
Note 2 in The Standard states the precision of this test method is dependent on the competence of the personnel performing the test and the suitability of the equipment and facilities used. Although this is a relatively obvious comment, lapses in the competence of the personnel frequently occurs when
642
communication is not well established between the engineering staff and the laboratory technician staff. Depending upon the nature of the testing and the purpose to which the data is to be put, variations in the testing procedures may be required. It is important for the engineering staff to be knowledgeable in oedometer testing, and to be fully involved in the laboratory testing that is being conducted. Laboratory technicians should not be the sole decision maker with respect to the laboratory procedures. This became evident several years previously when one of the authors of this paper was interviewing various consulting engineers during preparation of the book Nelson & Miller (1992). During one interview the engineer who was responsible for preparing geotechnical investigation reports, and for doing design calculations did not know what inundation pressure was being used in the laboratory. The sole laboratory procedures were dictated by the technicians in the laboratory. 2.3 Heave prediction
as is done in Method A, these samples could all be tested at the same value of i and the value of CH determined from Equation 1. Then the four different test results could be examined to evaluate the variability of the soil at the site. 2.4 Covering of specimen during initial loading In the Denver swell test, the specimen is initially loaded to the inundation pressure and allowed to remain under that stress for a period time until compression of the sample stops. This is necessary to allow consolidation of the sample under the applied load and to provide for closing of microcracks and fissures that have developed as a result of stress release. During this initial loading, The Standard indicates that plastic wrap, aluminum foil or moist filter paper should be used as a loose fitting cover to enclose the specimen so as to avoid water content loss. The authors have conducted a series of experiments in which the sample was covered using a loosely fitting plastic wrap. Water contents were compared with results of another sample prepared and loaded without any covering. The uncovered sample was observed to lose as much as 3 to 4% water content during a period of 4 to 6 hours. The sample that was covered with the plastic wrap had a negligible water content loss during the same period. The importance of this procedure is evident. 2.5 Selection of inundation pressure
Method A of The Standard indicates that 4 or more identical specimens should be assembled for testing. That Method has the advantage that it provides a direct means of establishing the relationship between heave and applied inundation stress. The equation for predicting heave of a layer of soil of thickness zi is given by Equation 2. CH zi log cv i vo i 1
n
(2)
where, free-field heave; CH heave index; zi thickness of the layer; cv constant volume swelling pressure; and vo overburden plus applied stress at the midpoint of the layer. The constitutive parameter, CH, is the slope of the line in Figure 3 which happens to be the same as the slope of the line from percent swell to cv in Figure 4. That method may work well for samples that are reconstituted and compacted into the oedometer ring using specified procedures. That methodology does not work well for insitu samples, because it is virtually impossible to obtain four identical specimens from the field. However, for undisturbed samples taken from the field, it is recommended that the slope of the line shown in Figure 4 be used. If the relationship given by Equation 1 is used, CH can be determined from a single test. Therefore, if one were to test 4 samples
The Standard generally calls for the specimens to be inundated under a stress equal to or close to the overburden stress or the stress to which the sample will be loaded after construction. However, the authors recommend that all samples should be tested at the same inundation pressure. It is common in soils reports in Colorado to indicate the risk of damage due to expansive soils at a site in terms such as low, moderate, high, and very high. Nelson, et al. (2007) define a term entitled expansion potential, EP, that express the expansion potential of the soil in a term that includes both the swell pressure and the percent swell of a sample. The expansion potential, EP, is given by Equation 3,
EP %S log
cv i
(3)
The original equation for expansion potential presented in Nelson, et al. (2007) was subsequently revised in the presentation of the paper to provide for comparison of samples tested at different inundation pressures.
643
Rather than having to correct for the inundation stress, testing of the sample at the same inundation pressure will provide for a more accurate comparison between different samples. The authors generally use an inundation stress of 48 kPa. This corresponds to the overburden stress at the normal depth of a house basement. 2.6 Time of inundation
Consolidation/Swell (%)
Section 12.3 of The Standard specifies the sample be inundated until primary swell or collapse volume change is completed and changes in deformation reading for secondary swell/collapse phase is small (typically 24 to 72 hours). In the Front Range area of Colorado, 24 hours is the typical inundation time. Due to the hysteretic effect of wetting soils, the soil sample will not fully saturate due to entrapped air in the pores. Unable to reach full saturation the soil will not exhibit the maximum swell potential. Three tests were conducted on a shale sample, with a liquid limit of 85 and plasticity index of 67, to determine the effect of inundation time and method on the swelling pressure and percent swell. An inundation pressure of 48 kPa was applied for all three tests performed. The first test followed The Standard procedure for inundation, filling the consolidometer base above the height of the porous stone and applied an initial load for 30 hours. For the second test the sample was saturated by inundation similar to The Standard, however the initial load was applied for 40 days before consolidation loads were applied. The third test was partially saturated by water immersion to the sample height for 31 days, then saturated by inundation for 9 days prior to consolidation loads being applied. The swelling pressure and percent swell of the three samples tested are summarized in Table 1. From Table 1, it can be seen that increasing the applied initial load time by 39 days and altering the method of saturation increased the percent swell and reduced the swelling pressure. Increasing the initial load time resulted in an increase of the percent swell by 10.3% and reduction of the swelling pressure by 3.3%. Altering the method of saturation by immersion for 31 days followed by 9 days
Table 1. Results of inundation procedure alteration. Swelling Percent pressure swell (kPa) (%) 409 395 443 3.9 4.3 5.2
of inundation caused the percent swell to increase 33.3% and the swelling pressure to increase 8.3%. Soil samples may not be fully saturated within the recommended inundation time provided by The Standard. Inundation allows air to be entrapped in the pores. Immersing the sample for 31 days prior to inundation is unrealistic. Therefore, it is recommended that samples first be inundated with water and the deformation or swell measurements taken for a minimum of 24 hours. If the sample continues to exhibit significant swell after 24 hours, readings should be continued until the rate of swell becomes insignificant. 2.7 Calibration of apparatus for system flexibility Figure 1 shows the apparatus commonly used in laboratories in the Front Range area of Colorado. Section 7.1.1 of The Standard indicates that consolidometers of high rigidity should be used. The apparatus shown in Figure 1 is not what one would normally consider to be an apparatus of high rigidity. Therefore, calibration of the apparatus and correction for system flexibility is important. Figure 5 shows consolidation-swell test data both with and without the corrections for tests conducted on apparatus of the type shown in Figure 1. As seen in Figure 5, the amount of correction required is greater at the higher stresses. Figure 5 indicates that the percent swell, S%, is not affected by the correction, but the swell pressure can be influenced greatly. For example, in Figure 5 for the sample that exhibited the highest swell pressure consolidation-swell swelling pressure increased from approximately 800 kPa to over 1,000 kPa. Reviews of raw data from several commercial laboratories in the course of ligation investigations have indicated that the calibration for system flexibility is not always performed. As shown in
12 10 8 6 4 2 0 -2 -4 -6 10 100 1,000 Applied Stress (kPa)
B-1 @ 9 m w/o Corrections B-1 @ 12 m w/o Corrections
Inundation test 30 hour inundation 40 day inundation 31 day immersion/9 day inundation
10,000
B-1 @ 9 m w/ Corrections B-1 @ 12 m w/ Corrections
Figure 5. Oedometer test data corrected for system flexibility.
644
Figure 5, this calibration is very important. Failure to perform that calibration may result in significant under-calculation of predicted heave. 2.8 Use of filter paper
If the porous stones used in the oedometer are too coarse soil may be pushed into the stones causing errors. The Standard recommends the use of very fine porous stones such as sintered brass disks to avoid the use of filter paper. However, if filter paper is used, The Standard states in Section 12.1 that dry filter paper should be used. There are particular errors that can be introduced by using dry filter paper. Figure 6 shows the results of oedometer tests conducted on a sample of filter paper in which a steel block was used in place of the soil. It is seen that during inundation of the filter paper and on the first loading significant deformation takes place. Subsequent unloading and reloading cycles show that after about the third loading the hysteresis loop becomes smaller and the load-unload cycles of filter paper are virtually the same for subsequent loadings. Therefore, it is recommended that if filter paper is used, each individual filter paper being used for the test should be calibrated. The filter paper should be inundated and it should be loaded and unloaded at least three times in the calibration apparatus. That calibration should be used in conjunction with the calibration for system flexibility. 2.9 Storage of samples
soils. Section 8.5 of The Standard states that containers for storage of extra samples may be either cardboard or metal and should be approximately 25 millimeters greater in diameter and 40 to 50 millimeters greater in length than the sample to be encased. The release of confinement by extruding the samples from the sampling tubes can lead to spoiling and deterioration of the sample causing significant disturbance. Common sampling techniques for the highly over-consolidated claystones in the Front Range area of Colorado is to use a lined split barrel sampler where brass rings are used as the liners. The use of brass liners eliminates the concern with the influence of rust. Furthermore, drilling fluid or free water should not be used when drilling in expansive claystones or expansive soils. It is recommended that continuous core be taken along with the California samples. The entire core should be retained and taken to the laboratory for review of the soil profiles in comparison to field logs. The core may be kept in cardboard or metal storage containers. However, such cores are usually too disturbed to be used for testing. 2.10 Wetting of subsoils Section 6.1.2 of The Standard presents a discussion of the wetting of subsoils in the field. It is stated that the wetting situation in the field rarely produces inundation; wetting is often caused by water percolation. Insitu water contents and degrees of saturation typically end up being somewhat lower than those caused by inundation in the laboratory. The degree of wetting of the subsoils in the field will depend on many factors. In highly over-consolidated claystone, cracks and fissures will influence the method by which water travels through the soil. Silty and sandy seams in the bedding planes influence lateral migration of waters significantly. Sources of deep subsoil wetting also exist in the form of sandstone lenses/strata and coal seams. Thus, wetting of the subsoils is a complex phenomenon. The entire depth of potential heave can become fully wetted within the design life of a structure depending on the site conditions and the structure. Analyses of the depth of wetting may take different approaches. One is to apply a macro approach where the subsoils are treated as a continuum that essentially averages out the soils properties over the cracks and fissures and the intact blocks of claystone forming the matrix. Alternatively, a micro approach may be used where the water content in the cracks and fissures is modeled separately from that in the blocks. The micro considerations of water migration can explain why low water contents in discrete samples can be observed even under perched water tables or ponds
Section 8.2 of The Standard states that storage in sampling tubes is not recommended for swelling
0 -50 Deformation (times 0.01 mm) -100 -150 -200 -250 -300 -350 -400 -450 -500 10 100 1,000 Applied Stress (kPa) 10,000
First Loading Second Loading Third Loading Fourth Loading Fifth Loading
Figure 6.
Compression of filter paper (MFG, 2003).
645
that have existed for long periods of time. Wetting to depths of 12 meters or more has been observed and documented by Diewald (2003). Without detailed analyses of water movement in the subsoils, it is prudent to assume that the full depth of the potential heave can become wetted. A companion paper being presented in this conference will discuss the analyses of the degree of wetting with depth. 3 CONCLUSIONS
guidelines and pointing out appropriate considerations and test methods. However, deviation from The Standards must be provided as necessary for specific cases. REFERENCES
ASTM D4546-08. 2008. Standard Test Methods for One-Dimensional Swell or Collapse of Cohesive Soils. ASTM International, West Conshohocken, PA, DOI: 10.1520/D4546-08. Diewald, G.A. 2003. A Modified Soil Suction Heave Prediction Protocol: With New Data from Denver Area Expansive Soil Sites: Denver, Colorado. University of Colorado at Denver, M.S. thesis, pp. 191. MFG. 2003. Technical MemorandumCalibration of Consolidometers. Fort Collins, Colorado. October 29. Nelson, J.D. and Miller, D.J. 1992. Expansive Soils: Problems and Practice in Foundation and Pavement Engineering. New York, NY: John Wiley & Son, Inc. Nelson, J.D., Reichler, D.K., and Cumbers J.M. 2006. Parameters for Heave Prediction by Oedometer Tests. Proceedings of the 4th International Conference on Unsaturated Soils. Carefree, Arizona. April, 951961. Nelson, J.D., Chao, K.C., and Overton, D.D. 2007. Definition of Expansion Potential for Expansive Soil. Proceedings of the 3rd Asian Conference on Unsaturated Soils, Nanjing, China. April.
A number of commentaries have been presented regarding The Standard Method of Testing for one dimensional swell of expansive soils. As noted these comments refer primarily to specific considerations related to highly over-consolidated claystones such as are observed in the Front Range of Colorado. Such highly over-consolidated claystones and clayshales have also been observed by the authors in various locations throughout the western United States, in South Africa, and other locales. It is noted that the procedures presented in The Standards do not always provide for proper handling of material of a specific nature and for the purposes for which the test results will be put. The Standards are important for providing
646
Detecting water content changes in sand samples by means of Electrical Resistivity Tomography
R.M. Cosentini, G. Della Vecchia, S. Foti & G. Musso
DISTR, Politecnico di Torino, Torino, Italy
ABSTRACT: The paper presents some preliminary experimental results concerning the use of the electrical resistivity tomography (ERT) technique to estimate the evolution of the water content in unsaturated laboratory samples. ERT relies on the measurement of the electrical potential drop between two electrodes immersed in an electrical field. By using a high number of electrodes, several independent measurements can be made along the boundaries of the body subjected to investigation. Afterwards, a tomographic reconstruction of the collected data allows for the estimation of the spatial distribution of the electrical conductivity. Such a technique has been applied to track the evolution of the water content sand samples undergoing wetting processes under imposed constant net vertical stress in an oedometer. The evolution of the average and local water content has been estimated on basis of the ERT reconstructions. Results of these experiments show the potentiality of this technique as a subsidiary tool for geophysical applications and for the interpretation of laboratory tests on unsaturated soils. 1 INTRODUCTION Moreover, the use of electrical methods is particularly appealing for the indirect determination of water content (Samoulian et al. 2005). In recent years, experimental studies to relate the electrical conductivity to the water content of homogeneous samples have been performed in the laboratory by Kalinsky & Kelly (1993), Dalla et al. (2004), Attia et al. (2008) and others. As for water content changes, Michot et al. (2003) applied ERT in situ to detect 2-D delimitations of soil horizons and to monitor soil water movement. More recently, Batlle-Aguilar et al. (2009) used ERT to monitor the infiltration of water in the vadose zone of a silty loam deposit. Those results appear promising for the quantitative use of the technique in characterizing the transport properties of unsaturated aquifers. Within this context, this paper presents some preliminary results of tests where ERT has been used to detect temporal changes of local water content in laboratory samples, showing the potentiality of the technique as an aid for the interpretation of both geophysical prospection and laboratory geotechnical tests. 2 EXPERIMENTAL TECHNIQUE AND EQUIPMENT
Electrical Resistivity Tomography (ERT) is a technique of common application in geophysical site investigations, generally adopted for the reconstruction of spatial distribution of the electrical conductivity, in order to obtain stratigraphic information. In the geotechnical field, standard electrical measurements at the laboratory scale have been used for instance to detect clay sample anisotropy (McCarter et al. 2005) and to monitor sedimentation processes (Blewett et al. 2001). ERT applications have evidenced the capability of the technique in visualizing liquefaction (Jinguuji et al. 2007), in individuating different kinds of heterogeneity (Borsic et al. 2005) and in monitoring diffusion processes of a saline solution (Comina et al. 2005, Damasceno et al. 2009). The previously mentioned works are characterized by the study of one-dimensional or two-dimensional geometries and allow for a limited analysis of spatial variability of the electrical conductivity. Such limitation has been overcome by an oedometer cell instrumented for the execution of threedimensional electrical tomography tests (Comina et al. 2008). The cell has been used for monitoring the evolution of the electrical conductivity during consolidation and in salt diffusion tests on homogeneous and heterogeneous samples. Those experimental results confirmed the ability of the technique in the spatial-temporal reconstruction of the evolution of the electrical conductivity field due to variation of chemo-mechanical conditions.
Tests described in the present work were run in the modified oedometer cell illustrated by Comina et al. (2008). The device allows for spatial and temporal monitoring of electrical and seismic properties
647
under controlled mechanical and hydraulic conditions (Fig. 1). The cell has a diameter of 130 mm and it can host samples with maximum height of 60 mm. For ERT measurements, 42 electrodes are hosted on the internal boundary of the cell, 16 equally-spaced on the sidewall and 13 on each base. The execution of conductivity measurements for tomography requires the injection of current at two electrodes, while other pairs of electrodes measure the induced electrical potentials. The measurements are used to reconstruct the electrical conductivity field within the sample. The procedure of reconstruction is carried out with a least-squares algorithm with a Tikhonov regularization inversion technique (Borsic et al. 2005). The drainage system of the cell is composed by three concentric non-conductive rings with a few microns of tolerance between them, posed on each base (Fig. 2). The equivalent permeability of the drainage system has been estimated to be about 6 10 6 m/s. Water pressure and water volume changes are measured by two transducers connected to the bottom drainage line. In the present work the cell has been used first to characterize the electrical behavior of homogenous specimens. Afterwards, it has been used to monitor
the evolution of soil electrical conductivity in samples of the same material subjected to wetting and water content homogenization processes. Since the electrical conductivity of soils is a function of the ones of its constituent phases (Santamarina et al. 2001), and the electrical conductivity of water depends on the amount of dissolved salts, a 0.5 M KCl solution was used in the characterization and wetting tests to impose constant water electrical conductivity. This choice simplifies the interpretation in terms of saturation. 3 3.1 EXPERIMENTAL INVESTIGATION Characterization of the soil used in the investigation
Experimental tests were run on a silty sand coming from the Ticino River alluvial plan in the NorthWest of Italy. Grain size distribution is 9.1% gravel, 78.8% sand and 12.7% silt, with D60 0.493 mm and D10 0.043 mm. Duplicate samples, prepared by moist tamping at a void ratio e 0.82, were used to characterize the hydraulic and electrical behavior of the material. The retention curve has been determined with a suction controlled oedometer cell (Romero et al. 1995) applying the axis-translation technique. Van Genuchten (1980) relation has been used to interpolate experimental data: Se Sr SrRES 1 RES n 1 ( s ) 1 Sr
m
(1)
Figure 1. Oedometer ERT cell used in the investigation.
where Se and SrRES are, respectively, the effective and the residual degree of saturation and , n, m three fitting parameters. The values of the parameters have been estimated to be 0.11 kPa1, n 3.5 and m 0.71 (Fig. 3). The electrical conductivitydegree of saturation relationship has been determined in the ERT oedometer by preparing homogeneous samples at increasing water contents. Experimental data (Fig. 4) have been fitted with Archies law (1942), which holds for porous media with non conductive solid grains. In particular, under conditions of constant porosity and water salinity, Archies law can be written as:
Srq sat
(2)
Figure 2.
Drainage system at the base of the cell.
Here, and sat are the current and the saturated electrical conductivities. The exponent q is a fitting parameter that takes into account the geometry of the interconnected porosity. For the material used
648
Table 1.
Specifications of the tests. Volume inflow (ml) 90 50 Time of flow (min) 0.67 115 Total time of ERT monitoring (min) 3000 115
Test 1 2
Imposed external pressure (kPa) 50 5
Figure 3.
Water retention curve.
Figure 4. Relationship between electrical conductivity and degree of saturation at constant void ratio.
in this investigation a value q 2.0 has been estimated. This value lays within the typical range for sandy materials (Mitchell & Soga 2005). 3.2 Experimental tests procedures
ment of the top of the sample has been measured by an LVDT transducer, in order evaluate the volumetric deformation. In the following, two of such experiments will be described and interpreted (Table 1). In Test 1, the sample has been wetted by allowing a water volume of 90 cm3 to enter the sample in 40 s under an external water pressure of 50 kPa. After that, the bottom drainage system has been closed and electrical measurements performed while the global water content of the sample remained constant. The spatial distribution of electrical conductivity has been monitored for 3000 minutes. This allowed for reconstructing the local redistribution of water content (homogenization). In Test 2, the inflow of water has occurred at a lower rate: a water pressure of 5 kPa has been applied to the base drainage, resulting in a water inflow of 50 cm3 in 115 minutes. Electrical measurements were performed during the whole wetting process. This allowed for monitoring the saturation process under the imposed flow condition. Measurements taken at different times were interpreted by means of the inversion algorithm, providing electrical conductivity data in space and time. These were then used to infer the local degree of saturation inside the sample by means of equation (2). Since the vertical displacement was negligible for all the tests, the link between degree of saturation and the quantity of water in the soil is straightforward. 4 RESULTS AND DISCUSSION
Wetting tests have been performed on cylindrical samples (height 4 cm, diameter 13 cm) prepared at a degree of saturation Sr 0.2 with the same compaction technique and at the same void ratio used in the characterization. Samples have been left undisturbed in the cell for at least 12 hours after preparation to ensure a uniform distribution of water content. After this phase of homogenization, wetting has been induced by imposing different inflows of water from the drainage at the base of the cell. During the tests a vertical stress of 2.54 kPa has been maintained and the displace-
In geophysical applications of ERT concerning the identification of transport parameters, one of the most debated issues is the so called mass balance problem. Indeed, water content changes estimated in situ through a direct quantitative application of the technique can be erroneous up to a 50% (Binley et al. 2002). Therefore, before attempting any other interpretation of the tests, the consistency between the imposed water content and the reconstructed one at given times has been checked. This has been done referring both to the homogenization and the inflow stages (i.e. with closed and open drainage lines).
649
By knowing the volume of water entering the sample, the average degree of saturation SrAV can be calculated as SrAV Vw Vsample (3)
where Vw is the volume of water inside the sample and Vsample the overall volume of the specimen. The average value of the reconstructed electrical conductivity AV* is
AV *
Vsample Vsample
dV
1 Vsample
e Ve
e 1
Ne
(4)
being e and Ve the electrical conductivity and the volume of each element used for the solution of the inverse problem and Ne the number of these elements. Inverting equation (2), the reconstructed average degree of saturation SrAV* has been calculated as SrAV * AV * sat
1/ q
Figure 6. Test 1Time evolution of conductivity (left) and degree of saturation (right) for a longitudinal cross-section: a) t 0 min, b) t 10 min, c) t 110 min, d) t 3000 min.
(5)
Figure 5 shows a pretty good consistency between the experimental and reconstructed values of the average degree of saturation for different tests performed in the same apparatus. Tests results have then been analyzed and interpreted in terms of the local evolution of electrical conductivity and saturation degree. With reference to Test 1, Figure 6 shows the tomographic reconstructions and the interpretation in terms of degree of saturation of a longitudinal
Figure 7. Test 1Estimated saturation degree along the symmetry axis.
Figure 5. Comparison between imposed and reconstructed average degree of saturation within the sample.
section before drainage opening (time t0 0 min), and during homogenization (t1 10 min, t2 110 min and t3 3000 min). Figure 7 reports the values of degree of saturation calculated along the axis of symmetry (r 0 cm). For increasing times, saturation degree tends to become uniform, leaning toward the average value Sr 0.58. The homogenization process is clearly shown by the decrease of the local water content in the zone close to the drainage system (z approximately between 0 and 1 cm) and by the corresponding increase for z greater than 1 cm. Time evolution of the degree of saturation along the diameter of the sample at two different heights (z 0.5 cm and z 3 cm) is shown in Figure 8. The represented times are t 0 min, t 10 min, t 270 min and t 510 min. From the figure, it is evident the greater water content characterizing the points at z 0.5 cm (continuum line),
650
Figure 8. Test 1Evolution of the saturation degree along the width of the sample for two different heights: z 3 cm (filled symbols, dashed line) and z 0.5 cm (empty symbols, continuum line).
Figure 11. Test 2Evolution of the saturation during infiltration: saturation contours at z 2 cm for t 115 min.
due to proximity of the base. Local maxima of saturation degree are found above the drainages lines (whose location is shown at the bottom of the figure). Again, for increasing times, water distribution tends to homogenize both in the vertical and the horizontal directions. For Test 2, isosaturation lines at three different times belonging to the infiltration stage are shown in Figures 9, 10, 11. Maps refer to a transversal section at the middle height of the sample (z 2 cm). Reconstructions show increasing values of electrical conductivity for increasing times, with higher values in correspondence of the center of the sample.
CONCLUSIONS
Experimental tests have been performed in order to check the ability of 3D ERT technique in monitoring water content changes in laboratory sandy samples. The investigation aimed at characterizing the dependency of electrical conductivity on saturation degree, and thereafter at verifying the possibility to infer the global and local water content evolution through ERT. As for the first aspect, Archies law was found to model satisfactorily the electrical behavior of the investigated soil. Its parameters, determined on the basis of ERT measurements, are in line with the range provided in the available literature. Imbibition and water content homogenization tests, performed under different hydraulic boundary conditions, were interpreted in terms of estimated average and local saturation degree. Consistency between the average reconstructed degree of saturation and the real value has been found.
651
The spatial and temporal evolution of water content has been monitored in two tests. The aim of the first one has been to check the evolution of local water content after the abrupt introduction of a certain amount of water. Repeated measurements allowed for correctly track the progressive homogenization of the water content. In the other test, the progressive increase in saturation at modest flow rate has been monitored. Reconstructions allowed for evidencing the shape of the developing saturation pattern, influenced by the drainage position. Laboratory ERT appear to be a useful tool to support in situ measurements for problems involving the distribution of water in the soil. In particular the proposed methodology can furnish preliminary laboratory calibrations and qualitative and quantitative interpretations useful to link electrical measurements with soil characteristics. This investigation suggests as well that ERT can be a support in laboratory geotechnical investigations. Among other data, it can provide valuable information on the actual homogeneity of samples, both inherent and induced. As a further step, laboratory 3D ERT reconstruction could be used as a basis for inverse analysis and estimation of soils hydraulic parameters in unsaturated conditions. The temporal and spatial evolution of reconstructed values of electrical conductivity during the inflow of water in the sample could be used as comparison values for the results of numerical simulations performed with a suitable discretization method. ACKNOWLEDGEMENTS This work has been developed as part of the research project SOILCAM, financed by the EU commission within the 7th Framework Program. REFERENCES
Archie, G.E. 1942. The electrical resistivity log as an aid to determining some reservoir characteristics. Trans AIME 146: 5463. Attia, A.M., Fratta, D. & Bassiouni, Z. 2008. Irreducible water saturation from capillary pressure and electrical resistivity measurments. Oil & Gas Science and Technology 63 (2): 203217. Batlle-Aguilar, J., Schneider, S., Pessel, M., Tucholka, P., Coquet, Y. & Vachier, P. 2009. Axisymetrical Infiltration in Soil Imaged by Noninvasive Electrical Resistivimetry. Soil Science Society of America Journal 73: 510520. Binley, A., Cassiani, G., Middleton, R. & Winship, P. 2002. Vadose zone flow model parameterisation using
cross-borehole radar and resistivity imaging. Journal of Hydrology 267: 147159. Blewett, J., McCarter, W.J., Chrisp, T.M. & Starrs, G. 2001. Monitoring sedimentation of a clay slurry. Gotechnique 51 (8): 723728. Borsic, A., Comina, C., Foti, S., Lancellotta, R. & Musso, G. 2005. Imaging heterogeneities with electrical impedance tomography: laboratory results. Gotechnique 55 (7): 539547. Comina, C., Foti, S., Lancellotta, R., Musso, G. & Borsic, A. 2005. Imaging heterogeneities and diffusion in sand samples. Proceedings of the 11th Int. Conf. of the International Association of Computer Methods and Advances in Geomechanics IACMAG2005 vol.2: 2734. Comina, C., Foti, S., Musso, G. & Romero, E. 2008. EIT oedometeran advanced cell to monitor spatial and time variability in soil. Geotechnical Testing Journal, ASTM 31 (5): 404412. Dalla, E., Cassiani, G., Brovelli, A. & Pitea, D. 2004. Electrical conductivity of unsaturated porous media: pore-scale model and comparison with laboratory data. Geophysical Research Letters 41. doi:10.1029/2003GL019170. Damasceno, V.M., Fratta, D. & Bosscher, P.J. 2009. Development and validation of a low-cost electrical tomographer for soil process monitoring. Canadian Geotechnical Journal 46: 842854. Jinguuji, M., Toprak, S. & Kunimatsu, S. 2007. Visualization technique for liquefaction process in chamber experiments by using electrical resistivity monitoring. Soil Dynamics and Earthquake Engineering 27: 191199. Kalinsky, R.J. & Kelly, W.E. 1993. Estimating water content of soils from electrical resistivity. Geotecnical Testing Journal 16 (3): 323329. McCarter, W.J., Blewett, J., Chrisp, T.M., & Starrs, G. 2005. Electrical property measurements using a modified hydraulic oedometer. Canadian Geotechnical Journal 42: 655662. Michot, D., Benderitter, Y., Dorigny, A., Nicoullaud, B., King, D. & Tabbagh, A. 2003. Spatial and temporal monitoring of soil water content with an irrigated corn crop cover using surface electrical resisitivity tomography. Water Resources Research 39 (5). Doi: 10.1029/2002 WR001581. Mitchell, J.K. & Soga, K. 2005. Fundamentals of soil behavior. John Wiley and Sons: New York. Romero, E., LLoret, A. & Gens, A. 1995. Development of a new suction and temperature controlled cell. In E. Alonso & P. Delage (eds), 1st Int. Conf. on Unsat. Soil, ParisFrance, 2: 553559. Samoulian, A., Cousin, I., Tabbagh, A., Bruand, A. & Richard, G. 2005. Electrical resistivity survey in soil science: a review. Soil & Tillage Research 83: 173193. Santamarina, J.C. in collaboration with Klein, K.A. & Fam, M.A. 2001. Soils and Waves. New York: John Wiley and Sons. van Genuchten, M.Th. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44: 892898.
652
Calibration of a TDR probe in an expansive soil

Stephen Fityus
Centre for Geotechnical and Materials Modelling, The University of Newcastle, New South Wales, Australia
Kalman Rajkai
Research Institute for Soil Science and Agricultural Chemistry, Budapest, Hungary
ABSTRACT: The measurement of soil water content in situ is not straightforward. Capacitance methods such as Time Domain Reflectometry (TDR) can be used to estimate the volumetric water content in situ using a passive probe which is embedded in a soil and connected via a cable to a signal processor/ recorder. In this study, a systematic series of laboratory measurements was undertaken to determine the relationship between true volumetric water content, gravimetric water content and TDR measured water content in uncracked Maryland clay. The results show not only that the TDR measured water contents differ from the true volumetric water contents, but that the TDR is potentially relatively insensitive over a significant interval of mid-range water contents in uncracked clay. Tests to estimate the effect on measurements from the presence of cracks of different width found that errors increase steadily as crack width increases. The study also determined that the measurements are relatively insensitive to misalignment of the waveguide prongs. 1 INTRODUCTION This involves making an assessment of its ability to accurately measure water contents in uncracked clay soils, and an assessment of errors that may result from the presence of cracks, or difficulties in installation. In particular, the research evaluates the performance of the TRASE System I (Soil Moisture Corporation, 1993) Time Domain Reflectometry (TDR) probe (product no. 6050XI) is tested, using a scanned 20 cm length buriable waveguides (product no. 6005 L2) (Evett and Heng, 2008). 2 BACKGROUND
The measurement of water content in situ is a challenging problem in any case, but it becomes all the more difficult when a soil is dominated by expansive clay minerals. In such cases, the dry density of the soil varies as its volume changes in response to water content changes, and so, there is no unique relationship between water content and void ratio, and no unique saturated water content. Most of the methods available for the measurement of water content in situ are indirect methods. That is, they measure some particular index property of the soil mass surrounding the measuring device, and from this measurement, infer a water content via a calibration equation. As this property is measured within some volume of soil, the measurements are usually derived as volumetric water contents: that is, they express the volume of water as a fraction or percentage of the total volume. Perhaps the more common methods for in situ water content determination are the neutron moisture meter, or neutron probe, and the socalled capacitance methods. Research has already determined that it is extremely difficult to calibrate the neutron probe (either theoretically or experimentally) to give reliable measurements in cracked expansive clay soils. (Fityus, et al., 2010) The purpose of the research presented here was to assess the effectiveness of a capacitance method in measuring water content in desiccated clay soils.
The TRASE TDR system uses measurements of the apparent dielectric constant of the soil, by analysing the time characteristics of the reflected response of an electrical pulse delivered to a threepronged waveguide, which is embedded in a soil. The TRASE analysis unit is a black box in that it returns a value of water content directly, according to its own pre-programmed analytical procedure and empirical calibration relationship. Because water content is linearly related to the quantity being measured for most soils, a twopoint calibration generally works well. However, it is relatively well established that TDR probes are unreliable, as clay soils shrink causing gaps to open between the waveguide and the soil. Figure 1 (Fityus, et al., 2010) shows a record of TDR
653
Figure 2. The TRASE system 1 TDR equipment used in this study. Figure 1. Continuous TDR readings over a 6 month period (Fityus and Wells, 2010). Table 1. Basic data for the Maryland expansive clay soil. (some data from Fityus et al. (2004), Fityus and Smith, (2004) and Moe et al. (2003)). Property Value 69.8 24.1 45.7 14 5.2 2.623 3940 3436 5.2 60 2/15
measurements in a basaltic clay. It can be seen that the readings fail when the water content drops below about 80%. It has also been reported that even though TDR probes have been used for monitoring the drying out of soil, they have proved erroneous even when the soil is close to saturation (Walker et al., 2004). The TDR waveguides can be installed in situ in a number of different ways. They can be pushed directly into bulk soil, if the stiffness of the soil is not so great that it prevents this; they can have neat holes pre-drilled and be inserted; or they can be placed in larger drilled holes and backfilled. Allman et al., (1998) have reported that the readings of TDR probes in expansive clay soils are sensitive to the method of installation. 3 3.1 APPROACH Soils and equipment
Liquid limit (%) Plastic limit (%) Plasticity index (%) Linear shrinkage (%) Shrink-swell index (% strain/pF) Gs Effective field saturated wc @ 0.35 m (%) Effective field saturated wc @ 0.9 m (%) Air-dry wc (%) Typical crack spacing (mm) Average/Maximum crack width (mm)
3.2 Reliability of measurements in uncracked soils Before considering the effect of cracks on TDR readings, it was necessary to produce an experimental calibration for the uncracked soil. To overcome the obvious difficulty that expansive soils crack as they are dried, it was decided to conduct this work on remolded samples. This approach had the added advantage that the samples could be thoroughly homogenized during the preparation process, and so, parallel samples would be the same in every way except for the water content at which they were remolded. A calibration was performed on the remolded uncracked soil by compacting a known mass of soil with a known volume into a known volume. This allows toe true volumetric water content to be determined, and this can be compared with the water content measured in the sample by the TDR. 3.3 The effect of cracking
As noted in the Introduction, this study is focused on the TRASE System 1 TDR equipment for the measurement of soil moisture. There is no particular reason for this other than the equipment was available for the study. The buriable waveguides for this device are 200 mm long, with three co-planar prongs, 3.16 mm in diameter and spaced at 22 mm, centre to centre, so that the waveguide unit has an overall width of around 47 mm. The TRASE TDR equipment is shown in Figure 2. The soils chosen for the study is the Maryland expansive clay. This soil was chosen because it has been the focus of many detailed studies into expansive soil behaviour (Fityus and Smith, 2004, Fityus et al., 2004), and so, there is already a wealth of associated information and data, to serve as a context for the present results. A summary of relevant data is presented in Table 1.
To assess the effect of cracking on TDR measured water contents, remolded samples were split and
654
the TDR probes placed loosely in the space created. As it is possible that crack geometries could take any form, and so affect the measured water content in any number of ways, it was necessary to choose a systematic crack form. An arrangement where the probe was positioned in the plane of a parallel-sided crack was adopted. 3.4 The effect of waveguide alignment
When the pronged waveguide is pushed into stiff soils, there is a tendency for the prongs to deviate from their ideal parallel arrangement. It is unclear what the effect of this is on the readings obtained. To assess the effect of prong out-of-alignment, a series of tests was carried out with the prong tips deliberately spread apart by various amounts. This was also done using split samples. 4 4.1 EXPERIMENTAL DETAILS Reliability of measurements in uncracked soils
Figure 3. Compacted sample with waveguide (left) and split sample with waveguide removed (right; unshaved).
On the basis that the field of influence around the waveguides extends out little more than 10 mm, it was decided that the tested samples could be remolded in heavy-walled PVC tubing, 235 mm long and 100 mm in diameter. This would ensure a soil cover of at least 25 mm over the length of the prongs, and 35 mm over the ends of the prongs. The soil sample was firstly conditioned to a high moisture content and left to equilibrate by breaking it into lumps of around 1020 mm and sealing it into a plastic drum for several days. The samples for testing were then created by compacting the soil into the PVC cylinder using a standard Proctor compaction hammer. Sample was added to the cylinder a small amount (9 or 10 lumps) at a time, and then compacted until it appeared that there were no macrovoids (i.e., no voids large enough to see). More soil was then added and compacted, and so on, until the cylinder was full. From measurements of the tube mass, with and without the compacted soil, the mass of the compacted soil was determined. Also, from measurements of the tube dimensions, the volume and the soil bulk density was determined. Subsequently, the volumetric water content could also be determined (Evett and Heng, 2008). A wave guide was then inserted into the sample while still in the tube (Figure 3, left). For the samples with higher water content (28%), waveguides were simply inserted by pushing. As the sample became drier and stiffer, the wave guides were then pushed in using the mechanical advantage of a hand press. As the samples became hard (water contents less
than about 24%) holes for the waveguide prongs needed to be pre-drilled. This was done using an extended 3.18 mm (1/8") twist drill, mounted in a drill press so that it drilled truly vertical holes. The holes were drilled to around 205 mm deep to allow for the small amount of scrapings that are pushed forward by the prongs as they move down the hole. Because of the neat fit, it was still necessary to use the press to insert the waveguide into the predrilled holes. It should be noted that during this process, up to 5 kN of force needed to be applied, and using a special device to transmit this pressure to the head of the probe, the probes proved to be robust enough to be used and reused without damage. In the 13 tests performed, only two probes were damaged and had to be discarded. At least 10 TDR measurements were made on each sample, in the tube. In general, the results returned were very consistent for a given sample, varying within a range of about 2% in any particular set of readings. At the end of measurement, a wire saw (tight, high-tensile wire) was pushed through the sample along one face of the prongs, effectively splitting it in two, and allowing the waveguide to be retrieved (Figure 3, right). This allowed the sample to be subsampled over its entire depth so that an average water content could be determined. After this, the soil was chopped up into 1020 mm lumps and dried down to a new target mass/moisture content. Then, it was left to equilibrate for several days before the whole sample preparation and testing cycle was repeated. A total of 13 tests were carried out, at gravimetric water contents from 15.5 to 42.2%. 4.2 The effect of cracking
The tests to assess the effect of cracking on TDR measured water contents used the same samples
655
as in the non-split assessment. After the samples had been split by cutting with a tight wire, and the TDR waveguide removed, the halves of the sample were shaved down so that smooth, flat surfaces were created. Then, the TDR waveguide was sandwiched in between the two halves so that there was even contact on each side, and readings taken. For softer samples, different crack widths were considered, by inserting spacer-plates (thinner than the waveguide prong diameter) with the waveguide, in between the two halves, and using firm pressure to push the halves together, so that the prongs of the waveguide were partially embedded in the faces of the block. This was repeated with successively thinner spacer-plates, until there was no spacerplate at all, and the waveguide was completely embedded in a block without a crack. 4.3 The effect of waveguide alignment
60
50
Volumetric WC (%)
40
30
20
TDR Measured Theoretical
10
Actual Poly. (TDR Measured)
0 0 10 20 30 40 50 Gravimetric WC (%)
The assessment of mis-aligned waveguide prongs was also carried out on the same samples used for the uncracked soil calibration. As with the assessment of crack width, the shaved split samples were used, this time with spreading of the waveguide prongs before sandwiching it in between the sample halves. In all cases, the misaligned waveguides were tested embedded in a crack of 3.18 mm wide: that is, no attempt was made to embed the waveguides into the soil. 5 5.1 RESULTS Calibration for uncracked expansive clay soil
Figure 4. Comparison of TDR measured and theoretical volumetric water content values with the actual volumetric water content.
The results of the calibration of the TDR in uncracked expansive soils are shown in Figure 4. These comprise the volumetric water content as measured by the TDR and the volumetric water content as determined from sample measurements, plotted against the gravimetric water content. The results of Figure 4 show that, despite the soil having no cracks, the TDR was unable to reliably predict the volumetric water content of the soil. Moreover, there are also potential problems with empirical calibrations for this particular soil, in that the relationship determined is unexpectedly flat between the significant gravimetric water contents of 20 and 35%, and hence, it is relatively insensitive: the TDR measurement at w 24.4% is 44.8, while the TDR measurement at w 35% is only 45.6. This suggests that in this heavy clay soil, the TDR is a poor discriminator of water content. Also shown in Figure 4 are the theoretical predictions of volumetric water content corresponding to particular values of gravimetric water content,
assuming that the clay soil is fully saturated and that the particle specific gravity for Maryland is 2,62, as measured by Micromeretics AccuPyc II 1340 V1.00. The results suggest that the recompacted clay soil remains effectively saturated at all water contents between about 15% and 40%, and that at a particle aggregate level, there is no air entry down to a water content of 15%, despite the development of substantial suctions (around 6 MPa according to Fityus and Smith, 2004). 5.2 The effect of cracking
The results for measurements in soils of particular water content with simulated cracks is shown in Figure 5. It appears from these results that the variation in readings obtained at any particular water content is as large as, and in some cases greater, than the variation in an uncracked soil due to water content. The effect of cracks is generally similar for soils of any water content. 5.3 The effect of waveguide alignment
The waveguide alignment effect was studied in samples of two different water contents, with waveguide prong spacings from about 2 mm to 35 mm. The results are shown in Figure 6. From Figure 6, it appears that TDR readings are relatively insensitive to the alignment of the
656
60.0
prongs of the waveguide, with no significant trend in measurements as the spacing is increased or decreased from the design spacing of 22 mm. 6 CONCLUSIONS
50.0
TDR water content (%)
40.0
30.0
20.0
w=19.2% w=24.4% w=26.5%
10.0
w=28.4% w=32.0% w=38.0% w=40.0% 0 1 2 3 planar crack width (mm) 4
0.0
Figure 5. The effect of crack width on TDR readings in expansive clays.
This study confirms that TDR methods of moisture measurements are generally unsuited to expansive clay soils. It appears that the magnitude of the effect that cracks have on measurements is of the same order of magnitude as the variations in the readings themselves. The problematic response due to cracks is compounded by the insensitivity of the method in heavy clay soils. This outcome was unexpected, since the significant variation in water contents in clay soils should be large enough to significantly affect the TDR response in a systematic way. More generally, however, the research confirms that some mis-alignment of the TDR probes is unlikely to further affect the reliability importance of device. Perhaps the most significant outcome from the research is the observation that in compacted expansive soils, the degree of saturation is likely to remain close to one over most of the significant range of water content change, despite the development of very large suction stresses. This has particular significance for volume change models in expansive soils. REFERENCES
Allman, M., Delaney, M. and Smith, D.W. 1998 A field Study of seasonal ground movements in expansive soil. Proceedings of the second international conference on Unsaturated Soils, Beijing. International Academic Publishers. Evett, S.R. and L.K. Heng. Conventional Time Domain Reflectometry Systems. In Field Estimation of Soil Moisture Content. A Practical Guide to Methods, Instrumentation and Sensor Technology. IAEA., Vienna, 2008. pp. 5571. Fityus, S.G., Smith, D.W. and Allman, M.A. 2004 An expansive soil test site near Newcastle., ASCE Journal of Geotechnical and Geoenvironmental Engineering Vol. 130, No. 7, 686695. Fityus, S.G. and Smith, D.W. 2004 The development of a residual soil profile from a mudstone in a temperate climate, Engineering Geology Vol. 74, No. 1, 3956. Moe, H, and Fityus S.G. Smith, D.W., 2003 `Study of a cracking network in a residual clay soil. In the Proceedings of Unsat Asia 2003, the 2nd Asian Unsaturated Soils Conference, Osaka. pp. 149154. Soil Moisture Corporation 1993. TRASE System 1 Operating Instructions. Walker, J.P., G.R. Willgoose, and J.D. Kalma. 2004. In situ measurement of soil moisture: a comparison of techniques. J. Hydrology. 293: 8599. doi:10.1016/ jhydrol2004.01.008.
30.0
25.0 TDR volumetric water content (%)
20.0
15.0 perfectly aligned 10.0
5.0 19.20% 24.40% 0.0 0 10 20 30 40 waveguide spacing (mm)
Figure 6. The effect of mis-alignment of TDR waveguide prongs on readings in expansive clays.
657
Effective thermal conductivity of partially saturated soils

S.S. Hamuda & M. Rouainia B.G. Clarke
School of Civil of Engineering & Geosciences, Newcastle University, Newcastle upon Tyne, UK School of Civil Engineering, The University of Leeds, Leeds, UK
ABSTRACT: The purpose of the present work is to develop a reasonable-cost, simple, and robust laboratory apparatus for determining the thermal properties of soils that can be replicated by others in geotechnical engineering field. The apparatus is then used to investigate the behaviour of thermal properties of partially saturated soils. The results of a series of laboratory tests have shown that below a certain degree of saturation (5% for the tested sand), the moisture in the porous started to migrate from the hotter region to the cooler region in shorter durations compared to higher degrees of saturations which affected the thermal steady state condition. 1 INTRODUCTION and the probe methods are in particular require much less time. The GHP method and the other methods that use the principle of the longitudinal heat flow at steady state condition, such as the unguarded hot plate method, the heat flow meter method and the comparative thermal conductivity methods are developed for structural and insulation materials, and have limited applications for soils. Other experimental testing programmes have been performed to measure the thermal conductivity of soils using the thermal single and dual probe methods and line heat source method (Abu-Hamdeh 2003; Nusier and Abu-Hamdeh 2003; Penner, Johnston, and Goodrich 1975; Abuel-Naga, Bergado, and Chaiprakaikeow 2006; Cote and Konrad 2005) among others. In partially saturated soils, the transport of heat is complicated by the fact that temperature gradients cause moisture movement, which will tend to redistribute itself when the temperature field changes (Pintado, Ledesma, and Lloret 2002). The moisture movement, which occurs both in the liquid and in the vapour phases, gives rise to a transport of sensible and latent heat, which again influences the temperature redistribution. A quantitative treatment of the combined transport of heat and moisture in porous media is complicated (De Vries, 1966). This phenomenon was comprehensively analysed by Philip and De Vries (1957). The authors have also made comparisons between the outcome of experiments and predictions using the developed theory. Reasonable or even good agreement between theory and experiments was found in many cases but there were some exceptions. Thomas and Sansom, (1995) made a fully
Migration of moisture will occur in partially saturated soils due to thermal gradients. Thermally induced moisture migration is a complex phenomenon involving the interaction of several physical mechanisms. These include heat conduction, latent heat, heat flow, vapour diffusion, and capillary induced liquid flow. Water movement occurs in natural soils either as liquid or vapour. In addition to heat transfer, water movement alters the soil properties because of the changes in the amount and type of water at particular position (Farouki 1986). Soil water exists in the form of an adsorbed film whose thickness is changing insignificantly until development of water rings occurs. This is followed by formation of bridges between contact points of the soil grains (Tarnawski and Leong 2000). These water bridges improve the heat transfer between the particles. A number of tests exist for the measurement of thermal conductivity of building materials. These include, the comparative, the guarded hot plate method, the hot wire method, the thermal needle test and the unguarded hot plate method described in the (ASTM 1997; BS 1988; Hanson, Neuhaeuser, and Yesiller 2004). Experimental methods of measuring thermal conductivity are steady state methods and transient state methods. The guarded hot plate method (GHP), the hot wire and probe methods are the most accepted in the industry for soils thermal conductivity measurements. The GHP method is generally regarded as accurate. However, it is usually quite time consuming. Compared to the GHP test, the hot wire
659
coupled analysis of heat, moisture and air in a partially saturated soil. 2 APPARATUSES DESIGN AND TESTING PROCEDURES
Steel studs Acrylic cover Aluminum sink disc Vaccum gap Inner and outer acrylic tubes Heater disc with heater opening at middle DC supply wires for heater and thermocouples wires go here Vaccum tube Aluminum sink disc Acrylic cover Studs holes
The purpose of the present apparatus is to find the thermal properties of soil. Several past and current experimental devices were studied and taken into consideration before attempting to design the new apparatus. The most important requirement in designing the new thermal conductivity cell is to use the soil samples obtained in routine investigations. This requirement limits the size of the specimen to the tested to 103 mm diameter. One dimensional heat flow is used in most of the methods of determining thermal conductivity so that Fouriers law of one-dimensional heat conduction can be applied once steady state conditions are established. The new thermal cell uses this principle. A thermal gradient is generated through the cylindrical soil specimen obtained from a U 100 sampler parallel to the axis of the cylinder. A heat source having the same cross sectional area of the specimen is used to create a uniform temperature across one end of the specimen thus creating the thermal gradient. A disk of aluminium with a relatively high conductivity is used to dissipate the heat from the other end of the specimen. This is known as the heat sink disk. Practically, there may be radial heat flow because it is impossible to create a boundary that is completely heatproof. Therefore to reduce the radial heat losses in this test, specimens were insulated with a material which has a low thermal conductivity material. This was further reduced by using a double wall cylinder with the air between the cylinder walls being evacuated. This cylindrical double walled vacuum tube was used to confine the specimens, the heater and the sink plates to limit the radial heat loss and create conditions of one-dimensional flow. A twin cell was developed so that the heater disk sits in between two specimens of soil. If the specimens are identical, potential heat losses are limited to radial heat loss only. Further by making the diameter to thickness of the heat disk as large as possible most of the heat will flow into the specimens. The thickness of the heat disk is a function of the thickness of the heater cartridge inserted into the heat disk. If two identical specimens of soil are used then it can be assumed that the heat flowing through each specimen is the same. In practice this will never be the case because of the variability of soil but for practical purposes it is assumed that remoulded specimens and specimens from the same sample will have similar structures, water content and density.
Figure 1. Thermal cell apparatus parts.
From the above discussion the main components for the new thermal conductivity testing cell is drawn schematically as shown in Figure 1. 2.1 Specimen preparation The method of preparation of specimens is dependent on the type of samples to be tested. Samples are classified into two main types, disturbed samples and undisturbed samples, according to their sampling techniques. U100 samples can be inserted directly into the cell. Disturbed samples are prepared directly inside the thermal cell by either compaction or deposition. Disturbed samples can either be fine grained soils (silts and clays) or coarse grained soils (sands and gravels). The water content of the soil is very important as it determines the method of setup or remoulding of the specimen in the thermal cell. Water content is also a major factor in the heat transfer mechanism. Dry and fully saturated specimens are considered two phase soils which mainly exhibit heat transfer by conduction; partially saturated soils are three phase soils which could involve conduction, latent heat and convection heat transfer mechanisms. Three-phase soils will be investigated in this paper. 2.2 Test set-up and procedure After the specimens were prepared and transferred to the thermal cell the volume of the specimens in the thermal cell should be precisely maintained by adjusting the two screws that held the sink disk in
660
place. The contact between the heater disk and the specimen should be visually checked through the transparent acrylic tubes to avoid any trapped air pockets which may create a high contact resistance to heat flow. In order to guarantee good contact between the heater and sink disks and the specimens a gentle pressure was applied to the top aluminium sink disks using the screw bolts at the steel holding plate to press the aluminium sink disks towards the specimens. 2.3 Measurement uncertainties
There are a number of errors that could occur during a thermal conductivity test. The following are those that are related to the test specimens preparation and test set-up: The dimensions of the specimens; The position of thermocouples; The weight of soil and water; Significant radial heat transfer due to vacuum leakage; Heat contact resistance due to entrapped air between specimen and the heater and sink disks; The measurement of temperature and power. In order to take into account the radial heat loss which cannot be eliminated completely, the temperature along the specimens will be measured in more than one position, and corrections will be applied for the resulted thermal conductivities. A typical configuration is to put one at end of the specimen adjacent to the sink disk and other two thermocouples positioned at one third and two thirds of the length of the specimen. 2.4 Test procedure
studs to avoid any vacuum leakage during the test. In order to achieve approximately the same density for both specimens, the weight of soil and the length of specimens should be the same. The first soil specimen is then placed in the upper chamber in the thermal cell. In case of dry coarse grained samples the soil was funnelled into the thermal cell directly. Once the specimen reached the correct length the sink disk was fixed in place. The sink plate is then placed on top of the specimen by the aid of the guide screws. The cell was inverted and the procedure repeated for the other specimen. The same procedure should be followed for both specimens to produce identical conditions. Thermocouples were then pushed into the specimens to the desired positions. A typical configuration is to put one at end of the specimen adjacent to the sink disk and other two thermocouples positioned at one third and two thirds of the length of the specimen. The temperature for the constant temperature chamber adjusted via the temperature controller to the desired ambient testing temperature. The complete testing set-up, as shown in Figure 3, is then allowed to reach equilibrium. This can be monitored graphically in a separate file using the Pico-logger software. After equilibrium has been reached, the DC power supply was switched on to feed the heater with power. The quantity of power, q in watts is controlled by changing the voltage and or the current by the controls in the DC power supply. The power was maintained until the steady state condition is reached. The input temperature, T2, and the output temperature, T1, are then recorded and plotted. At the steady state condition, the thermal conductivity coefficient could be then computed according to Fouriers law as follows: ke ql 2 A(T2 T1 ) (1)
The thermal cell is assembled as shown in Figure 2. It is important to carefully tighten the
Steel holding plate To vaccum pump Specimen 1 Specimen 2 Thermocouples
where l is the length of the specimen and A is its cross-sectional area.

Pico-Logger Constant temperature chamber Bulb Fan DC source Vacuum pump Thermal cell Temperature controller AC
heater
Figure 2.
To DC source
Assembled thermal conductivity cell.
Figure 3. Complete thermal conductivity set-up.
661
EXPERIMENTAL TESTING
Temperature C
28 26 24 22 20 18 16 0 20 40
Six experiments were designed to measure the effective thermal conductivity of medium-grained sand specimens with different initial water contents. The thermal cell apparatus at steady state condition was used. It should be noted that the steady state condition for a partially saturated soil, subjected to thermal gradient, is relatively hard to maintain for a long time particularly at low degrees of saturation. The average temperature of the tested specimens was kept as low as possible (ranging from 23C to 26C) in order to avoid the early evaporation of the moisture in the soil pores. This was to inhibit the redistribution of the water within the soil specimen until a temperature steady state could be attained. Several initial water contents were used to investigate the influence of water content (through degree of saturation) on the effective thermal conductivity of the soil. Initial water contents of 1%, 2%, 3%, 5%, 10% and 15% were used to prepare soil specimens. In order to eliminate the effect of soil dry density on the effective thermal conductivity, all tested specimens were prepared at the same porosity. Moreover, to keep the global water content of the specimens constant, thermal grease was applied in the clearance space between the aluminium sink discs and the inner acrylic wall of the tube. The temperature versus time relationship was used as an indication of any changes in the initial water content. When the water content changes at locations between thermocouples due to soil heating, the temperature at any thermocouple will change. This indicates that the water has started to redistribute itself within the soil specimen. The time required to start the redistribution of water is assumed to coincide with the end time of the steady state condition for that particular water content and temperature gradient. It should be acknowledged that this steady state condition was not stable if the initial water content was less than 1%. 3.1 Influence of initial water content
33 mm from top 1 66 mm from top 1 top of specimen 1 heater 1 heater 2 top of specimen 2 33 mm from top 2 66 mm from top 2 60 80 100 120 Time (hrs) 140 160 180 200
Figure 4. Temperature versus time curves for the 1% assumed initial water content.
Figure 4 shows the temperature versus time curve for 1% moisture content. It is obvious that the heater temperature started to increase continuously after a time of about 10 hours from the start of the test. This indicates that the moisture started to move from the hotter to the colder end of the specimen. This process continued up to the time of termination of test, which was about seven days. The other thermocouples placed at distances of 33 mm and 66 mm away from the heater disc, respectively, were showing a decrease in temperature because of the movement of water away from the heater. The increase in water content in
these locations increased the thermal conductivity of these segments, which is located between the 33 mm and 66 mm of soil, and resulted in a temperature decrease. The thermal equilibrium was expected to be established some 10 hours from the start of the test. In contrast, the hydraulic equilibrium due to the migration of water was expected to take a very long time compared to the thermal equilibrium. (Pintado, Ledesma, and Lloret 2002) found that hydraulic equilibrium in a partially saturated clay soil was achieved after 61 days compared to about 10 hours for the thermal equilibrium. It is obvious that at low water contents (less than about 1%) the heat transfer problem should be treated as transient and coupled heat and moisture analysis should be performed in order to accurately study the thermal properties of soils. For the sake of simplicity, a steady state condition was assumed to be reached at 10 hours and Fouriers law was applied to obtain the effective thermal conductivity of the soil. From Figure 5 it can be noted that, the steady state conditions were very clearly reached for initial water contents above 2%. At a water content less than 1%, the evaporation of water from the sand pores adjacent to the heater disc produced a very thin dry soil zone which reduced the thermal conductivity of that zone. This increased the temperature in that zone. The thickness of the zone increased gradually as water moved away from this zone. This phenomenon continued for up to seven days when the test was terminated. At the end of a test the specimens were allowed to cool down to the ambient temperature and the specimen was cut into five equal portions (cylinders) in order to determine the water content. The differences observed in the assumed initial water content and the average water content at the end of the test can be related to water loss during the mixing operation while preparing the specimens.
662
28 26 Temperature C 24
18 16 14 12
Ration Rw
35 40 45 50
22 20 18 16 0 5
33 mm from top 1 66 mm from top 1 top 1 heater 1 heater 2 top 2 33 mm from top 2 66 mm from top 2 10 15 20 25 30 Time (hrs)
10 8 6 4 2 0 0 0.01 0.02 0.03 0.04 0.05 0.06 0.07 0.08 0.09 0.1
Distance from heater (m)
Figure 5. Temperature versus time curves for the 2% assumed initial water content.
Figure 6. Water content distributions along the axial axis of the tested specimens.
2 1.8 1.6 1.4 1.2 1 0.8 0.6 0.4 0.2 0 0 5 10 Average initial water content (%) 15 Change in intial moisture content Medium sant (n = 0.364)
It has been observed that the water content distribution after seven days for a specimen that started with a 1% water content increases as the distance from the heater disk increases. This indicates that a marked water content gradient developed. It is important to mention that the water distribution is only representative of this particular time and could take other forms before or after this time. Figure 6 shows slight differences in the water content between the hot and cold edges. Those water content differences, between hot and cold edges (wc h) were estimated to be about 0.5%, 1.5% and 2.87% for initial water contents of 3%, 1% and 10%, respectively. In general the difference in water content between the hot and cold ends of the specimens were increasing as the initial water content increases. From the above discussion, it can be concluded that the change in the initial water content due to a temperature gradient through the longitudinal axis of the specimen increases with the increase in initial water content. The large change at 1% initial water content could be explained by relatively rapid movement of water. Another indicator for the water migration from the hot to the cold edges of the specimen was used. This indicator represents the ratio (Rw wc-h/wavg) of the difference in water content between the heater and sink edges (wc-h) to the average water content (wavg) of the specimen. As the ratio Rw increases the chance for the water redistribution increases, and the water in the pores migrates away from the hot edge in shorter times. This indicator was decreasing as the initial water content increases, as shown in Figure 7. The lowest ratio Rw in this figure corresponds to about 1% initial water content (about 5% degree of saturation). From this indicator, it can be observed that the soil below the 5% degree of saturation undergoes water redistribution in shorter times. This may explain the non applicability of Kerstens and Johansens empirical equations (Johansen 1975; Kersten 1949) below 5% degree of saturation for these types of soils.
Figure 7. Ratio of difference in water content between hot and cold ends to average initial water content.
The redistribution of water through the specimens results in a variation of effective thermal conductivity coefficient along the length of a specimen, which increases from the colder end towards the hotter end. As the water content of the specimen decreases the differential temperature between the specimen ends will be higher, hence the chance of early evaporation of water from the pore spaces increases. This clearly indicates that the initial water content of the soil affected the water redistribution, the temperature profile and hence the heat transfers through the soil. The experimental results of average effective thermal conductivity for the different initial water contents tested was evaluated and compared with Kerstens empirical equations as given in Figure 8. the figure shows that the experimental effective thermal conductivity was in good agreement with Kerstens empirical equation at degrees of saturation greater than 5%, while at lower degrees of saturation (less than 5%) the experimental values were higher than that predicted by the Kerstens equation. At a 1% initial water content, the deviation was high and exceeded by 44.3% that predicted by Kerstens empirical equation.
663
5.0 ke = 0.0008S2 + 0.097S + 0.7342 4.0 ke (WmC) 3.0 2.0 1.0 Kerstens emperical equation Experimental Poly. (Experimental) R2 = 0.9118
20
40 60 Degree of saturation S (%)
80
100
Figure 8. Experimental effective thermal conductivity compared to Kerstens equation. Table 1. Assumed wc (%) initial 0 1 2 3 5 10 15 Results of effective thermal conductivity. Average wc (%) after test 0.00 0.88 2.06 3.01 4.61 9.78 15.01 S (%) 0 4.53 10.51 14.81 22.93 48.25 76.97 ke Experim. (W/mC) 0.647 1.193 1.573 1.77 2.79 3.145 3.57 ke Kerstens (W/mC) NA 0.66 1.47 1.94 2.54 3.34 3.81
redistribution at any particular time. This helps in studying the interaction between energy piles with the surrounding soils, especially in energy piles used as cool sinks; where the water in the soils pores tends to escape away from the pile perimeter. A dry zone soil will develop near the pile perimeter, which affects the overall heat transfer from the pile to the surrounding soil. This phenomenon, in addition, can explain the difficulties in reaching steady state conditions in some thermal response tests in sites where the field soils are in low degrees of saturation. REFERENCES
Abu-Hamdeh, N. (2003). Thermal properties of soils as affected by density and water content. Biosystems Engineering 86(1), 97102. Abuel-Naga, H., D. Bergado & S. Chaiprakaikeow (2006). Innovative thermal technique for enhancing the performance of prefabricated vertical drain system. Geotextiles and Geomembranes 24(6), 359370. ASTM (1997). C518-98 standard test method for steady state heat flux measurements and thermal transmission properties by means of the heat flow meter apparatus. In Amer. Soc. for Testing Mat., pp. 211216. BS (1988). 874-2-2 determining thermal insulating properties (unguarded hot-plate method). In British Standards. Cote, J. & J.-M. Konrad (2005). Thermal conductivity of base-course materials. Canadian Geotechnical Journal 42(1), 6178. Farouki, O. (1986). Thermal properties of soils, Volume 101. Series on Rock and Soil Mech. Hanson, J., S. Neuhaeuser & N. Yesiller (2004). Development and calibration of a largescale thermal conductivity probe. Geotechnical Testing Journal 27(4), 393403. Johansen, O. (1975). Thermal conductivity of soils. Ph. D. thesis, University of Trondheim, Trondheim, Norway. Kersten, M. (1949). Laboratory research for the determination of the thermal properties of soils. In ACFEL Tech., Volume 23. University of Minnesota, Minneapolis. Nusier, O. & N. Abu-Hamdeh (2003). Laboratory techniques to evaluate thermal conductivity for some soils. Heat and Mass Trans. 39, 119123. Penner, E., G. Johnston & L. Goodrich (1975). Thermal conductivity laboratory studies of some mackenzie highway soils. Pintado, X., Ledesma & A. Lloret (2002). Back analysis of thermohydraulic bentonite properties from laboratory tests. Engineering Geology 64(2-3), 91115. Tarnawski, V. & W. Leong (2000). Thermal conductivity of soils at very low moisture content and moderate temperatures. Trans. Porous Media 41, 13747.
Table 1 shows the experimental results for the effective thermal conductivity of the sand and the predicted values by the Kerstens empirical equations. 4 CONCLUSIONS
Below a certain degree of saturation (5% for the tested sand), the moisture in the porous started to migrate from the hotter region to the cooler region in shorter durations compared to higher degrees of saturations which affected the thermal steady state condition. According to the above discussion, short-term transient techniques are preferable in principle for partially saturated sands with degrees of saturation less than 5%, to avoid the time it takes to reach steady state since that will change the thermal conductivity of the soil due to moisture migration. Steady state techniques can be used successfully for the determination of dry and saturated soils thermal conductivities as well as sands having degrees of saturation greater than 5%. For partially saturated soils, the apparatus could be used successfully to simulate the water content
664
A new unsaturated direct shear apparatus for measuring shear strength of unsaturated soils Part 1: Apparatus features and test procedures
E. Hefzi, S.S. Yasrebi & A. Mirzaii
ABSTRACT: In many geotechnical problems, one of the most common shear strength testing equipment is the direct shear apparatus to study shear strength behaviour of soils. This paper represents features of a new unsaturated direct shear apparatus designed and built in Tarbiat Modares University. The new apparatus enables using axis translation technique to control or measure soil matrix suction. This apparatus includes an air pressure chamber equipped with a shearing box inside. A particular pressure panel has been designed for the control and application of apparatus devices. The apparatus is also equipped with separate horizontal and vertical loading components. Horizontal loading device includes a hi-precision electro servo motor system, while vertical loading device is equipped with a pneumatic jack. A data acquisition system was designed to measure and acquire the variation of test data. Six conventional direct shear tests were then carried out on compacted unsaturated kaolin-sand mixtures. These tests were conducted in constant water content condition in different net normal stresses. The axis translation technique was used to measure or control soil matrix suction. 1 INTRODUCTION A controlled suction unsaturated direct shear apparatus was built by Escario (1980) to perform a series of shear strength tests. As shown in figure 1, a direct shear test box was placed inside a chamber.
Direct shear test is one of the most common laboratory and in situ geotechnical testing apparatus and has been in practical application for assessment of saturated shear strength behaviour of geomaterials since previous decades. Due to the simplicity, successful representation of certain soil behaviour, reasonable sample thickness for easy measurement of pore pressures, and consistency of the results with data attained from other tests, direct shear testing become also of interest to study shear strength behaviour of partially saturated soil media. However, common direct shear apparatuses should be approved for operation of unsaturated plane strain shear tests on porous soil media. As the early works, Donald (1956) carried out a series of unsaturated direct shear tests on fine sands and coarse silts with a modified direct shear box with separate control of pore air and water pressures during the shear. The top of the direct shear box was exposed to the atmospheric pressure while the pore water pressure was controlled at a negative value by application of a constant negative head to the water phase. The specimen was placed in contact with the water in the base of the shear box through the use of a colloidon membrane. The water in the base of the shear box was then connected to a constant head overflow tube at a desired negative gauge pressure.
Figure 1. Controlled suction unsaturated direct shear apparatus (Scario 1980).
665
A pressurized nitrogen flow was applied to the upper part of the soil sample through a coarse grained porous stone. The lower face of the sample was in contact with a high air entry value porous stone or a semi permeable membrane for higher values of the suction. Moreover, Gan et al. (1988) built a controlled suction direct shear apparatus and imposed desired values of matrix suction by the axis translation technique. A high air entry disk was glued to the down part of shear box for the control and measurement of pore water pressure. Vertical loads were applied to the specimen with an axial ram while a motor caused horizontal displacement with a constant rate for the application of horizontal shear loads. In a direct shear test, the sheared zone is very localized and thin, as compared to the mass of the unsaturated sample, and the application of axis translation technique has prosperous results for direct shear tests in unsaturated soils. Thus, this technique might not be suitable for investigation of unsaturated soils at high degree of saturation or for high values of matrix suction. Other techniques of controlling suction have also been applied to direct shear testing. The osmotic technique was adapted by Kassiff & Ben Shalom (1971) and took the advantage of the osmotic potential generated by a concentration gradient across a semi-permeable membrane to measure high values of matrix suction up to 10 MPa. Thus, this technique also has some major drawbacks such as corruption of the membrane in contact with soil, water evaporation, and difficulty in accurate calibration in particular for low suction values. As most remarkable, Boso et al. (2005) built a direct shear apparatus to measure suction by osmotic method. A porous plate was placed underneath the semi permeable membrane (on the osmotic solution side) to provide a rigid support for the sample. In addition, tensiometers were also installed on the loading pad for separate direct suction measurement. During the previous years, unsaturated direct shear tests popularity increased owing to its advantages for suction measurement. When analyzing the status of suction in an unsaturated sample submitted to direct shear stress under a slow shearing rate, it seems highly probable that a regulation of the suction in the sheared zone could be done by the mass of soil, since the volume of water to be exchanged to maintain suction equilibrium in the whole sample is small. Provided that the shearing rate is low enough, it seems highly probable that a slow constant water content direct shear test would also ensure a condition of constant suction. Thus, common problems associated in saturated direct shear tests (e.g. stress concentrations, definition
of the failure plane, rotation of principal stresses, and mandatory shear zones) were also appeared in unsaturated direct shear tests. This paper represents diverse features of a novel unsaturated direct shear apparatus designed and built in Tarbiat Modares University. The new apparatus enables using axis translation technique to control or measure soil matrix suction and includes a direct shear chamber, separate horizontal and vertical loading systems, control panel, and data acquisition system. A series of six constant water content direct shear tests were carried out on unsaturated compacted kaolin-sand mixtures to assess shear strength and mechanical behaviour of understudying soil. In these tests, soil samples were initially brought to the matrix suctions of 30 and 100 kPa and were in parallel consolidated in net confining stresses of 50, 100, and 200 kPa. At the end of equilibrium and consolidation stage, horizontal shear forces were applied to the specimens and were continued until they reached to the failure state. 2 APPARATUS FEATURES
The general layout of new unsaturated direct shear apparatus is shown in figure 2 and includes different components of direct shear chamber, horizontal and vertical loading systems, control panel, and data acquisition system. The details of each part are described in detail in the following sections. 2.1 Direct shear chamber The main part of the apparatus is the direct shear chamber where the soil specimen is placed inside and different load and boundary conditions can be applied to the specimens (figure 3). The chamber is constructed with aluminum to reduce the apparatus weight. The shear chamber is equipped with a set of shear box with a 5 bar HAED. A narrow
Figure 2. General layout of unsaturated direct shear apparatus.
666
Figure 3. The unsaturated direct shear chamber and shear box.
ceramic disk (Figure 4).Due to the thin thickness of ceramic disk, the water in the HAED compartment was regularly flushed (typically two times a day and also prior horizontal shearing) to remove any possible air bubbles from the water compartment. The shear boxes were made of stainless steel to eliminate from any possible erosion of shear boxes. HAED disk were mounted in the bottom shear box with a durable and water resistant epoxy resin. The shear loads are applied to the soil specimens within the horizontal movement of bottom shear box, and shear stresses are measured within the upper box. The confining air pressure and chamber air exhaust conduits are installed in the top of shear chamber. The vertical loading ram and pneumatic jack are also installed in the top of shear chamber. A glass-made window is installed in the front side wall of shear chamber to provide easy observation of soil sample and interior of the shear chamber during the tests. In addition, horizontal loading devices are also installed in the side walls of the chamber (figure 2). 2.2 Horizontal and vertical loading systems In this apparatus, two distinct electronic and pneumatic loading systems were used for the application of horizontal and vertical loads respectively (figure 2). A hi-precision electronic servo motor is used to induce horizontal shear forces. This device is connected to a programmable logic controller unit to produce different movement paths with desired speeds. A particular software was written for the PLC unit to regulate the horizontal movement path (e.g. monotonic or cyclic movement) and the speed of servo motor. This motor is connected to a helical gear box to increase the domain of loading speed and to reduce back lashes, particularly in small shear rates. The horizontal loading component of the apparatus is capable of the application shear loads within the loading speed of 0.0001 mm/min to 9.0 mm/min. The vertical loading device of the apparatus includes a pneumatic jack with the maximum sustainable capacity of 3.5 kN. A calibration table was prepared for appropriate application of desired vertical stress to the soil samples. 2.3 Control panel Control panel component behaves as the brain of the system and is designed for the control and application of pressure and volume changes in different part of the apparatus. In addition, a vacuum and de aired-water preparation system is also installed in the control panel. The general layout of control panel is shown in figure 5.
Figure 4.
Bottom shear box and its water compartment.
hole is put in the bottom of the chamber for the entrance and exit of de aired water for the saturation of HAED. Horizontal loads application and measurements instruments are installed in the side walls of the shear chamber while vertical loading ram is installed in the top of the chamber. The air and water pressure and volume change conduits of the soil specimen are installed in the top and side walls of the apparatus chamber. Square soil specimens with the dimension of 60 * 60 mm can be put in the shear box of direct shear chamber for assessment of shear strength and mechanical behaviour with respect to different boundary conditions (Figures 3 and 4). The shear box is composed of two upper and bottom moving shear boxes, and a thin high air entry stone is installed in the bottom moving shear box with the air entry value of 5 bars and the thickness of 4 mm. The thin value of HAEDs thickness significantly affected the response time of pore water control and measurement system. A continuous square water compartment is created beneath the HAED to control and measure the variation of pore water pressure and volume changes. The water compartment is both connected to the water pressure and volume change burettes and also DAVI conduits for the removal of accumulated air bubbles under the
667
Figure 5.
General layout of control panel device.
The application of pressures in different parts of the apparatus is carried out within the regulation of elevated air pressure by accurate diaphragm regulators. There air regulators are directly connected to a hi-precision transmitter, and the value of regulated pressures are shown in the digital indicator of control panel with the accuracy of 0.1 kPa. The pore water pressure and volume changes are controlled with a Perspex twin burette installed in the left side of control panel. The capacity of pore water pressure burette is 25 cm3 with the appropriate accuracy of 0.05 cm3. A vanturi valve is used to alter positive air pressure to desired values of negative air pressure. This negative air pressure was induced in the top of de-aired water tank in the control panel to remove trapped air in the matrix structure of water molecules. Typically, de-aired water is prepared with the application of 80 kPa vacuum pressure in the top of de-aired water tank for at least 48 hours. 2.4 Data acquisition system In most of geotechnical testing and in particular for unsaturated soils, test periods are typically long and the application of a convenient and reliable data acquisition system is mandatory. In this apparatus, data from different stages of unsaturated direct shear tests were monitored, measured, and acquired with a particular data acquisition system. This part of the apparatus included different components of electronic transmitters, a data logger, and a data acquisition software. Accurate pressure transmitters are used to monitor and assess the variation of pressures during the tests with an appropriate linearity and repeatability to measure pressures with the accuracy of 0.5 kPa. Two load cells were installed in the hori-
zontal and vertical loading systems to observe and acquire the variation of shear and normal forces with reasonable accuracy of 0.5 N. Horizontal and vertical displacements of soil samples were measured with two potentiometers with the displacement accuracy better than 0.01 mm. A particular 16 bit data logger instrument is designed to acquire data attained from electronic transmitters. This component of data acquisition system is made of 12 channels for the measurement of discharged signals from the transmitters. Data processing and acquisition are carried out with a particular acquisition software. This software enables the definition of different variables (e.g. matrix suction, net normal stress, shear stress, axial strain) based on primary input data from electronic transmitters. In addition, the software enables live observation of desired variables during the tests. The calibration of transmitters is also carried out with the acquisition software. 3 TEST PROCEDURE
A set of constant water content unsaturated direct shear tests is carried out to assess mechanical behaviour of compacted kaolin-sand soil samples in plane strain loading condition with the novel direct shear apparatus. In these tests, soil samples were brought to the initial matrix suctions of 30 and 100 kPa and were consolidated in vertical net stresses of 50, 100, and 200 kPa in the meanwhile. At the end of equilibrium and consolidation stage, horizontal shear loads were applied to the samples and were continued until they reached to the steady state. Further details of sample preparation and test procedures are described in the following sections. 3.1 Apparatus preparation
Prior the beginning of unsaturated direct shear tests, the ceramic disk installed in the bottom shear box was properly saturated. For this purpose, the shear chamber was filled with de-aired water up to the ceramic disk and then the air pressure inside the chamber was increased to 700 kPa (more than the HAEDs air entry value) and was maintained for 24 hours to achieve full saturation of ceramic disk. After the saturation of HAED, a constant water flow was discharged in the water compartment with a low pressure to remove any possible occulated air bubbles underneath the HAED. 3.2 Preparation of soil samples
The soil tested was a mix of 60% sand and 40% kaolin. The classification properties of the soil are given in Table 1.
668
Table 1. Classification properties of soil. Soil type: (unified system) Liquid limit % Plastic limit % Plasticity index % Specific gravity (GS) Clay percent % Optimum water content* % Maximum dry unit weight* (kN/m3) * From Standard proctor compaction test. Table 2. Physical properties of compacted soil samples. Wet unit weight (kN/m3) Dry unit weight (kN/m3) Initial Sr % Moisture content % Void ratio Porosity 18.81 17.17 49.11 9.6 0.52 0.342 SC 23.5 14 9.5 2.66 40 9.58 19.71
Table 3. Phase-state and stress-state values at the end of equilibrium and consolidation stage. Net confining stress* 50 100 200 50 100 200 Matrix suction 33.5 31.1 30.5 101.1 101.1 99.07 Sr % 65.3 66.7 71.6 56.3 56.6 60.8 e 0.508 0.499 0.463 0.513 0.505 0.470 wet (kN/m3) 19.46 19.60 20.06 19.12 19.21 19.70
Soil samples were brought to the matrix suctions of 30 and 100 kPa by using axis translation technique and by application of elevated pore air and water pressures. Compacted soil samples appeared to have greater initial matrix suction than 100 kPa and consequently a flow of water to the samples decreased the initial matrix of samples to the desired values. In parallel to equilibrium, vertical net confining stresses were applied to the samples and excess pore air and water pressures let to be drained. Equilibrium and consolidation of samples were typically longed between 34 days and the variation of pore air and water volume changes become constant. Table 3 shows the values of phase-state and stress-state of different samples at the end of equilibrium and consolidation stage. 3.4 Horizontal shear loading
Figure 6.
Preparation and mounting soil samples.
The soil samples were compacted and brought to a dry density that was 87.11% of the maximum dry density achieved in the Standard Proctor compaction test. The physical properties of soil samples are given in Table 2. After preparation of samples, remained water above the HAED removed before installation of the soil sample on the disk and consequently specimens were put above the saturated ceramic appropriately. A porous disk and vertical rigid loading plate were then placed above the sample (Figure 6). 3.3 Equilibrium and consolidation
After the installation of soil sample and setting up the apparatus, elevated pore air and water pressures were applied to the specimens to bring them to the desired values of matrix suction. During the equilibrium, specimens were consolidated in different net confining stresses in the mean while.
At the end of the equilibrium and consolidation stage, soil samples were horizontally loaded in constant water content condition and were sheared until they reached to the failure. During the shear, the variation of pore water pressure, matrix suction, horizontal and vertical displacements, and horizontal and vertical stresses were independently measured until the failure condition. In the shear stage, pore air let to be drained while the pore water was in undrained condition. The application of horizontal shear forces altered the value of pore water pressure and consequently the value of soil matrix suction. Typically, the value of water pressure in the pore spaces tended to increase within the application of horizontal shear loads and consequently decreased and became steady in the failure condition. This behaviour caused an initial decrease in the value of soil suction and reversely an increase in further until the soil suction reached to a constant value in the failure condition. Thus, the pore air let to be drained and soil samples tended to have total volume changes in
669
Table 4. Maximum sustainable shear stresses in constant water content direct shear tests. Net confining stress (kPa) 50 100 200 50 100 200 *Initial matrix suction (kPa) 33.5 31.1 30.5 101.1 101.1 99.07 Maximum shear stress (kPa) 58 83.8 140.8 79.8 105.4 161.9
* At the end of equilibrium and consolidation stage.
the shear stage. In all tests, soil samples tended to have a slight downward vertical settlement and became compacted at the beginning of shearing and with the continual of shear load application reversely tended to dilate and had upward vertical settlements. Horizontal shear loads were applied to the samples with the constant speed of 0.009 mm/min to have consistent reading of matrix suction in constant water content tests and were continued until the samples reached to the failure. The values of maximum sustainable shear stresses are shown in table 4 for samples in different matrix suction and net confining stress. 4 CONCLUSION
also installed in the control panel. In this apparatus, data from different stages of unsaturated direct shear tests are acquired with a particular data acquisition system composed of electronic transmitters, data logger, and data acquisition software. A series of six constant water content direct shear tests were carried out on unsaturated compacted kaolin-sand mixtures to assess mechanical behaviour of understudying soil in plane strain loading condition. In these tests, soil samples were initially brought to the matrix suctions of 30 and 100 kPa and in the meanwhile were consolidated in net confining stresses of 50, 100, and 200 kPa. At the end of equilibrium and consolidation stage, horizontal shear forces were applied to the specimens and were continued until they reached to the failure state. ACKNOWLEDGMENTS The authors are thankful for the financial support of Tarbiat Modares University for the construction of unsaturated direct apparatus. Kind discussions and guidance of Mr. Ardeshir Aminzadeh are also appreciated in mechanical design of the apparatus. REFERENCES
Boso, M., Tarantino, A. & Mongiov, L. 2005. A direct shear box improved with the osmotic technique. In A. Tarantino, E. Romero & Y.J Cui (eds), Advanced Experimental Unsaturated Soil Mechanics, Proc. of an International Symposium, Trento, 2729 June 2005. Rotterdam: Balkema. Donald, I.B., 1956. Shear strength measurements in unsaturated non-cohesive soils with negative pore pressures. In proceeding 2nd Australia-New Zealand Soil Mech. Found. Eng., Christchurch, New Zealand. Escario, V. 1980. Suction controlled penetration and shear tests. Proc. 4th Int. Conf Expansiue Soils II. Denver: American Society of Civil Engineers. Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics for unsaturated soils. New Jersey: John Wiley and Sons Pub. Gan, J.K.M., Fredlund, D.G. & Rahardjo, H. 1988. Determination of the shear strength parameters of an unsaturated soil using the direct shear test. Canadian Geotechnical Journal 25(3): 500510. Hefzi, E. 2009. Evaluation of shear strength of unsaturated soils by direct shear apparatus. M.Sc. Thesis. Tarbiat Modares University. Kassiff, G. & Ben Shalom, A. 1971. Experimental relationship between swell pressure and suction. Gotechnique 21(3): 245255. Ng, C.W.W. & Menzies, B. 2007. Advanced unsaturated soil mechanics and engineering. London: Taylor and Francis Group. Ning Lu & William, J. Likos 2004. Unsaturated soil mechanics. New Jersey: John Wiley and Sons.
In this context, diverse features of a novel unsaturated direct shear apparatus were introduced and a series of constant water content direct shear tests were carried out to assess mechanical behaviour of understudying soil in plane strain loading condition. The novel direct shear apparatus is designed and built in Tarbiat Modares University. This apparatus enables using axis translation technique to control or measure soil matrix suction and includes a direct shear chamber, separate horizontal and vertical loading systems, control panel, and data acquisition system. The main part of the apparatus is the direct shear chamber which is equipped with a set of upper and bottom shear boxes. A 5 bar HAED is mounted in the bottom shear box for the control and measurement of pore water pressure and volume changes. Horizontal and vertical loading system is composed of two distinct electronic and pneumatic loading devices respectively. The control panel component of the apparatus behaves as the brain of the system and is designed for the control and application of pressure and volume changes in different part of the apparatus. In addition, a vacuum and de aired-water preparation system is
670
Constitutive modeling of unsaturated soil behavior using a refined suction-controlled true triaxial cell: Preliminary observations
L.R. Hoyos, D.D. Prez-Ruiz & A.J. Puppala
University of Texas at Arlington, Arlington, Texas, USA
ABSTRACT: This paper presents the results from a comprehensive series of hydrostatic compression, conventional triaxial compression, and triaxial compression tests conducted on compacted specimens of unsaturated clayey sand under constant-suction states. The experiments were conducted using a straincontrolled true triaxial (cubical) apparatus that is suitable to test 3-in (7.5-cm) per side cubical specimens of unsaturated soil under controlled-suction states, and for a wide range of stress paths that are not achievable in a conventional cylindrical apparatus. The cell is a considerably refined version of the stresscontrolled systems previously reported by Hoyos et al. (2001, 2005). Target suction levels are induced and kept constant during testing using the axis-translation technique. Results from suction-controlled tests under axisymmetric conditions were used for calibration and further fine-tuning of the elasto-plastic, critical state-based framework postulated by the original Barcelona Model. Numerical predictions of soil response proved to be reasonably accurate, highlighting the potential of the critical state-based framework for use in the analysis of geotechnical boundary-value problems involving unsaturated soil deposits.
INTRODUCTION
The adoption of matric suction, s (ua uw), and the excess of total stress over air pressure, ( ua), as the relevant stress state variables, have facilitated the modeling of key features of unsaturated soil behavior via suction-controlled oedometer, triaxial, and direct shear testing (Alonso et al. 1990, Wheeler and Sivakumar 1992, Fredlund and Rahardjo 1993). The majority of these devices, however, allow for the application of loads along limited paths and/or modes of deformation, such as one-dimensional, hydrostatic or axisymmetric loading. In nature, pavement subgrades and shallow foundation soils well above the ground-water table may be subjected to three-dimensional stress gradients due to changes in the stress state variables (ij uaij) and (ua uw)ij, as illustrated in Figure 1. Therefore, accurate predictions of stress-strain response of geosystems resting on unsaturated ground require that all constitutive relations be valid for all stress paths that are likely to be experienced in the field. It is in this context that a true triaxial (cubical) cell, capable of inducing in the soil test specimens a wide range of simple-to-complex multiaxial stress paths under controlled-suction states, plays a fundamental role in the thorough stress-strainstrength characterization of this type of materials. This paper describes a servo-controlled true triaxial (cubical) apparatus that has been developed
Figure 1. Idealization of unsaturated soil deposits subject to multiaxial stress gradients.
to test 3-in (7.5-cm) side, cubical specimens of unsaturated soil under controlled-suction states for a wide range of stress paths that are not achievable in a conventional cylindrical apparatus. The
671
new cell is an upgraded, more elaborate version of the apparatus previously reported by Hoyos et al. (2005). Results from a series of suction-controlled tests under axisymmetric conditions (2 3) were used for calibration and further fine-tuning of the elasto-plastic, critical state-based framework postulated by the basic Barcelona Model (Alonso et al. 1990). 2 PREVIOUS WORK
Hoyos and Macari (2001) reported a first attempt to test unsaturated soil samples under constant-suction multiaxial loading. The cell used in this work, however, presented some important limitations, such as a highly corrosive steel frame, occasionally clogging the HAE ceramic disks; and the use of hydraulic oil to pressurize the latex membranes in contact with the sample, which resulted in occasional damage of the membranes when exposed to hydraulic oil for extended periods of time. Matsuoka et al. (2002) reported a rigid-boundary system that induces suction states in the specimens via negative pore-water pressure. This system also presented some limitations, including undesirable boundary effects due to rigid loading platens, reducing the capability of the cell to induce a wide range of stress paths; and the very use of negative pore-water pressure to induce suction states via HAE ceramics, which limits suction application to 100 kPa. More recently, Hoyos et al. (2005) developed a new cell similar to the one reported by Hoyos and Macari (2001), with a few enhanced features. However, the system still presents some important limitations, including low resolution of pressure transducers, which restricts load increments to a minimum of 1 psi (6.9 kPa); manual application of load increments, which limits the application of a continuous, ramped loading scheme; no real-time acquisition of soil deformation data from LVDTs; and occasional corrosion of springs attached to the LVDT extension rods when exposed to water over an extended period of time. 3 A REFINED SUCTION-CONTROLLED TRUE TRIAXIAL APPARATUS
bottom assembly to house a HAE ceramic disk and four symmetrically spaced coarse porous stones, as shown in Figure 2. The ceramic disk, with a 5-bar air-entry value, is saturated in place following a procedure similar to that suggested by Bishop and Henkel (1962). After saturation of the ceramic, the cubical soil specimen is gently slid in through one of the lateral cavities of the frame, with the specimen ultimately sitting right on top of the saturated ceramic disk. After setting of the compacted specimen into the inner cavity of the frame, the remaining five walls are assembled to the frame. The system does not require the use of LVDTs for monitoring soil deformation during testing. A full set of cubical latex membranes transmit the water-based hydraulic pressure to the top and four lateral faces of the specimen. Figure 3 shows a typical lateral wall assembly, including a picture illustrating the stretchability of the latex in a clayey sand sample that has been monotonically stressed to 35% of vertical strain. External hydraulic pressure is generated and controlled by a computer-driven PCP-5000-UNSAT panel, allowing independent control of major, intermediate, and minor principal stresses. A computer-driven PVC-100-UNSAT pressure volume
Figure 2. Bottom wall assembly with sealed HAE ceramic and four coarse stones.
The apparatus presented in this paper is a mixedboundary type of cell, with the specimen seating on top of a HAE ceramic disk and between five flexible membranes on the remaining sides of the cube. The cell consists mainly of a stainless steel frame featuring six pressure cavities to accommodate one top and four lateral flexible latex membranes, and a cubical base aluminum piece at the
Figure 3. latex.
Lateral wall assembly and stretchability of
672
Figure 5. Static compaction of cubical soil specimens.
Figure 4.
Panoramic view of true triaxial test layout.
control system maintains a constant suction on the specimen and allows for determination of soil deformation in all three principal directions. In this panel, three Model DC 7505000 sensors allows for accurate, real-time measurements of both the positive (compression) and the negative (extension) deformations experienced by the soil sample. Figure 4 shows a panoramic view of the entire suction-controlled true triaxial test layout, including the core cell, PCP-5000-UNSAT panel, PVC100-UNSAT panel, and DAS. 4 COMPACTION OF CUBICAL SAMPLES
Figure 6. HC tests at s 100 kPa on SP-SC specimens.
Test soil classifies as clayey sand (SP-SC) according to the USCS, with 70% poorly-graded sand and 30% low-plasticity clay. Cubical soil specimens were prepared via static compaction. The initial water content was kept at 10% while the target dry density was approximately 15.34 kN/m3. A triaxial load frame is then used to apply a quasi-static axial load to the loose sample until achieving the target volume, as shown in Figure 5. Cubical samples were prepared using a custommade stainless steel mold, as shown in Figure 5. The mold is first filled with 761.5 g of SP-SC soil mixed with water to the target 10% moisture. The loose mixed is then compressed by the loading frame at a constant rate of 1.0 mm/min. Immediately after compaction, the specimen is gently extruded from the mold and transferred to the cubical cell. A photograph of a typical statically-compacted sample is also shown in Figure 5. 5 INFLUENCE OF COMPACTION INDUCED DRY UNIT WEIGHT
A series of four drained (suction-controlled) hydrostatic compression (HC) tests were conducted at constant matric suction, s ua 100 kPa, to
experimentally determine the adequate compaction dry unit weight for SP-SC samples using static compaction. The intent was to reproduce identical samples with a relatively small preconsolidation pressure so that it was feasible to bring the test soil to a virgin state, hence inducing elasto-plastic deformations that would help identify the loading-collapse yield curve postulated by Alonso et al. (1990). All four specimens were prepared using static compaction method. Although all the samples were compacted at the same initial water content, w 10%, each sample had enough solid mass to attain a different dry unit weight at the same total volume. The selected dry unit weights, d 19.18, 17.26, 16.30, and 15.34 kN/m3, correspond to 100%, 90%, 85%, and 80% of the maximum standard Proctor dry unit weight d-max, respectively. Figure 6 shows the variation of specific volume, v 1 e, with net mean stress, p (1/3) (1 2 3) ua, from the four suction-controlled HC tests. All tests were carried out from an initial net mean stress of 20 kPa to a final net mean stress of 600700 kPa. Based on these results, a dry unit weight, d 15.34 kN/m3, which corresponds to 80% of maximum Proctor dry unit weight d-max, was found to yield lowest preconsolidation pressure. It is also worth noting that the initial (elastic) portion of all the v-p curves show a reasonably identical slope, including the rebound portion of the curve for d 15.34 kN/m3. Statically com-
673
pacted samples of SP-SC soil yield an average specific volume, v 1 e 1.72, and an average degree of saturation, S 37.5%. 6 REPEATABILITY OF TEST RESULTS IN TRIAXIAL COMPRESSION (TC)
Two triaxial compression (TC) tests were conducted at constant matric suction, s ua 200 kPa, to assess the repeatability of test results and the suitability of the developed apparatus for testing unsaturated soils under suction-controlled conditions via axis-translation technique. Results from the two TC trial tests are shown in Figure 7, which shows reasonably identical trends from both tests. It is also observed that the cell is capable of reproducing the compressive nature () of the major principal strain and the extensive nature () of the minor and intermediate principal strains, as dictated by the loading scheme of the TC stress path. 7 TEST RESULTS AND MODEL PREDICTIONS
(p ua) 50 kPa. Figure 8 shows the initial change in specific volume of the soil sample under sustained suction states, s ua 30, 100, 200, and 350 kPa. From these results, a 100-hr equalization time was deemed appropriate for all suction-controlled tests. During the shearing stages of CTC and TC tests, the soil was subject to a deviator-stress loading rate of 8 kPa/hr, until it was apparent that the soil had reached peak strength (Prez-Ruiz 2009). Figure 9 shows the variation of specific volume with net mean stress from HC tests at s 50, 100, 200, 350 kPa on compacted SP-SC soil. As it is expected, the suction state has a critical influence on the initial soil specific volume prior to initiating the HC test path. Moreover, the slopes of all elastic rebound curves are shown to be reasonably constant.
A comprehensive series of suction-controlled hydrostatic compression (HC), conventional triaxial compression (CTC), and triaxial compression (TC) tests was accomplished on several identically prepared samples of SP-SC soil. Details of the test procedures and results are given by Prez-Ruiz (2009). Samples were prepared via static compaction to a target density, d 15.34 kN/m3. The appropriate pore-fluid equalization time for a given, initial matric suction state was experimentally assessed under a net mean stress,
Figure 8. Assessment of pore-fluid equalization time in SP-SC soil.
Figure 7. TC trial tests at s 200 kPa on SP-SC specimens.
Figure 9. Volume change response from HC tests on SP-SC soil at s 50, 100, 200, and 350 kPa.
674
Figure 10 shows the experimental and predicted values of deviator stress at critical state condition. In this work, a peak deviator stress value corresponds to a cumulative shear strain q of 1520%, where q (2/3)(1 3). The best-fit values of Barcelona model parameters used for numerical predictions are summarized as follows (Prez-Ruiz 2009): (0) 0.146, slope of normal compression line in (v: p) plane for saturated case (s 0); k 0.031, elastic swell index; 8.95 (MPa)1, parameter controlling the rate of increase of (s) with matric suction; r 0.65, parameter defining maximum stiffness; pc 0.032 MPa, reference stress for which the LC curve becomes a straight line; G 5.5 MPa, shear modulus; M 1.103, slope of critical state line;
450 400 350
Deviatoric stress, q (kPa)
300 250 200 150 100 50 0 0 4 8 12 16 20 24
Total shear strain (%)
Figure 12. Experimental (dotted line) and predicted (solid line) soil response from CTC tests: p 50 kPa, s 200 kPa.
Figure 10. Experimental and predicted values of deviatoric stress at critical state.
450 400 350
k 0.616, parameter controlling increase in cohesion with suction; and, po(0) 0.065 MPa, yield stress for saturated case (s 0). A detailed description of the model yield loci, flow rules, hardening laws and elasto-plastic strain definitions is presented by Alonso et al. (1990). Figure 11 shows the experimental and predicted SP-SC soil responses from CTC tests at p 50 kPa and s 100 kPa. Likewise, Figure 12 shows the experimental and predicted responses at p 50 kPa and s 200 kPa. Predictions can be considered reasonably accurate. It can also be observed the significant influence that the matric suction state exerts on soil shear resistance, with a significant increase in soil shear strength for s 200 kPa. 8 CONCLUDING REMARKS
Deviatoric stress, q (kPa)
300 250 200 150 100 50 0 0 4 8 12 16 20 24
Total shear strain (%)
Figure 11. Experimental (dotted line) and predicted (solid line) soil response from CTC tests: p 50 kPa, s 100 kPa.
Preliminary testing on SP-SC soil has shown that the newly developed true triaxial cell is suitable for testing unsaturated soils under suction-controlled conditions via axis-translation. Numerical predictions of soil response using Basic Barcelona Model proved to be reasonably accurate. The new cell is to play a fundamental role in the complete characterization of unsaturated soil behavior under multiaxial stress paths that are likely to be experienced in the field. On-going testing involves stress paths not achievable in a cylindrical cell, including simple shear. The core of this system was developed under U.S. National Science Foundation Award No. 0216545. This support is gratefully acknowledged.
675
REFERENCES
Alonso, E.E., Gens, A., and Josa, A. (1990). A constitutive model for partially saturated soils. Gotechnique, 40(3), 405430. Bishop, A.W., and Henkel, D.J. (1962). The measurement of soil properties in the triaxial test. 2nd ed., London, England: Edward Arnold, 227 pp. Fredlund, D.G., and Rahardjo, H. (1993). Soil mechanics for unsaturated soils. John Wiley and Sons, Inc., NY. Hoyos, L.R., Laikram, A., and Puppala, A.J. (2005). A novel true triaxial apparatus for testing unsaturated soils under suction-controlled multi-axial stress states. CD-Rom Proc., 16th International Conf. on Soil Mechanics and Geotechnical Engineering, September 1216, 2005, Osaka, Japan, 387390. Hoyos, L.R., and Macari, E.J. (2001). Development of a stress/suction-controlled true triaxial testing device
for un-saturated soils. Geotechnical Testing Journal, ASTM, 24(1), 513. Matsuoka, H., Sun, D.A., Kogane, A., Fukuzawa, N., and Ichihara, W. (2002). Stress-strain behaviour of unsaturated soil in true triaxial tests. Canadian Geotechnical Journal, 39, 608619. Prez-Ruiz, D.D. (2009). A refined true triaxial apparatus for testing unsaturated soils under suction-controlled stress paths. Ph.D. dissertation, University of Texas at Arlington, 320 pp. Sture, S. (1979). Development of multiaxial cubical test device with pore-water pressure monitoring facilities. Rep. VPI-E-79.18, Dept. Civil Eng., Virginia Poly. Inst. & State U., Blacksburg, VA. Wheeler, S.J., and Sivakumar, V. (1992). Development and application of a critical state model for unsaturated soils. Predictive Soil Mech., eds: G.T. Houlsby & A.N. Schofield, 709728, London.
676
Modeling unsaturated soil response under large deformations using a novel suction-controlled ring shear device
L.R. Hoyos, C.L. Velosa & A.J. Puppala
ABSTRACT: In recent years, the key hypotheses of most elasto-plastic constitutive frameworks postulated for unsaturated soils have been validated via suction-controlled oedometer, triaxial, and direct shear testing. These techniques, however, allow for the application of loads along limited modes and levels of soil deformation. Today, it is well known that most geotechnical infrastructure resting on unsaturated ground involves a wide range of deformations. Calculation of foundation settlement, for instance, requires a good estimation of soil stiffness at relatively small strains. Analysis of slopes, embankments, and soil bearing capacity, on the other hand, requires an adequate estimation of shear strength, from peak to residual. To date, however, there is very limited experimental evidence of unsaturated soil response under mid-to-large strain levels under controlled-suction states. This paper introduces a novel suction-controlled ring shear apparatus suitable for modeling unsaturated soil response over a whole range of deformations using the axis-translation technique. The apparatus features a 500 N-m torque motor along with a 360-degree-range angular deformation sensor. The lower annular shear platen accommodates a 15.24-cm OD 9.65-cm ID 1.5-cm thickness specimen and also features custom-made grooves to house a full set of six 5-bar or 15-bar ceramics as well as diffused-air flushing ports. A preliminary series of suction-controlled ring shear tests was accomplished on several identically prepared, ring-shaped specimens of compacted clayey sand (SC). Matric suction was found to exert a critical influence on soils peak and residual strength responses under constant-suction ring shear testing. 1 INTRODUCTION lack of experimental evidence of this kind has driven motivation for the present work. 2 PREVIOUS WORK
In recent decades, several authors, including Alonso et al. (1990), Wheeler and Sivakumar (1995), Fredlund and Rahardjo (1996), Cui and Delage (1996), Hoyos (1998), among few others, have been able to validate the key hypotheses of several elastoplastic frameworks postulated for unsaturated soils via suction-controlled oedometer, triaxial, and direct shear testing. These techniques, however, are only feasible for the application of loads and/or stress paths along limited modes and levels of soil deformation. Today, it is very well known that most geotechnical infrastructure resting on unsaturated ground involves a wide range of deformations. Calculation of foundation settlements, for instance, requires a good estimation of unsaturated soil stiffness at relatively small strains. Analyses of slopes, embankments, and soil bearing capacity, on the other hand, require an adequate estimation of shear strength, from peak to residual. To date, however, there are virtually no research studies available that spans the whole range of shear deformations in unsaturated soils via ring shear testing under controlled-suction states. The
Recently, Vaunat et al. (2006) reported results from a modified Bromhead type ring shear apparatus used for soil testing under controlled high-suction states via vapor transfer technique. Shortly after, Infante et al. (2007) reported an attempt at modifying a Bromhead type ring shear apparatus for testing soils under controlled low-suction states via the axis-translation technique. In both cases, low- and medium-plasticity clayey soils were tested. The authors reported an increase in the residual friction angle with an increase in the applied suction. This was mainly associated to the partial aggregation of clayey particles, which was believed to have caused the soil to behave more like a granular material. The authors also observed that the unsaturated soil residual strength was independent of both the applied vertical stress and the suction stress history. These results, however, were far from conclusive, and the ring shear devices offered plenty of room for further refinement and development.
677
A NOVEL SUCTION-CONTROLLED RING SHEAR DEVICE
This paper introduces a novel suction-controlled ring shear (RS) device that is suitable for testing unsaturated soils over a whole range of deformations using the axis-translation technique. The general design of the RS device is based upon the original design concept introduced by Bromhead (1979), in which the soil specimen seats inside the annular space formed between two concentric rings of a lower rotational cell. The RS device features two independent servocontrolled actuators: a pneumatic actuator for normal loads and an electrical actuator for shear loads. It allows the application of vertical stresses up to750 kPa, monotonic torque up to 500 N-m, constant suction states up to 500 kPa, and more than 360-degree-range of continuum angular deformation. The entire RS test layout is composed of three main modules: (1) Main cell with rotational shear system, including normal load and torque actuators; (2) Data acquisition and process control (DA/PC) system with performance and data reduction software for real-time calculation of shear stresses and average linear displacements; and (3) PCP-15U suction control panel. Details are summarized in the following. 3.1 Main cell
Figure 1. Panoramic view of complete suction-controlled ring shear test layout.
The main cell of the RS apparatus features all of the following: (a) A pneumatic servo-controlled actuator for application of vertical loads up to 8000 N; (b) An electrical servo-motor actuator for application of torsional loads up to 500 N-m; (c) A linear variable differential transducer (LVDT) to measure soil vertical deformations; (d) A full set of self-contained electrical sensors for real-time calculations of shear torque, shear angular deformation, and vertical load; (e) Adjustable top and bottom stainless-steel loading platens; and (f) Core confining cell with an 1000-kPa air pressure capacity. Figure 1 shows a panoramic view of the entire suction-controlled ring shear test layout. The upper annular platen of the main cell houses a sintered, rough-surfaced bronze stone for application of pore-air pressure ua. The lower annular platen houses a total of six HAE 5-bar ceramics for control of pore-water pressure uw. All ceramics are sealed with sintered stainless-steel rings to prevent pore-air leakage into the water system, as shown in Figure 2. Soil samples conform to the annular space between the two concentric rings of the lower platen, having a 9.65-cm inner diameter, a 15.24-cm outer diameter, and adjustable heights between 0.5 and 3.0 cm. Vertical pressure, prior to shearing, is applied via the upper annular platen. During shearing,
Figure 2. Upper annular platen (shown upside down), lower annular platen, and fully-assembled bottom torsional gear.
the upper annular platen is monotonically rotated by an electrical drive unit while the lower annular platen remains restrained. The core system of the RS apparatus was manufactured by Geotechnical Consulting and Testing Systems (GCTS), Tempe, Arizona. 3.2 DA/PC system
The DA/PC system consists of all of the following: (1) One SCON-1500 digital servo controller and acquisition system; (2) One DSB-11 universal signal conditioning board (for load cells and LVDTs); (3) GCTS software for real-time measurement and control of normal stress, shear stresses, and angular deformations; and (4) IBM-PC unit.
678
3.3
PCP-15U suction control panel
A PCP-15U model pressure control system (Fig. 1) is used for direct control of pore-air pressure ua at the top of the soil specimen with dual pressure regulators and gauges for precise measurement and control of matric suction, s ua (uw 0). The panel also features a flushing mechanism for removal of diffused air beneath the HAE ceramics during constant-suction RS testing. 4 SAMPLE PREPARATION VIA IN-PLACE STATIC COMPACTION
The test soil used in this work classifies as clayey sand (SC) according to the USCS (80% sand, 20% clay). Soil samples were compacted directly into the lower annular platen using in-place static compaction, as shown in Figure 4. The upper platen is used to compress the loose soil-water mix until achieving a target density of 95% of maximum Proctor dry density, d-max 1.84 g/cm3. During compaction, a monotonic force is applied to the top platen by means of a triaxial loading frame. Samples were prepared in a single lift under a constant compaction displacement rate of 1.25 mm/min. The final height and weight of the compacted samples were determined in order to verify the desired dry unit weight. A total of five samples were prepared with water contents equivalent to initial suction states of 25 kPa (two samples), 50 kPa (two samples), and 100 kPa (one sample) according to the soil-water characteristic curve (SWCC) shown in Figure 3, which was obtained via pressure plate testing. In this figure, solid and dashed lines represent best-fitting SWCC model curves based upon Van Genuchten (1980) and Fredlund and Xing (1994) equations, respectively.
25 SWCC Data
Figure 4. Static compaction of ring-shaped sample using conventional triaxial frame.
Two more samples were prepared with a moisture content that is 4% greater than Proctor optimum and subsequently soaked in distilled water in the main RS cell for testing under zero suction state (s 0). 5 SUCTION-CONTROLLED RING SHEAR TESTING PROCEDURE
5.1 Single-stage ring shear testing Typical single-stage (SS) ring shear test paths at different controlled-suction states are shown schematically in Figure 5. The tests were performed on four identically prepared samples, SS1, SS2, SS3, and SS4, under constant suction states, s 0, 25, 50, and 100 kPa, respectively. All samples were tested under a constant net normal stress, (n ua) 25 kPa. A typical single-stage procedure can be summarized as follows. After full assembly of the RS apparatus (Fig. 1), a vertical load was monotonically applied via the upper annular platen to induce a normal stress 25 kPa greater than the corresponding value of suction. The soil was then allowed to consolidate under this load for at least 24 hours. Pore-air pressure ua was then increased via compressed air in the main cell (Fig. 1) until achieving the desired suction state. The external normal stress was adjusted accordingly to keep a constant net normal stress, (n ua) 25 kPa. After complete equalization of pore fluids under a constant pore-air pressure, which may take up to 48 hours, the shearing stage was initiated. Shearing was continued at a constant angular displacement
Gravimetric water content, w (%)
20
Van Genuchten (1980) Fredlund and Xing (1994)
15
10
0 0.1 1.0 10.0 100.0 1000.0 10000.0
Figure 3. (SC) soil.
SWCC best-fitting models for clayey sand
679
200 200
Shearing stage IV
Monotonic shearing stage

Net normal stress, ( ua): kPa
100
SS4
100
Shearing stage III
50
SS3
50
Shearing stage II
25
SS2
25
Shearing stage I
SS1 Horizontal displacement (mm)

0 0
Figure 5. Typical suction-controlled single-stage RS test paths for constant net normal stress, (n ua) 25 kPa.
Figure 6. Typical suction-controlled multi-stage RS test paths for constant suction states, s 0, 25, or 50 kPa.
rate of 0.024o/min until it was apparent that a welldefined residual state had been reached. At this point, shearing was stopped and the normal load completely removed. Finally, the air pressure in the main cell was slowly brought back to zero. The top annular platen was then removed and the sample inspected for possible shear band formations. This procedure was followed for samples SS2, SS3, and SS4, tested at constant suction, s 25, 50, and 100 kPa, respectively. The same procedure summarized above was followed for sample SS1, except that this time the sample was kept soaked in distilled water in the main RS cell, under a constant 25-kPa normal stress, prior to initiating the shearing stage. Saturation of the sample was allowed for at least 24 hours in order to ensure a zero suction (s 0) state. 5.2 Multi-stage ring shear testing
A typical multi-stage (MS) ring shear test path at a constant suction state is shown schematically in Figure 6. The tests were performed on three identically prepared samples, MS1, MS2, and MS3, under constant suction states, s 0, 25, and 50 kPa, respectively. Sample MS1 (s 0 kPa) was subjected to a multi-stage path involving a range of net normal stresses, (n ua) 25, 50, 100, and 200 kPa. Sample MS2 (s 25 kPa) was subjected to a multi-stage path involving a range of net normal stresses, (n ua) 25, and 75 kPa. Finally, Sample MS3 (s 50 kPa) was subjected to a multistage path involving a range of net normal stresses, (n ua) 25, 50, and 100 kPa. A typical multi-stage test procedure can be summarized as follows. After initial consolidation and equalization of the pore fluids to achieve a net normal stress, (n ua) 25 kPa, the first shearing stage is initiated at a constant angular displacement rate of 0.024o/min until it is apparent that a residual state had been reached. At this point, shearing is stopped
and the net normal stress further increased to the next corresponding level (Fig. 6), depending on the sample being tested. The sample is then allowed to consolidate under this new load for a period of at least 8 hours, while the initial suction state is kept constant. Shearing is then reinitiated until it is again apparent that a new residual state has been reached. The same procedure is followed for the remaining value(s) of net normal stress (n ua). Upon completion of the last shearing stage, both the torque and vertical load are slowly brought back to zero and the air pressure in the main RS cell fully removed. The upper platen is then removed and the top surface of the sheared sample carefully examined for significant features such as shear band formations. This procedure was followed for samples MS2 and MS3. The same procedure was followed for sample MS1, but the sample was first soaked in distilled water in the main RS cell, under a constant 25-kPa normal stress, prior to initiating the first shearing stage. 6 ANALYSIS OF TEST RESULTS
Results from single-stage RS tests on SC soil under net normal stress, (n ua) 25 kPa, and matric suction states, s 0, 25, 50, and 100 kPa, are shown in Figure 7. Test results are presented in terms of shear stress (kPa) versus horizontal displacement (mm), as well as vertical displacement (mm) versus horizontal displacement (mm). As it is expected, both the peak and residual shear strengths are significantly influenced by the level of matric suction, with a considerable increase in both values for a suction, s 200 kPa. SC soil also seems to exhibit a more brittle-like behavior with increasing suction: at lower suction, the residual strength is just below the peak strength; whereas at higher suction, there is a more
680
140 120
140 120 N o n -b ri ttl e B ri ttl e -l i k e
Shear stress (kPa)
100 80 60 40 20 0 0 -0.40 20
Shear strength (kPa)
s = 100 kPa
100 80 60 40 20
Peak Residual
s = 50 kPa s = 25 kPa s = 0 kPa
40
60
80
100
0 0 25
50
75
100
125
-0.30 s = 100 kPa -0.20 -0.10 0.00 0.10 0.20 0 20 s = 25 kPa s = 0 kPa s = 50 kPa
Figure 8. Variation of peak shear strength and residual shear strength of SC soil with suction, s ua (uw 0).
140 120
n - u a) = 200 kPa
Shear stress (kPa)
100 80 60 40 20 0 0 20 40 60 80 100
40
60
80
100
n - u a) = 100 kPa
Figure 7. Results from single-stage RS tests on SC soil under (n ua) 25 kPa, and s 0, 25, 50, and 100 kPa.
n - u a) = 50 kPa n - u a) = 25 kPa
appreciable difference between peak and residual shear strength. This is further substantiated by the more pronounced dilation observed under higher suction states. Figure 8 shows the variation of peak and residual shear strength of SC soil with matric suction, s. This figure further highlights the more appreciable difference between the peak and residual shear strength as suction increases. In light of these preliminary results, it can be partly concluded that the SC soil experiences a transition from a ductile-like behavior to a more brittle-like behavior when subjected to a suction state between 25 kPa and 50 kPa. Results from multi-stage RS tests on the saturated SC soil (s 0) and unsaturated SC soil (s 50 kPa) are shown in Figures 9 and 10, respectively. As it is expected, the residual strength increases with the net normal stress, regardless of suction state. For higher suction, s 50 kPa, the SC soil seems to exhibit a more contractile response as the net normal stress is increased (Fig. 10), which can be attributed to the nature of the multi-stage testing scheme followed in this work, where a preexisting failure surface is induced on the sample before increasing the net normal stress to the next corresponding level. Figure 11 shows the effect of suction on the position and slope of residual failure envelopes for compacted SC soil. Experimental points represent residual strength values obtained from singlestage and multi-stage RS tests as described in the previous two sections.
-0.40
-0.30 -0.20 -0.10 0.00 0.10
n - u a) = 100 kPa n - u a) = 200 kPa
n - u a) = 50 kPa n - u a) = 25 kPa
0.20 0 20
40
60
80
100
Figure 9. Results from multi-stage RS tests on saturated (s 0) SC soil under (n ua) 25, 50, 100, and 200 kPa.
It is observed that matric suction exerts a critical influence on the position of the residual failure envelope, with a considerably higher position for s 50 kPa. Likewise, suction shows a significant influence on the residual apparent cohesion, with apparent cohesion almost negligible in the saturated soil (s 0). Even though there appears to be a slight increase in the residual friction angle r at higher suction states, the slope of residual failure envelopes remains virtually constant, regardless of suction state. Figure 12 shows the effect of net normal stress on the position and slope of residual failure envelopes projected onto the residual shear stress versus matric suction plane. Similarly to the trends observed in residual friction angle 'r from Figure 11, the
681
140 120
Shear stress (kPa)
n - u a) = 100 kPa
100 80 60 40 20 0 0 -0.40 20 40 60 80 100
soils, similar to that postulated for peak strength (Fredlund and Rahardjo 1996) could eventually be devised. 7 CONCLUDING REMARKS
n - u a) = 50 kPa n - u a) = 25 kPa
-0.30 -0.20 -0.10 0.00 0.10 0.20 0 20
n - u a) = 25 kPa
n - u a) = 50 kPa n - u a) = 100 kPa

40 60 80 100
Preliminary suction-controlled RS testing on compacted SC soil has shown that the newly developed RS apparatus is suitable for testing unsaturated soils under constant suction states via the axistranslation technique. Preliminary observations also suggest that a conceptual residual strength framework for unsaturated soils, similar to that postulated for the peak strength (Fredlund and Rahardjo 1996), may well be devised when additional experimental evidence of this kind becomes available. The core of the system was developed under U.S. National Science Foundation Award No. 0626090. This support is gratefully acknowledged. REFERENCES
Alonso, E.E., Gens, A., and Josa, A. (1990). A constitutive model for partially saturated soils. Gotechnique, 40(3): 405430. Bromhead, E.N. (1979). A simple ring shear apparatus. Ground Engineering, 12 (5): 4044. Cui, Y.J., and Delage, P. (1996). Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique, 46(2): 291311. Fredlund, D.G., and Rahardjo, H. (1993). Soil mechanics for unsaturated soils. John Wiley, NY. Fredlund, D.G., and Xing, A. (1994). Equations for the soil water characteristic curve. Canadian Geotechnical Journal, 31(3): 521532. Hoyos, L.R. (1998). Experimental and computational modeling of unsaturated soil behavior under true triaxial stress states. Ph.D. dissertation, Georgia Institute of Technology, Atlanta, GA, 352 pp. Infante, J.A., Vanapalli, S.K., and Garga, V.K. (2007). Modified ring shear apparatus for unsaturated soil testing. Geotechnical Testing Journal, ASTM, 30(1): 19. Van Genuchten, M.T. (1980). A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal, 44: 892898. Vaunat, J., Amador, C., Romero, E., and Djeren-Maigre, I. (2006). Residual strength of a low plasticity clay at high suctions. Proceedings of the Fourth International Conference of Unsaturated Soils, vol. 1, 12791289. Wheeler, S.J., and Sivakumar, V. (1995). An elasto-plastic critical state framework for unsaturated soil. Gotechnique, 45(1): 3553.
Figure 10. Results from multi-stage RS tests on SC soil under s 50 kPa, and (n ua) 25, 50, and 100 kPa.
150
s = 50 kPa
Residual shear stress (kPa)
s = 25 kPa s = 0 kPa
' r = 33.3 ' r = 30.6 ' r = 30.3
100
50
0 0 25 50 75 100 125 150 175 200
Net normal stress (kPa)
Figure 11. Effect of suction on position and slope of residual failure envelopes for compacted SC soil.
150
Residual shear stress (kPa)
r = 39.35
100
r = 39.35 r = 40.8
b
50
Net normal stress = 100 kPa Net normal stress = 50 kPa Net normal stress = 25 kPa
0 0 25 50 75 100 125
Figure 12. Effect of net normal stress on position and slope of residual failure envelopes for compacted SC soil.
angle br remains virtually constant, regardless of net normal stress level. All of these preliminary observations suggest that a residual strength framework for unsaturated
682
Small-strain stiffness of compacted silty sand using a proximitor-based suction-controlled resonant column device
L.R. Hoyos, E.A. Suescn, J.A. Pineda & A.J. Puppala
ABSTRACT: Small-strain soil stiffness properties such as shear wave velocity, shear modulus, and material damping, are key subsoil parameters for an adequate analysis and design of unsaturated earth structures subject to static and non-static loading. Most conventional geotechnical testing techniques, however, are not able to capture this small-strain behavior and, hence, vastly underestimate the true soil stiffness, mainly due to inaccuracies in small-strain measurements. In the United States, a great deal of research efforts has been devoted to field and laboratory based measurements of soil suction, assessments of soil-water retention properties, and analyses of swell-collapse behavior. However, very few efforts have been focused on small-strain response of unsaturated soils and their dynamic characterization at small strains. This paper introduces a suction-controlled, proximitor-based resonant column device featuring a PCP-15U pressure control panel that allows for the implementation of the axis-translation technique via the independent control of pore-air and pore-water pressures in the specimen. A preliminary series of resonant column tests were conducted on statically compacted samples of silty sand for a range of constant suction states between 50 kPa and 400 kPa, at different net confining pressures. Results show the critical role of matric suction on the soils small-strain response. The apparatus also features a full set of self-contained bender elements for simultaneous testing under both techniques. However, preliminary results from an on-going bender element testing program are beyond the scope of the present work. 1 INTRODUCTION infrastructure is now widely acknowledged (Atkinson 2000). As most soil materials involved in these geosystems remain partially saturated, and the real strains experienced are relatively small, there is a great need for a better understanding of the smallstrain behavior of soils under such conditions. Moreover, in the unsaturated soil practice, a thorough understanding of the critical influence that the season-dependent suction states have on small-strain soil stiffness properties, such as shear wave velocity (Vs), small-strain shear modulus (Gmax), and small-strain material damping (Dmin), is of paramount importance. These are key sub-soil parameters for the adequate analysis and/or design of unsaturated earth structures subject to static as well as non-static loading, as depicted in Figure 1. Since the static/dynamic response of unsaturated soils is known to heavily depend on suction states, the lack of consideration of these suction effects in the small-strain dynamic characterization of unsaturated soils may lead to erroneous property measurements and, ultimately, as stated earlier, faulty and/or excessively conservative designs of earth structures. Most conventional soil testing devices and techniques, however, are not able to adequately capture this small-strain behavior and, hence, vastly
In every state of the nation (U.S.A.), civil engineers face problems with road and railway embankments, riverbanks, earth dams, and shallow foundation materials that remain under partially saturated conditions throughout any given year. The lack of education and training among engineering graduates and practitioners to properly deal with unsaturated soil conditions has resulted in faulty or excessively conservative designs, construction delays, and deficient long-term performance of built infrastructures. Recently, the unsaturated soil mechanics discipline has begun to receive increasing attention from researchers and practitioners nationwide, thus providing better explanations for soil behavioral patterns than the conventional saturated soil mechanics. A great deal of research efforts has been devoted to field and laboratory based measurements of soil suction, assessments of soil-water retention properties, and analyses of swell-collapse behavior. However, very few efforts have been focused on small-strain response of unsaturated soils and the dynamic characterization of these materials at small strains. The critical role played by soil stiffness at small strains in the analysis and/or design of geotechnical
683
Traffic load Cyclic load
Pavement
Machine foundation
Cross-hole testing
Lumped mass
Unsaturated soil
Unsaturated Unsaturated soil soil Unsaturated Unsaturated soil soil
Idealization
G(s)
D(s)
Figure 1. Idealization of common geotechnical infrastructure above ground-water table subject to non-static loading.
underestimate the true soil stiffness, mainly due to inaccuracies in the small-strain measurements. This paper introduces a suction-controlled, proximitor-based resonant column (RC) apparatus featuring a PCP-15U pressure control panel that allows for the implementation of the axis-translation technique via independent application and control of pore-air and pore-water pressures in the soil. A preliminary series of RC tests were conducted on statically compacted samples of silty sand for a range of constant suction states between 50 kPa and 400 kPa at different net confining pressures. The RC apparatus also features a full set of self-contained bender elements for simultaneous testing under both techniques. However, preliminary results from an on-going, bender element based testing program are beyond the scope of the present work. The recent focus of the majority of U.S. Departments of Transportation has been towards proposing pavement design procedures based on a mechanistic-empirical approach using the resilient modulus (MR) as the primary soil parameter. However, a more rational procedure should be based on a thorough understanding of the effects of seasondependent suction states (i.e., seasonal variations that include wet-dry and/or freeze-thaw cycles) on small-strain soil stiffness properties. The present work is an attempt to contribute towards this goal. 2 PREVIOUS WORK
The following is not intended to be a fully inclusive literature review of all research efforts accomplished to date on small-strain stiffness characterization and response of partially saturated soils. The main intent is to highlight the importance of the subject and the extensive room still available for further research. Brull (1980) and Wu et al. (1985) reported linear relationships between initial shear modulus Go and soil suction for silty and sandy soils in the 080 kPa suction range. In both cases, the degree of
saturation Sr of the soil was assessed immediately after measuring soil stiffness via resonant column testing. The testing procedure consisted in applying a confining pressure on the unsaturated specimen, under drained conditions, and then measuring Go after 1,000 minutes. The Go vs. Sr functions show a distinct peak for initial shear modulus Go corresponding to Sr values between 10%20%, for any given confining stress. Qian et al. (1991) studied the effects of capillarity on small-strain shear modulus of partially saturated sands. A Hall-type resonant column apparatus was used to perform the experiments. They reported that the capillary stress significantly increase the shear modulus of unsaturated sands. Void ratio, confining pressure, degree of saturation, grain shape and size distribution were identified as the primary factors affecting the soils shear modulus. Marinho et al. (1995) performed bender element based measurements on London clay specimens assessing soil suction via filter paper technique. Picornell and Nazarian (1998) also reported some results obtained from reconstituted samples of silt and clay using bender elements inside a suction plate. In both cases, a power law was shown to best fit the experimental Go vs. suction functions, and that the small-strain moduli tent to a constant value when the soil moistures moves towards residual water content. Cabarkapa et al. (1998) used the bender element technique adapted to a cylindrical triaxial cell to test silty soil samples under controlled suction states via axis-translation technique. They concluded that, for normally-consolidated quartz silt, the unsaturated Go values can be readily assessed by multiplying the saturated Gsat values corresponding to the same net pressure, (p ua), by a factor that depends only on matric suction, s (ua uw). Cho and Santamarina (2001) performed a series of bender element based experiments on unsaturated particulate materials with emphasis on pendular menisci stage. Small-strain stiffness was continuously measured on soil specimens subject to drying while changes in stiffness were related to changes in interparticle forces. Soils included glass beads (spherical particle), a mixture of glass beads and kaolinite (flat particlesclay), granite powder (angular particlessilt), and sandboil sand. They concluded that while capillary forces affect smallstrain stiffness (shear-wave velocity), they may not contribute to the large-strain stiffness or strength. As shown by this brief review, significant efforts have been reported since the early 1980s to study the influence of capillarity and degree of saturation on the small-strain stiffness of unsaturated soils using either resonant column or bender element based testing technique. Even though all of these works have made important contributions in this area, virtually none has
684
directly dealt with resonant column testing under suction-controlled conditions. Only until very recently, Mancuso et al. (2002) and Vasallo et al. (2007) performed a series of suction-controlled resonant column/torsional shear tests on unsaturated silty soil using an RC/TS apparatus originally developed at the University of Napoli, Italy. Matric suction, s (ua uw), was induced via the axis-translation technique, while the torsional torque was steadily increased to study soil stiffness response at small, mid, and high-shear strain amplitude levels. Results within the small-strain range were similar to those reported by Cabarkapa et al. (1998) using bender element based technique, and no attempt was made to study the effects of suction state on material damping D of the tested unsaturated silty sand. 3 A SUCTION-CONTROLLED PROXIMITOR-BASED RESONANT COLUMN DEVICE
The cell accepts samples having a diameter of 70 mm and 35 mm with a length of 2 to 2.5 times the diameter. The bottom pedestal features a full set of HAE ceramics, and one bender element crystal for shear-wave velocity readings, as shown in Figure 3. An electrical servo motor actuator is used for the application of torsional loads with a / 2.33 kN-m (peak) capacity and 300-Hz frequency. Mounted on an internal floating frame to allow for large vertical specimen deformations, the actuator includes a servo amplifier for closed-loop control of torsional load or angular deformations, as shown in Figures 4 and 5. 3.2 Data acquisition and process control system The SCON-1500 model DA/PC system consists of a microprocessor based digital servo controller, function generator, data acquisition, and digital I/O unit. It includes software for performance and data reduction of resonant column tests with automatic calculation of shear modulus, shear strain, and damping ratio at strain levels ranging from 10% to 104 %. The signal conditioning mother
The resonant column based technique was first used to study dynamic properties of rock materials in the early 1930s, and it has been continuously evolving since then for the dynamic characterization of a wide variety of geologic materials. The RC test essentially consists of a soil column in fixed-free end conditions that is excited to vibrate in one of its natural modes. Once the frequency at resonance (fr) is obtained, the shear wave velocity (Vs) and, hence, shear modulus (Gmax) of the soil can readily be determined. Material damping ratio (Dmin) can be obtained from logarithmic decay curves or the half-power method using the frequency response curve (Huoo-Ni 1987). The suction-controlled proximitor-based resonant column device introduced in this paper is composed of four main modules: (1) THS-100 model main RC cell, (2) GCTS data acquisition/process control system, (3) PCP-15U model pressure control panel for matric suction application via axis-translation, and (4) HM-414 model confining pressure panel. Further details are summarized in the following. 3.1 Main RC cell
Figure 2. Panoramic view of entire suction-controlled resonant column test layout.
A panoramic view of the suction-controlled resonant column testing layout is shown in Figure 2. The core system of the apparatus was manufactured by Geotechnical Consulting and Testing Systems (GCTS), Tempe, Arizona. The THS-100 model main cell consists of a four-column stainless steel construction with an external reinforced acrylic cell of 1000-kPa confining pressure capacity. It includes feed-trough connectors for internal angular displacement, velocity, torque, and axial deformation transducers.
Figure 3. Bottom RC pedestal with HAE ceramics and bender element (receiver).
685
constant compaction displacement rate of 1.0 mm/ min (Venkatarama and Jagadih 1995). The soil was compacted to a target void ratio, e 1.0; dry density, d 1.313 kN/m3; and average water content of 26%, in order to achieve similar conditions under natural undisturbed state. 5 TEST VARIABLES AND PROCEDURE
Figure 4. Partly assembled and fully assembled RC main cell.
Figure 5. Close-up view of proximitor mounting (internal angular displacement transducer).
board accepts up to eight universal signal conditioning module slots (Fig. 2). 3.3 PCP-15U and HM-414 pressure panels
A PCP-15U model pressure control system (Fig. 2) is used for direct control of pore-air pressure ua at the top of the soil specimen with dual pressure regulators and gauges for precise measurement and control of matric suction, s ua (uw 0). The panel also features a flushing mechanism for removal of diffused air beneath the HAE ceramics during constant-suction RC testing. Finally, an HM-414 model pressure panel is used for application of external confining pressure (Fig. 2). 4 TEST SOIL AND SAMPLE PREPARATION
The test soil used in this work classifies as silty sand (SM) according to the USCS (70% sand, 30% silt). The passing No. 200 sieve fraction has a liquid limit, LL 26.4%, and plastic limit, PL 22.2%. The soil has a specific gravity, Gs 2.71, with coarse fraction having particle sizes between 0.51.2 mm. Samples were compacted into a 70-mm diameter, 130-mm height, compaction split mold using static compaction. During compaction, a monotonic force is applied using a triaxial loading frame. Samples are prepared in three lifts under a
In the present work, a series of constant-suction RC tests has been accomplished on four identically prepared samples of SM soil using the static compaction process summarized above. Each sample was tested for a different value of net confining pressure, (p ua) 50, 100, 200, or 400 kPa. RC tests were then performed on the same sample under a constant net confining pressure (p ua) and for four different suction states induced by axis-translation, that is, s (ua uw) 50, 100, 200, and 400 kPa. A typical RC test procedure can be summarized as follows. After full assembly of the RC apparatus (Fig. 2), a confining pressure is applied to the sample via the HM-414 panel in order to induce a confinement that is 50 kPa greater than the first desired value of suction. The sample is then allowed to consolidate under this confinement for at least 2 hours. Pore-air pressure ua is then increased via the PCP-15U panel until achieving the desired suction state. The external confinement was adjusted accordingly to keep a constant net confining pressure, (p ua) 50 kPa. After complete equalization of pore fluids under a constant pore-air pressure, which may take up to 24 hours, a RC test is conducted by inducing a torsional vibration at the top of the soil specimen via the servo motor actuator (0.5 kN-m torque). The frequency of the vibration is changed by sweeping the entire input frequency scale until obtaining a frequency response curve, normally between 60 Hz and 240 Hz. Frequency response curves are presented in terms of shear strain fraction (cm/cm) versus frequency (Hz). The steady torsional vibration is then completely cut off in order to obtain the logarithmic decay curve for the assessment of material damping. Logarithmic decay curves are presented in terms of shear strain fraction (cm/cm) versus decay time (sec). Pore-air pressure ua is then further increased until achieving the next desired matric suction state. After complete equalization of the pore fluids, a new RC test, as described above, is conducted. The same test procedure is repeated for suction states, s (ua uw) 50, 100, 200, and 400 kPa, on the same sample at constant net confining pressure, (p ua). This identical procedure was followed in testing SM soil samples at constant net confining pressures, (p ua) 100, 200, and 400 kPa.
686
6 6.1
ANALYSIS OF TEST RESULTS Frequency response curves
A family of frequency response curves from suctioncontrolled RC tests conducted on one SM soil sample at net confining pressure, (p ua) 50 kPa, and matric suction states, s 50, 100, 200, and 400 kPa, is shown in Figure 6. The magnitudes of induced shear strain amplitude range from 0.01% to 0.001%. It can be observed the critical effect that the matric suction state has on the resonant frequency of the soil column, with a considerable increase in resonant frequency for suction, s 400 kPa. This can be directly attributed to an increase in effective stress and, hence, an increase in the rigidity of the SM soil skeleton at higher suction states. It can also be noted that the maximum shear strain induced by a 0.5 kN-m torque (peak response in the frequency response curve) decreases at higher suction states. A family of frequency response curves from suction-controlled RC tests conducted on four different SM soil samples at net confining pressures, (p ua) 50, 100, 200, and 400 kPa, respectively, and matric suction state, s 100 kPa, is shown in Figure 7. As expected, the level of net confinement has an even more pronounced effect on frequency response. 6.2 Logarithmic decay curves
vibration. Since the initial peak amplitude is higher at suction, s 200 kPa, this is indicative of lower material damping at higher suction states, which further confirms the increase in rigidity of the soil skeleton with an increase in matric suction state. Likewise, Figure 9 shows the logarithmic decay curves from two different SM soil columns at net confining pressures, (p ua) 50 kPa and 400 kPa, and matric suction state, s 50 kPa. Again, it is observed that the level of net confinement has an even more
2.0E-04 (p-ua) = 50 kPa, s = 100 kPa (p-ua) = 100 kPa, s = 100 kPa (p-ua) = 200 kPa, s = 100 kPa 1.5E-04 (p-ua) = 400 kPa, s = 100 kPa
Shear strain fraction (cm/cm)
1.0E-04
5.0E-05
0.0E+00 60 100 140 180 220
Frequency (Hz)
Figure 7. Frequency response curves from suctioncontrolled RC tests on SM soil at (p ua) 50, 100, 200, and 400 kPa, and s 100 kPa.
2.0E-04
Figure 8 shows logarithmic decay curves from one SM soil column at net confining pressure, (p ua) 50 kPa, and matric suction states, s 50 kPa and 200 kPa. It can be noticed that the initial peak shear-strain amplitude at time, t 0, that is, as soon as the vibration is cut off, is lower at higher suction states, in accordance to test results shown in Figure 6. It can also be noted that both decay curves appear to be fully attenuated after 0.08 seconds of free
2.0E-04 (p-ua) = 50 kPa, s = 50 kPa (p-ua) = 50 kPa, s = 100 kPa (p-ua) = 50 kPa, s = 200 kPa (p-ua) = 50 kPa, s = 400 kPa
1.5E-04 1.0E-04 5.0E-05 0.0E+00
(p-ua) = 50 kPa, s = 50 kPa (p-ua) = 50 kPa, s = 200 kPa
-5.0E-05 -1.0E-04 -1.5E-04 -2.0E-04 0 0.04 0.08 0.12
Time (s)
Figure 8. Logarithmic decay curves from SM soil columns at (p ua) 50 kPa, and s 50 kPa and 200 kPa.
2.0E-04 1.5E-04 1.0E-04 5.0E-05 0.0E+00 -5.0E-05 -1.0E-04 -1.5E-04 (p-ua) = 50 kPa, s = 50 kPa (p-ua) = 400 kPa, s = 50 kPa
1.5E-04
1.0E-04
5.0E-05
0.0E+00 60
100
140 Frequency (Hz)
180
220
-2.0E-04 0 0.04 0.08 0.12
Time (s)
Figure 6. Frequency response curves from suctioncontrolled RC tests on SM soil at (p ua) 50 kPa, and s 50, 100, 200, and 400 kPa.
Figure 9. Logarithmic decay curves from SM soil columns at (p ua) 50 kPa and 400 kPa, and s 50 kPa.
687
pronounced effect on the decay response. The initial peak shear-strain amplitude at time, t 0, is considerably lower at higher net confining pressures. 6.3 Small-strain stiffness response
amplitude of the nth free-vibration cycle. The trends also confirm those shown in Figures 8 and 9. 7 CONCLUDING REMARKS
Figure 10 shows the variation of small-strain shear moduli Gmax with net confining pressure, (p ua), on the basis of matric suction state, s (ua uw) 50, 100, 200, and 400 kPa. The trends further substantiate those shown in Figures 6 and 7. It is worth noting that at higher net confinement levels, the effect of suction state appears to become less critical. Similarly, Figure 11 shows the variation of smallstrain damping ratio Dmin with suction, s (ua uw), on the basis of net confining pressure, (p ua) 50 kPa and 200 kPa. Damping ratio can be readily obtained from a logarithmic decay curve as follows: Z 1 D ln o 2 n Zn (1)
Preliminary suction-controlled RC testing on compacted SM soil has shown that the new RC device is suitable for testing unsaturated soils under constant suction states via axis-translation technique. Results highlight the critical influence that soil suction states have on small-strain soil stiffness properties. The core system of the RC device was developed under the Research Equipment Fund (REF) program of the University of Texas at Arlington. This support is gratefully acknowledged. REFERENCES
Atkinson, J.H. (2000). Nonlinear soil stiffness in routine design. Gotechnique, 50(5): 487588. Brull, A. (1980). Caracteristiques mcaniques des sols de fondation de chausses en fonction de leur tat dhumidit et de compacit. Proc., International Conf. on Soil Compaction, Paris, vol. 1: 113118. Cabarkapa, Z., Cuccovillo, T., and Gunn, M. (1998). A new triaxial apparatus for testing unsaturated soils. Proc., 2nd International Conf. on Unsaturated Soils, Beijing, China, vol. 2: 194195. Cho, G., and Santamarina, J.C. (2001). Unsaturated particulate materials: Particle-level studies. J. of Geotechnical and Geoenvironmental Engineering, ASCE, 127(1): 8496. Huoo-Ni, S. (1987). Dynamic properties of sand under true triaxial stress states from resonant column/torsional shear tests. Ph.D. Dissertation, Univ. of Texas, Austin, TX. Mancuso, C., Vassallo R., and dOnofrio A. (2002). Small strain behavior of a silty sand in controlled-suction resonant columntorsional shear tests. Canadian Geotechnical Journal, 39: 2231. Marinho, E.A., Chandler, R.J., and Crilly, M.S. (1995). Stiffness measurements on high plasticity clay using bender elements. Proc., 1st International Conf. on Unsaturated Soils, Paris, vol. 1: 535539. Picornell, M., and Nazarian, S. (1998). Effects of soil suction on low-strain shear modulus of soils. Proc., 2nd International Conf. on Unsaturated Soils, Beijing, vol. 2: 102107. Qian, X., Gray, D.H., and Woods, R.D. (1991). Resonant column tests on partially saturated sands. Geotechnical Testing Journal, ASTM, 14(3): 266275. Vassallo, R., Mancuso, C., and Vinale, F. (2006). Effects of net stress and suction history on small strain stiffness of a compacted clayey silt. Canadian Geotechnical Journal, 44(4): 447462. Venkatarama, R.B., and Jagadish, K.S. (1993). The static compaction of soils. Gotechnique, 43(2): 337341. Wu, S., Gray, D.H., and Richart, Jr., F.E. (1984). Capillary effects on dynamic modulus of sands and silts. Journal of Geotechnical Engineering, ASCE, 110(9): 11881203.
where, D material damping ratio, Zo peak amplitude of the first free-vibration cycle, and, Zn peak
120
Small-strain shear modulus, Gmax (MPa)
100
80
60 s = 400 kPa 40 s = 200 kPa s = 100 kPa s = 50 kPa 0 0 100 200 300 400 500
20
Net mean stress, (p-ua): kPa
Figure 10. Variation of small-strain shear moduli Gmax with net confining pressure, (p ua), on the basis of matric suction state, s (ua uw).
10
Small-strain damping ratio, Dmin (%)
(p-ua) = 50 kPa (p-ua) = 200 kPa
1 0 100 200 300 400 500
Matric suction, s = (ua-uw): kPa
Figure 11. Variation of small-strain damping ratio Dmin with matric suction, s (ua uw), on the basis of net confining pressure, (p ua).
688
Efficient measurement of unsaturated hydraulic conductivity using a wetting front advancing method
X. Li
Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong School of Civil Engineering, Beijing Jiaotong University, Beijing, China
L.M. Zhang & D.G. Fredlund
Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong
ABSTRACT: Unsaturated hydraulic conductivity is one of the most difficult parameters to measure because of the time involved and the limited suction measurement range. In this study, a new wetting front advancing method was developed for measuring unsaturated hydraulic conductivity. In this method, an initially air-dried soil column is subjected to a wetting event. During the wetting process, a wetting front advances with time and gives a wetting front advancing velocity function. A new interpretative procedure is developed to calculate the unsaturated hydraulic conductivity based on the monitored water content, suction, and wetting front advancing velocity. The proposed technique is used to measure the unsaturated hydraulic conductivities of five colluvial soils. The results indicate that the proposed technique is timesaving (i.e., requires several days for a complete test) and can be applicable over wide ranges of suctions and unsaturated hydraulic conductivities. The measured unsaturated hydraulic conductivity using the wetting front advancing method is similar to that obtained when using the instantaneous profile method. 1 INTRODUCTION and a wider range of soil suction. To achieve this objective, a new experimental method called the wetting front advancing method was developed. In this method, a large-scale soil column test is used to simulate the flow of water into an unsaturated soil column. The monitored data are used to calculate the soil hydraulic conductivity curve with the help of a newly developed analytical interpretation procedure. The newly developed experimental and interpretation procedure were used to measure the unsaturated hydraulic conductivities of a sandy silt.
The soil permeability function (sometimes referred to as hydraulic conductivity curve) is the primary soil parameter function required when performing seepage analysis for saturated/unsaturated soil systems (Fredlund et al. 1994). The soil hydraulic conductivity curve is required when modeling contaminant migration or water flow in vadose zones, studying the triggering of landslides due to rainfall infiltration, and designing capillary barriers and cover systems. There are three primary methods proposed in the literature for the measurement of unsaturated hydraulic conductivities in the laboratory or in situ; namely, the steady-state method (e.g. Klute 1965; Samingan et al. 2003), the instantaneous profile method (e.g., Richards and Weeks 1953; Hamilton et al. 1981), and the parameter estimation method (e.g., Kool and Parker 1987). The existing methods for measuring soil hydraulic conductivity curves are often time-consuming, limited in the available measurement ranges of soil suction (e.g. 01500 kPa in a test using the steady-state method) or dependent on the selected permeability function models for analyzing experimental data (Li et al. 2009). It is desirable to develop a method that is time-saving and applicable to a wide range of hydraulic conductivity
DEVELOPMENT OF A LARGE-SCALE COLUMN DEVICE FOR MEASURING SOIL HYDRAULIC CONDUCTIVITY Soil column device
2.1
A soil column device was developed to measure the soil hydraulic conductivity curve. The soil column device is shown in Fig. 1, which consists of a column cell, a system for suction and water content monitoring, and a water flow control and measurement system. A 120 mm diameter, 1000 mm high acrylic column cell was used to hold the soil sample and support soil suction and water content monitoring sensors. Four ML2x Theta-probe soil moisture sen-
689
(a)
2.2 Wetting front

Valve
Soil Tensiometer
Theta-probe
In this study, sandy silt derived from completely decomposed granite was used. The wet soil appeared as a different color comparing to the dry soil as shown in Fig. 2a. During the wetting process, the dry soil in the soil column turned darker as shown in Fig. 2b. A clear edge appeared between the wetted soil and the initially drier soil. This demarcation is called the wetting front and can be clearly identified and measured during the wetting process. The location of the wetting front was taken from two graduation labels placed on the outer surface of the column cell (See Fig. 2b). 2.3 Experimental procedure In the wetting process simulation using the soil column device, the soil specimen was prepared in two steps. First, oven-dried soil was exposed in the air for two days and the water content was measured. Second, the soil was dynamic compacted into the column in layers, each layer being 40 mm thick. Four Theta-probes for water content measurement were installed into the soil during soil compaction. After compaction of the soil specimen was complete, a flux boundary condition was applied. In a capillary-rise test, a constant water head about 5 mm higher than the bottom of the soil was applied at the bottom of the soil column. The column at the beginning of a capillary-rise test is shown in Fig. 1b. A wetting front (Fig. 2b) can be observed and recorded. When the soil suction at a monitoring section was estimated to be equal to or smaller than about 85 kPa, a miniature-cup tensiometer limited in a measurable suction range of 085 kPa was installed in the soil column through
Acrylic disk with holes
Water supply
Figure 1. Soil column device: (a) sketch of the soil column device; (b) photograph of the soil column device.
sors from Delta-T Devices Ltd., UK were used to monitor the water content and four tensiometers were used to measure the soil suction. The sensors were installed along the column cell at heights of 100, 300, 600, and 800 mm above the pedestal.
Figure 2. Observed wetting front: (a) the colors of wet soil and dry soil; (b) observed wetting front in a wetting process.
690
a pre-drilled hole. In an infiltration test, a constant water head (e.g., 100 mm higher than the top of the soil column) was applied at the top of the soil column. The infiltration test was terminated when steady-state conditions were achieved, (i.e., when the water flowed out from the bottom of the soil column at a constant flow rate). 2.4 Data interpretation
reaches section F2. The flux passing through section B between two time intervals can be calculated as, QB WB A QA (1)
Figure 3 shows a case of capillary rise from the bottom of the column. The top of the soil column is at section A. The water content and suction measurement sensors were installed at section B. During the wetting process, the wetting front is observed to rise with time. The recorded heights of the wetting front are plotted in Fig. 4. Figure 3 shows that at time t1, the wetting front reaches section F1 and at time t2, the wetting front
where QB is the flux passing through section B from t1 to t2; QA is the flux passing through section A from t1 to t2, which is assumed to be zero during the capillary-rise process; WB-A is the water storage change between section B and section A in this period and can be calculated as, WB A
hA hB
(h,t2 )Acdh (h,t1 )Acdh

hB
hA
(2)
where (h, t) is the water content distribution function; Ac is the cross-sectional area of the soil column; t is time; h is the section height from the bottom of the column; hB is the height at section B; and hA is the height at section A. Figure 3 shows that in a short period of t (i.e., t2 t1), the wetting front advances a distance of h (i.e., the height difference between F1 and F2). If the water content contour is assumed to advance smoothly during the time interval, then the water content distribution function obeys
( h,t t ) ( h h,t )
and the suction distribution function obeys ( h,t t ) ( h h,t )
(3)
(4)
Then the total flow volume passing through section B in eq. 1 can be written as, QB
Figure 3. Wetting fronts at two moments during a capillary-rise process in an initially air dried soil column.
hA hB
(h h,t1 )Acdh (h,t1 )Acdh

hB hB
hA
hB h
( h,t1 )Ac dh
hA hA h
(5)
( h,t1 )Ac dh
The first part of eq. 5 can be calculated as,

hB hB h
( h,t1 )Ac dh 0.5 ( hB ,t1 ) ( hB ,t2 ) Ac h
(6)
Figure 4.
Wetting front advancing process.
If the initial water content is very low and therefore the coefficient of permeability is extremely small, the water content change in the initially airdried soil zone can be ignored during the limited wetting period. Therefore, the water contents in this zone remain at the initial water content, i, during the entire wetting process. The second part of eq. 5 can be calculated as follows,
691
hA hA h
( h,t1 )Ac dh i Ac h
(7)
QB can be calculated by substituting eqs. 6 and 7 into eq. 5, QB 0.5 ( hB ,t1 ) ( hB ,t2 ) 2 i Ac h (8)
The flow quantity can be written using Darcys law as, Q kiAc dt (9)
conductivities. It is not necessary to have a series of measurement sections to provide water content / suction profiles as is the case for the instantaneous profile method. It should be mentioned that this interpretation procedure requires the wetting front advancing velocity. The method depends on an observed wetting front. Therefore, this method is limited to a wetting test with soils that are initially at high suctions and change color when wetted. 3 EXPERIMENTAL RESULTS
where k is the instantaneous coefficient of permeability at a specific section along the soil column; i is the hydraulic gradient at that section, which is calculated as, i dh d / w d 1 w dh dh (10)
3.1 Wetting process A capillary rise test was conducted to a sandy silt (ML) using the column device. The ML contained 60% fines, 25% sands and 15% gravel. The capillary-rise process in the soil column is shown in Fig. 5. The elapsed time was taken as zero when the water supply was first connected to the soil column. Figure 5 shows that it takes about 3 days for the wetting front to pass through a soil column 400 mm in height, about 8 days through a soil column 600 mm in height, and about 17 days through a soil column 800 mm in height. 3.2 Water content changes during wetting process The water content changes at heights, H, of 100, 300, 600, and 800 mm were monitored during the experiment. The monitored water content profiles are shown in Fig. 6. The water contents in the soil columns show three distinct stages: namely, the initial stage, the wetting stage, and the wetted stage. In the initial stage, before the wetting front reached a particular section, the water content in the section remained at the initial water content corresponding to the placement compacted conditions. After the wetting front reached a particular section, the soil columns turned to the wetting stage. The suctions
1000
The hydraulic gradient at t2 at section B can be obtained by substituting eq. 4 into eq. 10: i ( hB ,t2 ) ( hB h,t2 ) 1 w h ( hB ,t1 ) ( hB ,t2 ) 1 w h
(11)
The average coefficient of permeability between t1 to t2 at section B can be calculated by substituting eqs. 8 and 11 into eq. 9: kave QB iAc (t2 t1 ) ( hB ,t1 ) ( hB ,t2 ) 2i w h2
(12)
2 ( hB ,t1 ) ( hB ,t2 ) w h (t2 t1 )
The wetting front advancing velocity (referring to Fig. 4) is defined as, v h / t Using v instead of h, eq. 12 becomes, kave (13)
800
(hB ,t2 ) (hB ,t1 ) 2i wv2 (t2 t1 ) 2 ( hB ,t1 ) ( hB ,t2 ) wv(t2 t1 )
Wetting front height, h (mm)
600
(14)
400
So far, the unsaturated hydraulic conductivity corresponding to a suction of 0.5[(hB, t1) (hB, t2)] can be calculated using eq. 14. Only the monitored data at one section and the wetting front advancing velocity are required in eq. 14. The monitored data at each section can be used to calculate a series of unsaturated hydraulic
200
400000
800000
1200000
1600000
2000000
2400000
Time, t (s)
Figure 5.
Capillary-rise process in the ML column.
692
0.5
60
(b) ML, s = 0.45 Wetted stage

Soil suction, (kPa)
50
0.4
Volumetric water content,
H = 100 mm H = 300 mm H = 600 mm H = 800 mm
40
0.3
Wetting stage
0.2
30
20
0.1
Initial stage
0.0 0 500000 1000000 1500000
H = 100 mm H = 300 mm H = 600 mm H = 800 mm

2000000
10
0 0 500000 1000000 1500000 2000000 2500000
Time, t (s)
Time, t (s)
Figure 6. Monitored water content profiles at different sections in the ML column.
Figure 7. Monitored soil suction profiles at different sections in the ML column.
in the wetted soil decreased rapidly and large hydraulic gradients were present. Therefore, water was able to enter into the initially dried soil quickly and the water content increased rapidly. Finally, the soil columns came to the wetted stage. The water content in the wetted soil increased slowly and approached the steady-state water content. Tensiometers were inserted into the soil column after the wetting front passed through the respective sections. The monitored suctions were relatively stable and decreased slowly with time. The monitored suction profiles are shown in Fig. 7, where suctions decreased with time and approached a steady-state. At the 100 mm-high section, suction decreased to a constant value of 1.2 kPa, which was close to the steady-state value of 1.0 kPa at that section. 3.3 Measured hydraulic conductivity curve
1E-5
ks = 1x10 - 3x10
1E-6
-7
-6
ks
Coefficient of permeability, k (m/s)
1E-7
1E-8
1E-9
1E-10
1E-11
1E-12
Gardner fit Monitored suction, H = 100 mm Monitored suction, H = 300 mm Monitored suction, H = 600 mm Monitored suction, H = 800 mm Inferred suction, H = 100 mm Inferred suction, H = 300 mm Inferred suction, H = 600 mm Inferred suction, H = 800 mm 1 10 100
= 2.15
1E-13 0.1
1000
10000
Soil suction, (kPa)
The water contents and soil suctions during the capillary-rise or infiltration process are used to calculate the unsaturated hydraulic conductivities using eq. 14. Both monitored suctions and suctions inferred from a SWCC curve can be used. The computed unsaturated hydraulic conductivities for ML are shown in Fig. 8. The Gardner SHCC model (Gardner, 1958) was used to fit the calculated hydraulic conductivities using the saturated hydraulic conductivities measured in the laboratory. If the suction values measured using the tensiometers installed on the column (Fig. 1) are inserted directly into eq. 14, the unsaturated hydraulic conductivities are calculated as shown in Fig. 8 using open symbols. Soil suctions larger than 85 kPa could not be monitored due to the limited measurement range of the tensiometers. The unsaturated hydraulic conductivity calculation based on monitored water contents and soil suctions is therefore limited in a narrow suction range.
Figure 8. for ML.
Unsaturated soil hydraulic conductivities
If the suctions are estimated from the soil-water characteristic curves (SWCCs) based on the water contents monitored using the Theta probes installed on the column (Fig. 1), the unsaturated hydraulic conductivities can be calculated over a much wider suction range as shown in Fig. 8 using solid symbols. The hydraulic conductivities calculated from the monitored suctions and the suctions inferred from the SWCCs agree well. In addition, the hydraulic conductivities interpreted based on the measurements at the four monitoring sections at different heights are similar. The hydraulic conductivities are found to locate along a log-linear hydraulic conductivity band, where the upper value is about 38 times of the lower value at the same suction. The Gardner SHCC model can fit the data quite well.
693
1E-6
(b) ML Coefficients of permeability, k (m/s)

1E-7 1E-8 1E-9 1E-10 1E-11 1E-12 1E-13 0.1
Instantaneous profile method Wetting front advancing method

1 10 100 1000 10000
Soil suction, (kPa)
is required for hydraulic conductivity calculations when using the method proposed. The proposed method makes it possible to measure the unsaturated hydraulic conductivity curve over a wide suction range in a few days, resulting in considerable time-saving as compared to the traditional steady-state method and instantaneous profile method. The proposed method is applicable to a wide range of soil suctions. The proposed method depends on an observed wetting front. Therefore, this method is limited to a wetting test with soils that are initially at high suctions and change color when wetted. ACKNOWLEDGEMENT This research was substantially supported by the Research Grants Council of the Hong Kong Special Administrative Region (Projects No. 622206 and N-HKUST611/03). REFERENCES
Fredlund, D.G., Xing, A.Q. & Huang, S.Y. 1994. Predicting the permeability function for unsaturated soils using the soil-water characteristic curve. Canadian Geotechnical Journal 31: 533546. Gardner, W.R. 1958. Some steady state solutions of the unsaturated moisture flow equation with application to evaporation from a water table. Soil Science 85: 228232. Hamilton, D.C., Gloeckler, G., Krimigis, S.M. & Lanzerotti, L.J. 1981. Composition of nonthermal ions in the Jovian magnetosphere. Journal of Geophysical Research 86: 8301. Klute, A. 1965. Laboratory measurement of hydraulic conductivity of saturated soil. In Methods of Soil Analysis, Part 1., Edited by Black, C.A. et al., Agronomy, 9: 210220. Kool, J.B. & Parker, J.C. 1987. Development and evaluation of closed-form expressions for hysteretic soil hydraulic properties. Water Resources Research 23: 105114. Li, X., Zhang, L.M. & Fredlund, D.G. 2009. Wetting front advancing column test for measuring unsaturated hydraulic conductivity. Canadian Geotechnical Journal 46(12): 14311445. Richards, L.A. 1931. Capillary conduction of liquids through porous medium. Physics 1: 318333. Samingan, A.S., Leong, E. & Rahardjo, H. 2003. A flexible wall permeameter for measurement of water and air coefficients of permeability of residual soils. Canadian Geotechnical Journal 40: 559574.
Figure 9. Hydraulic conductivities calculated using the instantaneous profile method and the wetting front advancing method.
3.4
Comparison with instantaneous method
The instantaneous profile method can also be used to calculate the unsaturated hydraulic conductivities. The calculated unsaturated hydraulic conductivities for ML are compared in Fig. 9. Comparing to the hydraulic conductivities calculated using the wetting front advancing method, the hydraulic conductivities calculated using the instantaneous profile method are located in a limited suction range of 1 to 100 kPa. The flow rate used in this study was too fast to measure the hydraulic conductivities at suctions above 100 kPa using the instantaneous profile method. A low flow rate test could be used to measure the hydraulic conductivities at suctions above 100 kPa using the instantaneous profile method, which on the other hand would be time-consuming. More research on such a comparison for other soils would be of value. 4 SUMMARY AND CONCLUSIONS
A wetting front advancing method is developed for the measurement of unsaturated soil hydraulic conductivities. In the new method, an initially air dried soil column is subjected to a wetting event. During the wetting process, a wetting front advances with time and gives a wetting front advancing velocity function. With the help of wetting front advancing velocities, the monitored water contents/suctions can be used to calculate a series of unsaturated hydraulic conductivity. Only one measuring section
694
Advances in tensiometer-based suction control systems

S.D.N. Loureno D. Gallipoli
School of Earth and Ocean Sciences, Cardiff University, Cardiff, UK Department of Civil Engineering, University of Glasgow, Glasgow, UK
D.G. Toll & C.E. Augarde F.D. Evans
School of Engineering and Computing Sciences, Durham University, Durham, UK Controls Testing Equipment Ltd, Wykeham Farrance Division, Tring, Hertfordshire, UK
ABSTRACT: Cunningham (2000) and Jotisankasa (2005) pioneered the development of tensiometerbased suction control systems. In these systems, wetting and drying of the soil are achieved by water injection and circulation of air in contact with the specimen while suction is monitored by sample-mounted high suction tensiometers. Unlike the axis translation technique, these systems avoid using elevated air pressures and better reproduce the drying and wetting conditions occurring in the field. Building upon these earlier works, this paper describes an automated tensiometer-based suction control system that enables direct measurement of water content changes inside the sample. A diaphragm pump forces air to flow inside a closed loop that runs across the sample while a moisture trap ensures that the relative humidity of the circulating air is kept low. As the circulating air dries the soil, the amount of abstracted water is measured by continuous weighing of the desiccant inside the moisture trap. Wetting of the sample is instead achieved by controlled injection of water through a solenoid valve connected to a pressurized volume gauge. The changes of soil water content are given by the difference between the amounts of water injected by the volume gauge and that retained by the desiccant. The system is used to impose cycles of drying and wetting on compacted clayey specimens and results from preliminary tests are presented. 1 SUCTION-CONTROLLED TESTING OF UNSATURATED SOILS In the suction control system presented in this paper, the measurement devices are: a) a samplemounted tensiometer for measuring pore water pressure on the specimen surface, b) a continuously weighted moisture trap for measuring removal of water from the system and, c) a pressurized volume gauge for measuring addition of water into the system. Controlled wetting or drying of the sample is achieved by forced circulation of dry air or direct injection of liquid water inside a closed loop that runs across the sample. Unlike the axis translation technique, the proposed system does not require application of elevated pore air pressures. It therefore ensures testing conditions that closely resemble those existing in the field allowing development of negative pore water pressures below the water cavitation threshold. The use of sample-mounted high suction tensiometers provides direct measurements of negative pore water pressures on the sample surface. This overcomes limitations due to the indirect nature of suction measurements in the osmotic and vapour
The selection of an appropriate suction control technique for testing unsaturated soil samples depends on the range and type of suction (i.e. total or matric) to be controlled. Conventional techniques for controlling suction in unsaturated soils (e.g. axis translation, osmotic flow and vapour equilibrium) allow direct imposition of suction at the boundary of the sample. In recent years, however, alternative techniques have been developed to provide indirect control of suction or water content. These use a computerized feedback system that includes a device to measure pore pressure or water content and a control device to impose wetting or drying. Software is usually employed to compare measurements with target values and activate control as required. The feedback system continuously adjusts the control device to maintain the measured variables within a set tolerance of the target.
695
equilibrium techniques, where suction is correlated to the chemical concentration of a control fluid or to the relative humidity in equilibrium with a saline solution. Cunningham (2000) and Jotisankasa (2005) were the first to develop a suction control system for laboratory testing of unsaturated soils based on circulation of air along the sample surface and direct suction measurement by sample-mounted miniature tensiometers. The suction range of these systems depends on type and pre-conditioning of the high suction tensiometer employed but it can typically cover negative pore pressures down to 1.5 MPa, which is adequate for most geotechnical engineering applications. This contribution initially reviews two tensiometer-based systems developed by Cunningham (2000) and Jotisankasa (2005) to control drying or wetting of the sample by means of air circulation. It then describes an enhanced suction control system, which has been developed in the present work. The enhanced system operates along similar lines as those of previous proposals but incorporates modifications to improve measurement of sample water content and effectiveness of drying or wetting. 2 PREVIOUS TENSIOMETER-BASED SUCTION CONTROL SYSTEMS
Figure 1.
Drying system by Cunningham (2000).
Cunningham (2000) was the first to develop a tensiometer-based system that employed air circulation for controlling suction in triaxial tests (Figure 1). The system was only able to dry the soil and it could therefore be employed for the application of constant suction stress paths only if pore water pressure showed a tendency to increase during loading at constant water content. The system used two sample-mounted tensiometers to read suction at the side and at the top of the specimen. Air was circulated along the base of the sample by applying a positive pressure at one end of the cell pedestal while leaving the other end open to atmosphere. A spring valve, placed in the tubing, automatically opened when air pressure exceeded a given value. The system also included a transducer to monitor pressure inside the air line. If the average pore water pressure measured by the top and mid-height tensiometers was higher than the target value, the control software would automatically open the spring valve by raising the line pressure so that air would start flowing and the sample would start drying. When pore water pressure dropped back within a set tolerance of the target, the control system would lower air pressure again to close the valve and stop air circulation leaving the sample to equalize.
Figure 2. System by Jotisankasa (2005); a) drying, b) wetting.
Jotisankasa (2005) further developed the system designed by Cunningham (2000) through incorporation of a relative humidity sensor at the outlet of the air line and by extending the system to impose wetting as well as drying (see Figure 2). The system
696
developed by Jotisankasa (2005) was also able to control suction automatically at a target value. Similarly to Cunningham (2000), the system developed by Jotisankasa (2005) for soil drying imposed air circulation along the base of the sample and used two tensiometers to measure pore water pressures at two different points on the surface of the sample (see Figure 2a). Changes of water content were estimated by comparing measurements of relative humidity at the outlet and inlet of the air line running along the sample base. Wetting was initially based on the circulation of moist air but, as this proved ineffective, it was subsequently replaced by direct injection of liquid water (see Figure 2b). Water was manually injected in stages at the top of the sample by using a peristaltic pump. After each injection, the readings of suction and strains were monitored until stable values were achieved before moving to the next injection stage. 3 3.1 NEW TENSIOMETER-BASED SUCTION CONTROL SYSTEM General set-up The adoption of a closed air circulation loop, such as that shown in Figure 3, achieved a twofold objective of maximizing the sample surface exposed to air flow while, at the same time, isolating air circulation from the surrounding environment. In this way, changes of sample water content could be evaluated by measuring the respective amounts of water captured by the moisture trap and injected by the volume gauge. A hygrometer was also placed in the loop to measure the relative humidity of the circulating air. An automated feedback system using the software TRIAX (Toll, 1999) was employed to impose suction controlled stress paths to the sample. Manual control could be alternatively adopted by relying on a person to take measurements and operate the control devices. In addition to controlling pore water pressure, the computerized triaxial system could also impose target values of confining pressure, vertical load and water content while measuring air pressure, volumetric strain and axial strain in a fully automated way. The system included a double walled triaxial cell to measure sample volume changes during tests on unsaturated soils. 3.2 Drying control
Figure 3. samples.
System for drying and wetting of soil
The suction control system here proposed is similar to those developed by Cunningham (2000) and Jotisankasa (2005) but includes additional features to improve effectiveness of air circulation during drying or wetting and to provide more accurate measurements of the changes of the soil water content. The proposed system was developed for controlling suction during triaxial tests but could be adapted for use with other types of equipment, such as oedometers or shear boxes. A schematic representation of the system is presented in Figure 3, which shows that a single high suction tensiometer is used to measure pore water pressures at the base of the sample. The high suction tensiometers used in this work were designed by Durham University and Wykeham Farrance Limited (Loureno et al., 2006) and were able to read minimum values of pore water pressure between 1.5 MPa and 2 MPa. Comparison of Figures 1, 2 and 3 indicates that the proposed experimental set-up differed from previous systems because air was circulated inside a closed loop running through the entire sample rather than across an open line at the specimen base. During drying, relative humidity inside the closed loop was kept low by means of a moisture trap containing silica gel (a commonly used laboratory desiccant) through which air flowed. During wetting, controlled amounts of liquid water were injected inside the air circulation loop while the moisture trap avoided condensation in the tubes.
During drying, air at low relative humidity was circulated inside the closed loop shown in Figure 3 without water injection.
697
The moisture trap ensured that water vapour was continuously removed from the circulating air so that relative humidity was kept low and drying potential remained high. The moisture trap consisted of granular silica gel (with an approximate dry mass of 400 g) continuously weighted by a digital balance sealed inside an air-tight metal box, which was connected to the air circulation loop (see Figure 3). By assuming that all water removed from the sample is captured by the silica gel, it was possible to relate any increase in weight registered by the digital balance to a decrease of soil water content. The moisture trap proved particularly efficient in drying the air due to the high specific surface area of the silica gel. Figure 4 shows the results from a preliminary experiment carried out on a sample of approximately 400 g dry silica gel left exposed to the laboratory atmosphere at an ambient relative humidity of about 55% over a period of several days. Figure 4 indicates that the water absorption rate decreased with time as the silica gel came into equilibrium with the environmental humidity after an increase in weight of approximately 32 g (corresponding to a water content of about 8%). This experiment confirmed that relatively small amounts of silica gel could effectively remove vapour from the surrounding air even when exposed to a lower relative humidity than that existing inside the air circulation loop of the proposed suction control system. Some difficulties were encountered in making the metal box of the moisture trap completely vapour leak-proof. In one experiment, about 785 g of dry silica gel was placed on the balance and sealed in the air-tight metal box with all connections shut. Despite taking precautions to prevent leakage of vapour into the box from the outside environment, the moisture absorbed by the silica gel increased with time with an apparently linear trend (see Figure 5). This, however, does not pose
791 790
mass (g)
789 788 787 786 0 1 2
time (day)
Figure 5.
Leakage though moisture trap metal box.
35 30 mass of water (g) 25 20 15 10 5 0 0 1 2 3 time (days) 4 5 6
Figure 4. Adsorption of water by approximately 400 g of initially dry silica gel at an ambient relative humidity of about 55%.
an insurmountable problem as long as the leakage rate is reasonably constant over time and can therefore be calibrated. A diaphragm pump was used to force air around the loop running across the sample and the moisture trap. Unlike the previous systems proposed by Cunningham (2000) and Jotisankasa (2005), where air was pumped along the specimen base, flow was imposed in this case across the entire sample height to maximize soil exposure to drying and to avoid large spatial variations of pore water pressure across the height of the specimen. During initial tests, it was evident that a large difference of pore air pressure could occur between the two extremities of the sample depending on the sample air conductivity, which is in turn controlled by the intrinsic permeability and degree of saturation of the soil. This is of course undesirable as it creates a variable net stress field across the height of the sample. In order to ease flow and to reduce pore air pressure gradients, geotextiles were subsequently placed between the lateral surface of the sample and the latex membrane (Figure 3). Controlled drying was achieved by setting a target value for the pore water pressure measured by the tensiometer at the bottom of the sample. The software TRIAX automatically activated the pump, initiating flow of dry air, when pore water pressure rose above target and stopped air circulation when pore water pressure dropped below target. Figure 6 shows a typical test performed on a sample of compacted sandy clay, which was dried to a target pore water pressure of 300 kPa from an initial almost saturated condition while subjected to a cell pressure of 300 kPa. Once the measurement of pore water pressure reached the target, the control software continued switching the pump on or off as the pressure read by the tensiometer became either higher or lower than target (within a set tolerance of 1 kPa). Given that tensiometer readings did not respond immediately to the activation or deactivation of air flow, the measured pressure often exceeded the set tolerance band by
698
pore water pressure (kPa)
0 -50
0 -100 -200 -300 -400 -500 0 6
wetting drying 1injec/3min
cumulative drying injected water cumulative trapped water
6 mass of water (g) 5 4 3
-150 -200 -250 -300 -350 0 5 10 time (hours) 15 target pore water pressure = 300kPa
pressure (kPa)
-100
pore water pressure targets
pore water pressure
2 1 0
12
18 24 30 time (hours)
36
42
48
Figure 6. Automatic drying to a pore water pressure of 300 kPa.
Figure 7. Automatic wetting and drying cycle.
few kPas. Nevertheless, the degree of control still appears of acceptable accuracy. Figure 6 also shows that the periods of the subsequent pressure cycles increased as the soil water content stabilized at the target pore water pressure of 300 kPa, signifying a progressive equalization across the sample. 3.3 Wetting control
An initial attempt was made to wet the soil by circulation of humid air but this proved unsuccessful as the sample tended to become drier rather than wetter. In addition, significant condensation occurred in the tubing, which made impossible to quantify accurately the amount of water retained by the sample. Because of these difficulties, it was then decided to wet the sample by automatic injection of liquid water into the air circulation loop through a solenoid valve connected to a pressurized volume gauge. The solenoid valve was opened at regular intervals so that consecutive water pulses of approximately 0.05 cm3 were introduced from the pressurized volume gauge into the loop. Air was circulated to help distribution of water on the sample surface and to facilitate moisture adsorption by the soil. The moisture trap avoided condensation or accumulation of water inside the tubes during wetting. Any change of soil moisture was then quantified as the difference between the water injected by the volume gauge and that retained by the moisture trap. The silica gel acted as a sink preventing storage of unaccounted moisture inside the system, hence ensuring that the any difference between injected and retained water could be entirely attributed to changes in the sample moisture content. A typical wetting-drying cycle is shown in Figure 7, where a sample of compacted sandy clay was initially wetted to a target pore water pressure of 200 kPa and then dried back to a target pore water pressure of 400 kPa, while subjected to a cell pressure of 300 kPa.
As water injection progressed at a rate of one pulse every three minutes, air was continuously circulated from the top to the base of the sample and into the moisture trap along the closed loop shown in Figure 3. Figure 7 indicates that the pore water pressure measured at the base rose while water injection took place (see light grey shadow in Figure 7). As soon as pore water pressure reached the target of 200 kPa, water injection was stopped but pressure kept on rising overshooting the target by more than 100 kPa. Circulation of dry air was maintained while the target had been overshot (see dark grey shadow in Figure 7), forcing eventually the soil to dry and the pore water pressure to be corrected back towards the target. Air circulation was stopped as soon as the tensiometer read a value of 200 kPa but, again, this did not prevent the pressure from reducing further, hence undershooting the target. Pore water pressure eventually stabilized at approximately 250 kPa and air circulation was then started again to dry the sample to a pressure of 400 kPa. Once more, the reading of the tensiometer undershot the target by about 50 kPa despite air circulation had been stopped as soon as the pressure target had been achieved. The results observed during the drying path in Figures 6 are somewhat different from those observed during the drying paths in Figure 7. In the former case, the measured pore pressure never significantly undershot the target while this is not true for the drying paths shown in Figure 7. In both instances, air circulation was stopped when a water pressure equal to the target was measured at the base of the sample. In the test shown in Figure 6, however, air was circulated from the base to the top of the sample while, in the test shown in Figures 7, air was circulated in the opposite direction. Hence, in the latter case, pore water pressure was highest at the base because the above soil had been exposed to drier air coming from the top. This would explain the decrease of pore water pressure
699
at the base as pore water pressure tended to equalize throughout the soil mass following the end of air circulation. It was also found that wetting was much more difficult to control than drying and target pressures were overshot by a large margin during wetting regardless of the direction of air flow and the location where pore water pressure was measured. Figure 7 also shows the two curves giving the cumulative amounts of water injected by the volume gauge and absorbed by the moisture trap. Assuming no external water losses or gains, the variation of water content inside the sample at equilibrium could be measured as the difference between these two curves. 4 CONCLUSIONS
The paper presents a computerized system that uses automated air circulation and water injection to control suction during triaxial tests on unsaturated soil samples. Unlike the axis translation technique, the proposed method does not require application of elevated air pressures and enables development of negative pressures inside the soil pores, which closely replicates real field conditions. In addition, unlike other suction control techniques where pore air is maintained at atmospheric pressure, pore water pressures are not correlated to relative humidity or osmotic pressure but they are directly measured by a miniature tensiometer at the base of the sample. This work extends the suction control systems previously developed by Cunningham (2000) and Jotisankasa (2005). It introduces additional features to improve measurement of water content changes inside the sample and to enhance effectiveness of drying and wetting. In the present system, air is circulated inside a closed loop across the height of the sample rather than across an open line at the specimen base. A moisture trap is incorporated within the air circulation loop to keep relative humidity low during drying and to avoid water condensation during wetting. A new approach is also employed to obtain continuous measurements of injected water volume and water mass absorbed by the moisture trap. Changes in the sample water content are evaluated as the difference between these two measurements. Drying or wetting of the sample is governed by means of a computerized feedback routine, which commands air circulation and water injection
based on the comparison between measured and target values of pore water pressure. Results from initial wetting and drying tests on compacted clay samples are presented to demonstrate the performance of the proposed suction control system. The system is capable of accurately imposing drying of the sample to a target suction level. However, the inertia of wetting paths is more difficult to control and, when pore water pressure is increased, the target is often overshot by a significant amount. Further research is required to devise a wetting control method that is not only capable of increasing the pore water pressure but also of accurately halting or reversing the pressure trend at any point during the imposed wetting path. Although the suction control system presented in this paper was specifically designed for use inside a triaxial cell, similar principles could be equally employed for controlling suction in other testing equipment such as oedometers or shear boxes. ACKNOWLEDGEMENTS This research was funded by Engineering and Physical Sciences Research Council (EPSRC) of the United Kingdom through a CASE research grant. The support from the European Commission via the Marie Curie Research Training Network contract number MRTN-CT-2004506861 is also gratefully acknowledged. The Authors also wish to thank Mr C.B. McEleavy and Mr S. Richardson for their useful technical support. REFERENCES
Cunningham, M.R. (2000). The mechanical behaviour of a reconstituted unsaturated soil, Ph.D. Thesis, Imperial College of Science, Technology and Medicine, London. Jotisankasa, A. (2005). Collapse behaviour of a compacted silty clay, PhD Thesis, Imperial College. Loureno S.D.N., Gallipoli D, Toll D.G. and Evans F.D. (2006). Development of a commercial tensiometer for triaxial testing of unsaturated soils, Geotechnical Special Publication No.147, Reston, ASCE, Vol. 2, 18751886. Toll, D.G. (1999). A Data acquisition and control system for geotechnical testing, Computing developments in civil and structural engineering (eds. B. Kumar and B.H.V. Topping), Edinburgh: Civil-Comp Press, 237242.
700
A new permeameter for the determination of the unsaturated coefficient of permeability

M.P.H. Moncada & T.M.P. de Campos
Catholic University of Rio de Janeiro, Rio de Janeiro, Brazil
ABSTRACT: This paper describes design details of a flexible wall permeameter for measuring the water permeability of unsaturated soils. The new permeameter works with constant head or constant flow methods, i.e., the common steady-state methods. The suction control is based on the axis translation technique. Total volume change of soil specimens during the test are determined by an accurate measure of the volume change of the confining cell liquid. The applicability and limitations of the new system are illustrated with analyses of test measurements conducted in one tropical soil.
INTRODUCTION
In saturated soils, the water coefficient of permeability is related to the voids ratio of the soils. Changes in the stress-state variables result in changes in void ratios of soils and therefore in the magnitude of the water coefficient of permeability. In unsaturated soils, the water coefficient of permeability is a function of both the voids ratio and the degree of saturation of the soil. In saturated permeability tests, the total volume change of the soil specimen can be directly inferred from the volume of water entering or leaving the soil specimen. In unsaturated permeability tests, the total volume change of the soil specimen is due to the change in volume of air and of water in the soil specimen. However, it is not easy to measure the change in volume of air, as it is very sensitive to pressure and temperature. Laboratory measurements of the unsaturated coefficients of permeability of soils involve the use of high-air entry disks as a separator between the two immiscible fluids, water and air. The tests can be divided into two main categories, steadystate and unsteady-state methods. Both categories involve the measurement of flow rates of the fluids through soil pores. Though it is time consuming, the steady-state method is the most accurate for the determination of the coefficient of permeability following Darcys law (Benson & Gribb 1997). Several researchers have measured the water and air coefficients of permeability of unsaturated soils, e.g. Klute (1965), Barden & Pavlakis (1971), Fleureau & Taibi (1994), Huang et al. (1998), Gan & Fredlund (2000), Samingan et al. (2003), Bicalho et al. (2005), Lu et al. (2006).
The goal of this paper is to describe a new flexible wall permeameter for measuring the water coefficient of permeability of unsaturated soils. 2 DESCRIPTION OF THE FLEXIBLE WALL PERMEAMETER
The flexible wall permeameter system, shown in Figure 1, consists of a permeameter cell, pressure and flushing lines, water pump flow and a total volume change measuring device. Data collection is via a data acquisition system (DAS in Figure 1) at constant time intervals. The flexible wall permeameter chamber comprises a thick metallic wall cylinder to prevent volume change of the chamber under cell pressure variations. To minimize volume changes of the metallic chamber due to temperature variations, testing is run in an air conditioned room with temperature control. The top cap and pedestal were specially designed to have a reservoir and a recess for the high-air entry disks, and an annular recess for the air porous stones. Two 1-bar HEV disks with an average coefficient of permeability of 1.07 106 cm/s were installed on the top cap and pedestal using epoxy glue. Annular grooves of 1 cm wide and 1 cm deep were also cut onto the face of top cap and pedestal (Figure 2). The annular grooves were designed to allow an uniform distribution of the air across the cross section of the soil specimen and to separate the air flow and water flow such that the equilibrium time of matric suction across the sample is reduced. Both, the top cap and pedestal were each connected to an air pressure line, a water pressure line, and a flushing line. The top and
701
Figure 1.
Permeameter system.
Figure 3.
Total volume change system.
Figure 2.
Pedestal and top cap.
bottom water pressures were applied through volume change transducers type Imperial College. The water pressures were monitored using a pore-water pressure transducer (pressure range: 01000 kPa) and a differential pressure transducer (differential pressure range: 1000 to 1000 kPa) connected to the top-bottom connection line.
The air pressure lines were connected to an air pressure controller (pressure range: 2120 psi; 13,78827,37 kPa). A pressure transducer (pressure range: 01000 kPa), and a differential pressure transducer (differential pressure range: 1000 to 1000 kPa), was also attached to the top and bottom pressure lines. The confining pressure of the soil sample was provided by an air-water interface and monitored using a pressure transducer (pressure range: 01000 kPa). The top and bottom flushing lines were designed in such a way that diffused air collected in the space between the top cap, pedestal, and the highair entry disks could be flushed out of the system (Carrillo, 2000). Two automatic volume change measuring apparatuses were used for water flow measurements.
702
The total volume change of soil specimen was inferred from the change in the volume of water inside the pressure cell. The total volume change system, shown in Figure 3, consists of a balance (0,001 g), a reservoir and an air-water interface (the same used for controlling the confining pressure). However, the volume measurement was found to fluctuate with the ambient temperature of the laboratory room in which the tests were conducted. This was observed and a correction procedure was adopted. 3 EXPERIMENTAL PROCEDURE AND ANALYSIS
The unsaturated coefficient of permeability with respect to the water phase was determined after the unsaturated consolidation stage was completed. There are two methods to determinate permeability, i.e., constant head and constant flow methods. In this study both were used. In the constant flow method, a controlled water flow rate is applied at the bottom of the permeameter, while the water pressure at the top is maintained constant. The induced pressure change is measured. The air pressures applied, on the top and on the bottom, are the same. The water inflow and outflow, the total volume change and the induced pressure change, were continuously monitored until the steady-state condition was established, i.e., when the outflow rate was approximately equal to the inflow rate and the induced pressure change was approximately constant. In the constant head method, a controlled hydraulic gradient was applied by reducing the top water pressure or increasing the bottom water pressure such that the difference was equal to the pressure head corresponding to the desired hydraulic head gradient. The air pressures applied, on the top and on the bottom, were the same. The water inflow and outflow, the total volume change and the induced pressure change, were continuously monitored until the steady-state condition was established, i.e., when the outflow rate was approximately equal to the inflow rate. In this study, the net confining stress was maintained at a constant value since only the relationships between the coefficients of permeability and matric suction were examined. The water coefficient of permeability was computed using Darcys law, which is also valid for unsaturated flow. Numerical modeling of flow within the soil specimen was conduced in the validation phase of the permeameter system. Modeling was conduced using the multi-phase fluid and heat flow simulator TOUGH2 (Pruess et al., 1999). The use of a
multi-phase flow code allows taking explicitly into account the air flow to the simulated permeameter tests. Possible influences of the air phase in the test results are therefore incorporated. For reasons of space, the results of these simulations are not presented here. They can be found in Moncada (2008). In order to apply Darcys law, and based in the simulation results, it was adopted a corrected (modified) area, smaller than the cross-sectional area and equal to the HEA disks area. Also, as the coefficient of permeability of the HEA disks may affect the determination of the water coefficient of permeability of the soil specimen, the impedance of the disks was taken into account. The water coefficient of permeability of the soil specimen was determinated by treating the disk-soil-disk arrangement as a three-layered system, using the following equation: kp
H k
Hs
Ht kt
Hb kb
(1)
where Hs is the height of the soil specimen; Ht is the thickness of the top HEV disk; Hb is the thickness of the bottom HEV disk and H is the total height of this system, i.e., H Ht Hs Hb; kt is the coefficient of permeability of the top disk, kb is the coefficient of permeability of the bottom disk; and k is the coefficient of permeability of the three-layered system. To calculate the average (the reference) matric suction corresponding to the unsaturated hydraulic permeability it was adopted the following equations proposed by Steger et al. (2008): s 1 1 sref ( sb st ) ( sb st ) if ( sb st ) b 2 8 2 and s 1 1 sref ( sb st ) ( sb st ) if ( sb st ) b 2 6 2 (3) (2)
where sb higher suction at the top (or bottom), and st lower suction at the bottom (or top). 4 RESULTS AND DISCUSSION
The flexible wall permeameter system was used to determinate the water coefficient of permeability for one Brazilian tropical soil. The basic soil properties of the residual soil used are listed in Table 1. Saturated permeability tests were performed under constant gradient using a conventional triaxial cell.
703
Table 1.
Average soil properties. Residual soil 15.57 2.691 0.726 1.14 106 62.0 44.0 1.5 27.2 27.1 44.3
Soil property Dry unit weight, d (kN/m3) Specific gravity, Gs Voids ratio, e0 Saturated permeability, ks (cm/s) Limit liquid (%) Plastic limit (%) Gravel (%) Sand (%) Silt (%) Clay (%)
Figure 4.
Water retention curve of the residual soil.
The water retention curve of the soil, shown in Figure 4, was obtained from pressure plate and filter paper tests. The van Genuchten equation (1980) was used to get the curve that fits the data shown in Figure 4. The air entry values and other relevant parameters associated to this curve are summarized in Table 2. Block samples of 200 mm of side were obtained in the field. Permeability testing samples were trimmed carefully to the required diameter (100 mm) using a sharp blade. Then, each specimen was inserted in a 100 mm mould and their ends were trimmed flat. The saturation system process consisted of the saturation of the HAE disks and the saturation of the confined system, water lines, etc. The disks were saturated by applying a reservoir pressure (or back pressures) smaller that the cell pressure Water was allowed to flow through the disks until having a constant outflow rate. Air was dissolved into the water or would be collected into the water reservoirs beneath the disks. Then, the water in the system was flushed and the permeameter was ready for testing. In most tests, the specimens were in their natural conditions, i.e., at their natural moisture content. In cases in witch this was not possible (very dry specimens) the sample was submitted to a wetting process until the required moisture water content was reached. The wetting process was executed using a pressure plate cell. Then the specimen was placed on the bottom HAE disk in the permeameter cell. After, the top cap was placed on the soil specimen and the sample was laterally covered with a rubber membrane secured by O-rings. The cell was then carefully filled with water in order to avoid any presence of air inside the chamber. After set up, the soil specimen was isotropically consolidated to the desired net normal stress 3 ua, and matric suction, ua u w, where 3 denotes the applied confining pressure and ua and uw represent the applied pore-air and pore-water
Table 2. Water retention curve parameters. Parameter Air entry value (kPa) Saturated gravimetric water content (%) Residual suction (kPa) Residual gravimetric water content (%) Residual soil 300 5000 25 19.20 1000 26466.55 19.20 3.92
Table 3. Specimen dimensions and applied stresses. Method test no. Constant head 1 2 3 4 Const. flow 7 8
Height (cm) 3 3 3 3 6 6 3 6 40 17 33 72 47 47 15 30 uw (kPa) Top 48 20 35 43 51 51 16 32 Bottom ua (kPa) 95 95 95 95 76 76 95 75 Top 95 95 95 95 67 75 96 77 Bottom ua uw(1) (kPa) 60 76 72 37 26 25 78(2) 44(2) 3 ua (kPa) 139 139 120 120 120 120 152 120 (1) Average matric suction; (2) In equalization stage.
pressures, respectively. During the unsaturated consolidation process, water drained out of the top and bottom water pressure lines. The unsaturated consolidation process was completed when there was negligible water flow either into or out of the soil specimen and its total volume was not changing anymore. After the consolidation stage, water permeability measurements were taken with the constant head method (CH) or the constant flow method (CF). The stress states and the methods used in the tests performed are shown in Table 3.
704
Table 4. Test No.
Experimental data for soil specimens. 1 2 3 4 5 6 7 8
Matric suction(1), ua u w (kPa) Rate flow (cm3/s) inflow outflow Total volume change(2) (cm3) Gravimetric water content (%) Water permeability, kw (cm/s) Total time (h)
60.22 76.80 72.13 37.27 26.24 25.31 74.94 31.54 4.95 106 2.18 106 2.98 106 1.32 105 3.83 105 1.33 105 2.09 105 2.09 105 5.14 106 2.03 105 0.179 20.63 0.193 23.16 0.167 23.06 0.064 25.16 0.235 23.41 0.291 23.71 0.157 22.93 0.344 21.43
4.63 109 8.39 109 1.27 108 4.37 109 5.70 107 6.54 107 6.53 109 1.66 108 158 294 234.2 168.5 297 290 288.7 360.5
(1) Average matric suction; (2) positive value compression.
The reference matric suction, the water coefficient of permeability, the total testing time and the total volume variations are indicated in Table 4. The experimental investigation showed that the measurement of kw under unsaturated conditions requires a fairly long time as compared to conventional saturated measurements. It is noticeable, however, that no important time difference was observed between both the steady-state testing methods employed. The range of kw value that can be measured is limited by the permeability of the HEV disk used. The disks also limit the maximum value of suction that can be used. In the present work that was 100 kPa. The total sample volume change during the water permeability tests was quite small (Table 4) and could be neglected. The water coefficient of permeability versus matric suction for the residual soil is shown in Figure 5. In this Figure, the matric suction has been computed assuming a linear suction distribution throughout the sample, i.e., by taking an arithmetic mean of the top and bottom applied suctions. Table 5 shows a comparison of such mean suction with the reference suctions computed using Equations 2 and 3. The application of Steger et al (2008) equations in Table 5 shows the relative importance of the non-uniform distribution of suction that occurs in the sample to define the matric suction associated to each obtained water permeability coefficient. The soil shows a decrease in permeability with increased suction as expected. However, such decrease is more sensitive to the change in suction than that observed in the corresponding water retention curve (decrease in water content with increasing suction). An air entry value retrieved from the permeability test would be lower than that obtained from the water retention curve. This
Figure 5. Water coefficient of permeability versus matric suction (linear suction distribution).
Table 5.
Average and reference matric suction values.
Water permeability Average suction Reference suction kw (cm/s) (kPa) (kPa) 5.70 107 6.54 107 1.27 108 1.66 108 4.37 109 4.63 109 6.53 109 8.39 109 26.24 25.31 72.13 31.54 37.27 60.22 78.66 76.80 21.09 26.13 61.04 27.25 33.69 50.21 78.22 76.45
would indicate that the use of the water retention curve to obtain the permeability function would not apply in the case of the tested material. Indeed, as shown in Figure 6, that was the case. In Figure 6 the permeability function was fitted with the van Genuchten-Mualem (1976, 1980) model using the
705
Figure 6. Best fit with van GenutchenMualem model.
same parameters ( and n) of the water retention curve, i.e., modifying only the t parameter. The best fittings to the experimental data are not satisfactory for the range of water permeability coefficients measured. A second approach, changing all the curve-fitting parameters, also did not show any satisfactory representation of the experimental data. 5 CONCLUSIONS
The new flexible wall permeameter system has been found to work well for measuring the water coefficient of permeability under suction control. The system also allows the continuous measurement of the total volume change of the sample in different stages of the unsaturated permeability test, with adequate accuracy. The limitations of the system are related to the permeability and air entry values of the HEV disks. Numerical simulations (not presented here) showed that additional corrections are necessary in order to apply the Darcys law in the interpretation of the tests. The experimental results indicated that the time taken to run the test under constant flow rate is equivalent to that taken under constant head. The experimental data showed that, in the low suction range, the permeability function of the tropical soil studied is more sensitive to changes in suction than the corresponding moisture function, with a lower air entry value being found in the first case. Such aspect has to be further investigated. REFERENCES
Barden, L. & Pavlakis, G. 1971. Air and water permeability of compacted unsaturated cohesive soils. Journal of Soil Science. 22. pp. 302318.
Benson, C.H. & Gribb, M.M. 1997. Measuring unsaturated hydraulic conductivity in the laboratory and field. In Unsaturated soil engineering practice. Edited by S. Houston and D.G. Fredlund. American Society of Civil Engineers. Geotechnical Special Publication. No. 68. pp. 113168. Bicalho, K.V., Znidarcic, D. & Ko, H-Y. 2005. An experimental evaluation of unsaturated hydraulic conductivity functinos for a quasi-saturated compacted soil. In: Proceedings of international symposium on advanced experimental unsaturated soil mechanics. Balkema, pp. 325329. Carrillo, D.C.W. 2000. Desenvolvimento de um equipamento para o estudo da permeabilidade efetiva em meios multifsicos. Tese de Doutorado. Pont. Univ. Catlica. do Rio de Janeiro, DEC. Fleureau, J.M. & Taibi, S. 1994. A new apparatus for the measurement of polyphasics permeabilities. In Proceedings of the 1st. International Conference on Environmental Geotechnics (ICEG). Edmonton. Alta. 1114 July. Edited by the Technical Program Committee of the 1st. ICEG. BiTech Publishers. Richmond, B.C. Vol. 1. pp. 227232. Gan, J.K.M. & Fredlund, D.G. 2000. A new laboratory method for the measurement of unsaturated coefficients of permeability of soils. In Unsaturated soils for Asia. Edited by H. Rahardjo, D. Toll and E.C. Leong, A.A. Balkema. Rotterdam. pp. 381386. Huang, S., Fredlund, D.G. & Barbour, S.L. 1998. Measurement of the coefficient of permeability for a deformable unsaturated soil using a triaxial permeameter. Canadian Geotechnical Journal. 35. pp. 426432. Klute, A. 1965. Laboratory measurement of hydraulic conductivity of unsaturated soils. In Method of soil analysis. Edited by C.A. Black, D.D. Evans, J.L. White, L.E. Esminger and F.E. Clark. American Society of Agronomy. Monograph 9. Madison. Wis.. Vol. 1. pp. 253261. Lu, N., Wayllace, A., Carrera, J. & Likos, W. 2006. Constant flow method for concurrently measuring soil-water characteristic curve and hydraulic conductivity function. Geotechnical Testing Journal. Vol. 29. No. 3. pp. Moncada, M.P.H. 2008. Evaluation of hydraulic properties of unsaturated tropical soils. D.Sc. Thesis. Catholic University of Rio de Janeiro (in portuguese). Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research., 12, 513522. Pruess, K., C. Oldenburg & G. Moridis, TOUGH2 users guide, Version 2.0, Lawrence Berkeley Laboratory Report LBL-43134, Berkeley, California, 1999. Samingan, A.S., Leong, E. & Rahardjo, H. 2003. A flexible wall permeameter for measurements of water and air coefficients of permeability of residual soils. Canadian Geotechnical Journal. Vol. 40. No. 3. pp. 559574. Steger, G., Semprich, S., Moncada, M.P.H. de Campos. T.M.P. & Vargas, Jr. E. 2008. A numerical investigation of steady-state unsaturated conductivity tests. In Unsaturated Soils: Advances in Geo-Engineering. Edited by D.G. Toll, C.E. Augarde, D. Gallipoli and S.J. Wheller. Taylor & Francis Group. pp. 747753. van Genuchten, M.T. 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal, 44, pp. 892898.
706
Suction measurements on a natural unsaturated soil: A reappraisal of the filter paper method
J.A. Muoz-Castelblanco, J.-M. Pereira, P. Delage & Y.-J. Cui
Ecole des Ponts ParisTech, UR Navier/CERMES, Universit Paris-Est, Marne la Valle, France
ABSTRACT: Soil suction measurements on an unsaturated soil were performed using the filter paper method and a high capacity tensiometer so as to analyse the reliability of the filter paper technique. The results show that an alternative approach of the filter paper method based on using a previously wetted filter paper can fruitfully be used to measure suction provided an appropriated calibration curve is used. This method was compared to the standard filter paper method in which the paper is initially dry and a good agreement was found between the two approaches, being the wet method somewhat faster in terms of suction equilibration. Suction data also agree well with results obtained independently using a home made high capacity tensiometer. The two methods finally provide a first outlook of the water retention properties of an unsaturated loess from Northern France, observing some evidences of hydraulic hysteresis on it. 1 INTRODUCTION be different in view of the possible differences in either wetting or drying processes. This paper investigates the filter paper method by direct measurement of suction on a natural unsaturated loess from Northern France (Cui et al., 2004, Delage et al., 2005, Yang et al., 2008). Experimental results show a good agreement between both approaches of the filter paper method and also with the tensiometer measures. Shorter periods of suction equilibration were observed for the wet filter paper method. 2 MATERIAL AND EXPERIMENTAL METHODS
Soil suction measurements by the filter paper technique were compared with experimental data obtained by a high capacity tensiometer to analyse the reliability of the filter paper method and additionally to characterize some water retention properties of a natural unsaturated loess. An alternative approach of the filter paper method, proposed by Parcevaux (1980) and based on using a previously wetted filter paper, was experimentally tested to measure suction by utilizing an appropriated calibration curve. Results obtained by this approach were compared to data obtained by the standard filter paper method (ASTM, 2003, D 529803) in which the paper is initially dry. Hysteresis effects in filter paper have been observed, taking into account that initial water content is different for both approaches. Experimental data obtained by different researchers (Fawcett & Collis-George, 1967, Parcevaux, 1980, Hamblin, 1981, Greacen et al., 1987, Ridley, 1995, Harrison & Blight, 1998, Leong et al., 2002) were presented showing the influence of filter paper hysteresis effects in soil suction measuring. Suction equilibration time between filter paper and soil was also analysed for both filter paper approaches. In according with the ASTM procedure, a period of seven days is recommended for the dry filter paper approach whereas there are not enough experimental data for the wet filter paper approach. Because initial water content is different for both approaches, the equilibration time could
2.1 Bapaume loess The soil used in the experiments is an intact and unsaturated loess extracted from a deposit located in an area near to the city of Bapaume (Northern France). These loess deposits were formed during the Quaternary period under periglacial conditions, concerning the aeolian transport of silt particles eroded by a constant North West wind (Antoine, 2002, 2003). This soil is characterized by a relative homogeneity, a low plasticity, a high porosity, a weak carbonate bonding and an open structure that explains its susceptibility to collapse when saturated (Cui et al., 2004, Delage et al., 2005, Yang et al., 2008). Exploration works were made by manual cutting in order to keep the natural structure of loess, considering its high fragility. Intact loess blocks
707
were extracted in cubic boxes of 300 mm side and then protected by paraffin and plastic film to avoid evaporation. The void ratio at natural state is about 0.85, the natural water content is 14% and the initial degree of saturation is 44%. The sand, silt and clay fraction are equal to 2, 82 and 16% respectively. The dominant clay minerals are kaolinite, illite and interstratified illite-smectite (Karam 2006). The carbonate CaCO3 content is 6%. The plasticity index is quite low (Ip 9, wp 37%, wl 28%) and the Casagrande classification of the loess is CL. 2.2 The filter paper method
The filter paper technique was standardized by the ASTM (2003, D 529803) to measure suction values ranging from 0.01 to 100 MPa. Filter paper measurements are performed by placing a piece of filter paper in contact with the soil sample so as to attain suction equilibrium between the filter paper and the sample. The suction value is derived from the calibrated water retention curve of the filter paper. In this work, Whatman No. 42 filter paper was employed. The initial water content under ambient laboratory conditions of the Whatman No. 42 filter paper in original boxes is around 6% (Marinho & Oliveira, 2006) that corresponds to a suction of 29 MPa, which is almost the upper limit for measuring suction. An alternative method using initially wet filter papers was reported by Parcevaux (1980). This approach is here referred to as wet filter paper method compared to the ASTM dry filter paper method. When an initially dry paper is used as recommended by the ASTM procedure, the filter paper is submitted to a wetting path during suction equilibration. Suction measurements with the wet filter paper approach concern drying paths during suction equilibration from a wet state near to 150% of filter paper water content. The tests were performed on samples of 70 mm diameter and 19 mm height. They were cut carefully to get a surface as planar as possible in order to ensure a good contact with the filter paper piece. A set of three filter paper pieces of 50 mm diameter was prepared and placed between two soil samples. The central piece was smaller to avoid pollution due to soil particles. The overall set was then covered with paraffin wax and enveloped in a plastic film to avoid evaporation during the equilibration time. The set was put in an insulated plastic box reducing temperature perturbations. After seven days of suction equilibration, the set was opened and both soil samples and the central piece of filter paper were weighed with a precision balance (1/10,000 g). Weighing was as quick as possible to
avoid any evaporation. Finally, suction was calculated from the filter paper water content using the appropriate calibration curve depending upon the initial state of the filter paper. Figure 1 presents the results of filter paper calibration curves made by different researchers (Fawcett & Collis-George, 1967; Parcevaux, 1980; Hamblin, 1981; Greacen et al., 1987; Ridley, 1995; Harrison & Blight, 1998; Leong et al., 2002) using either drying or wetting processes. Data below 1 MPa were calibrated by suction control techniques such as suction plate, pressure membrane extractor, psychrometer and Richards apparatus whereas higher suctions were controlled with vapour equilibrium technique using saline solutions. The calibration curve of the wet filter paper method was obtained from drying paths and that of the dry filter paper method from wetting paths. These paths correspond to the water transfer processes occurring within the filter paper during the paper/soil equilibration periods, concerning water extraction from the soil in the dry method and water adding from the filter paper to the soil in the wet method. A good agreement is observed in Figure 1 between data obtained independently with different suction control techniques by different authors at different times. The data describe a space between both wetting and drying curves that concerns the hysteresis of the water retention curve of the Whatman No. 42 filter paper. The phenomena of hysteresis in the filter paper have not influence in the soil suction measurement if the correct calibration curve is used for each approach.
1000000 Ridley (1995)
drying
wetting
100000
Harrison and Blight (1998) drying wetting Leong et al. (2002) drying wetting
10000
Suction (kPa)
1000
100
Parcevaux (1980) - drying Fawcett & Collis -George (1967) - wetting Hamblin (1981) - wetting Greacen et al. (1987) - wetting
10
WET FILTER PAPER (Parcevaux, 1980) DRY FILTER PAPER (ASTM D 5298-03)
0.1
20
40
60
80
100
120
140
160
180
200
Filter paper water content, wf (%)
Figure 1. curves.
Whatman No. 42 filter paper calibration
708
2.3
High capacity tensiometer
In order to compare filter paper suction measurements, another suction measurement technique was used. It consists on a high capacity tensiometer developed in the CERMES laboratory, inspired on the original idea of Ridley and Burland (1995) and previously detailed in Mantho (2005) and Cui et al. (2008). The CERMES HCT (high capacity tensiometer), detailed in Figure 2, is an integral strain gauge tensiometer (Tarantino, 2004, Delage et al., 2008) composed of a porous high air entry value (1.5 MPa) ceramic disk with strain gauges glued to a metallic diaphragm and a water reservoir of 0.1 mm thickness. This sensor is characterized by a range of suction measurement from 0 to about 800 kPa. As proposed by Tarantino and Mongiovi (2001), a process of saturation was conducted by the application of 4 MPa of positive water pressure to get rid of any air trapped in the system. Then, the HCT was submitted to a process of cavitation by putting it in contact with a dry sample and then resubmitted to a 2 MPa positive water pressure. As suggested by Mantho (2005), cycles of cavitation saturation are highly recommended to improve the tensiometer functioning. Suction measuring was performed by using a modified oedometer cell that had been already utilized to perform oedometer testing, measuring the suction evolution (Delage et al., 2006, Tarantino and De Col, 2008). The HCT was inserted into the base of the modified cell. A soil sample of 70 mm diameter and 19 mm height was inserted in a metallic ring and placed over the oedometer base. The set
was then placed over a precision balance to register the soil weight changes. A fine layer of loess slurry was placed on the tensiometers surface to avoid the early cavitation of the HCT and to improve the contact between it and the soil specimen. A piston applying 1.5 kPa of vertical pressure was placed over the soil sample to the same reasons. 3 3.1 EXPERIMENTAL RESULTS AND DISCUSSIONS Wetting and drying processes
For both suction measurement methods, hydraulic paths were started from the loess natural water content (14.2%), and followed by either drying or wetting stages. Drying was performed by allowing evaporation of the soil specimen under laboratory conditions for periods of time comprised between 1 and 4 hours. Wetting was attained by adding small quantities of water to the soil sample. In this case, a piece of filter paper was placed on top of the sample and water drops were uniformly distributed over the surface using a syringe in order to ensure homogeneous wetting. Once the target water content was obtained, a new suction measurement was taken. 3.2 Equilibration time for the filter paper method
Figure 2. Peeling off of high capacity tensiometer (HCT CERMES).
Suction measurements on loess samples were performed by using both the dry and wet filter paper methods to compare obtained values and required equilibration times. Testing periods of time varied from 1 to 12 days. Tests were carried out with four samples at the following initial water contents: 23.8, 13.33, 7.88 and 7.19%, corresponding to suction levels about to 13, 46, 144 and 350 kPa respectively. Figure 3 presents the time evolutions of suction for all tests. Each point corresponds to a measurement of the water content of the filter paper at a given time (between 1 and 12 days) after the contact between the filter paper and the soil has been ensured. After each weight measurement, a new filter paper was put in contact with the soil sample and a longer equilibration time was allowed before performing a new measurement. Suction variations with time, related with filter paper water content, are the consequence of water transfers between the soil and the filter paper. For the dry method (resp. wet method), it is observed that the filter paper extracts water from the soil (resp. releases water to the soil). Water transfers tend to stabilize after a few days, generally before the seven days period mentioned in the ASTM recommendation. Monotonic changes in suction are observed at 13.33% and 7.88% whereas some oscillations are
709
Time (days)
20 0 2 4 6 8 10 12
Table 1. Ratio of water exchange between filter paper and soil. wi* (%) 23.80 13.33 7.88 7.19 WFP* ()* 0.74 2.86 9.89 10.02 DFP* ()* 2.04 1.93 2.58 2.15
w = 23.8%
Suction (kPa)
16
Dry filter paper Wet filter paper
12
8 50
w = 13.33%
Suction (kPa)
48
wi*: Initial water content; WFP*: Wet filter paper; DFP*: Dry filter paper; ()*: Water exchange ratio (per thousands).
46
44 180 0 2 4 6 8 10 12
Suction (kPa)
170 160 150 140 480
w = 7.88%
w = 7.19%
Suction (kPa)
440 400 360 320 0 2 4
Time (days)
10
12
Figure 3. Time evolution of suction measured with filter paper techniques.
observed at 23.8% and 7.19%. In the driest samples (w 7.88% and 7.19%), a significant drying of the initially wet filter paper occurs very quickly with relative changes higher than 0.15 g of water mass occurring during the first day. Not substantial reasons were found for these oscillations, especially at the water content level of 7.19% corresponding to 350 kPa of suction around. However, a good agreement in terms of final suction is observed at the end of all the tests with slightly higher values given by the dry filter paper method. This good agreement is also observed in the case where oscillations are observed. In wet samples, the difference appears to be smaller than 1 kPa for suction values of 12.6 (w 23.8%) and 46.8 kPa (w 13.33%). The highest difference is observed at 7.19% (44.6 kPa with the wet paper and 111.5 kPa for the dry one). It is observed a good agreement between the two protocols for the filter paper method with the
minor influence of the presence itself of the filter paper on the measured suction. In Table 1, equilibrated values of water exchange ratio between filter paper and soil are presented for each initial water contents and for each method at all final measurement stages. This rate corresponds to the variation of the mass of water in the filter paper during the FP ) divided by the total mass of measurement (Mw soil ). The low water present in the soil sample (Mw observed values confirm that the filter paper technique is a reliable technique of measuring suction and that both protocols provide comparable results in terms of measured suction. From the point of view of equilibration time and variability of measures, it appears (see Figure 3) that the wet filter paper method gives less variable measures than the dry filter paper method. The wet filter paper method also seems to provide a quantitatively shorter measure of suction time as compared to the dry filter paper method. In almost all cases, measurements confirm the 7 days equilibration time standardized by ASTM. 3.3 Suction measurements Suction measurements were carried out by the HCT and the wet and the dry filter paper methods to compare the obtained experimental results. Hydric paths, consisting on either drying or wetting processes, were started from the natural loess water content of 14.2% that corresponds to an initial degree of saturation of 44% and a suction of 40 kPa as measured by the filter paper (wet or dry) and the HCT. After each change in water content in the HCT hydric paths, an equilibration time of 1 hour was waited for before measuring suction with the HCT. Experimental observations showed 4 to 10 min of HTC response time in wetting stages, and 45 min to 1 hour in drying stages. Various hydraulic paths corresponding to wetting and drying processes into the suctionwater
710
content space were performed. Three wetting hydraulic paths starting from the natural state were made at low suction values (Figure 4). Other three hydraulic paths starting by drying were performed at high suction values (Figure 5). Paths in Figure 4 were firstly conducted by wetting process followed by drying (paths 1, 2 and 3). Paths in Figure 5 started by drying followed by some wetting steps (paths 4, 5 and 6). Observed hydraulic paths show that suction increases currently when water content decreases, and conversely. A good agreement is observed between results obtained by the techniques used. Wetting and drying data obtained from HCT measurements show relatively smooth evolutions of suction with respect to water content.
CONCLUSIONS
Degree of Saturation
100 0.35 0.4 0.45 0.5 0.55 0.6 0.65 0.7 0.75 0.8
Natural state
Wetting stages Drying stages
10
Path 1 - HCT
wetting wetting wetting
drying drying
Path 2 - DFP Path 3 - WFP

1
12
14
16
18
20
22
24
26
Two approaches of the filter paper method were identified in this work. The first one was the standardized technique by ASTM (2003, D 529803) consisted on putting a dry filter paper piece in contact with a soil specimen. Suction values were obtained by measuring the filter paper water content at final stage of the equilibration time between filter paper and soil, which was seven days as recommended by ASTM. The second one was proposed by Parcevaux (1980) consisting on using an initially wetted filter paper piece instead of a dry one. The calibration curves for both methods, giving soil suction from filter paper water content, were presented following the analysis made by Leong et al. (2002). Hysteresis phenomena in filter paper were also identified. The suction equilibration time by both protocols of the filter paper technique was also investigated. Shorter equilibration periods were observed for the wet filter paper approach. Measures confirmed the seven days equilibration time proposed by ASTM. Suction measurements on a natural unsaturated loess were performed by both protocols of filter paper, having experimental data obtained by the CERMES high capacity tensiometer as a reference. Experimental data obtained by both filter paper approaches match well with the HTC measurements. From Figures 4 and 5, it can be seen that drying and wetting branches are not parallels, but apparently converge towards a point corresponding to the natural state (w 14%, s 40 kPa). The gap existing between them is better known as hydraulic hysteresis, and it is more remarkable at higher levels of suction (s 100 kPa). ACKNOWLEDGEMENTS
Figure 4.
Suction (kPa)
Hydraulic paths wetter than natural state.
0 0.05 0.1 0.15 0.2 0.25 0.3 0.35 0.4 0.45
10000
Wetting stages Drying stages
Suction (kPa)
The present study is part of the first author PhD work. It was supported by the European Alan Program of high level scholarships for Latin America, scholarship N E07D402297CO, through grants to Mr. J. Muoz. The support of the French Railways Company SNCF is also acknowledged. REFERENCES
Antoine, P. 2002. Les loess en France et dans le Nord-Ouest europen. Revue Franaise de Gotechnique 99: 321. Antoine, P., Cattt, J. & Somm, J. 2003. The Loess and Coversands of Northern France and Southern England. Journal of Quaternary Science 18: 309318. ASTM International. 2003. Standard Test Method for Measurement of Soil Potential (Suction) Using Filter Paper, D 529803. ASTM International.
1000
Path 4 - HCT
drying
100
wetting
Path 5 - DFP drying wetting Path 6 - WFP drying wetting Natural state
6 8 10 12 14
10
Figure 5.
Hydraulic paths drier than natural state.
711
Cui, Y.J., Marcial, M., Terpereau, J.M., Delage, P., Antoine, P., Marchadier, G. & Ye, W.M. 2004. A geological and geotechnical characterisation of the loess of Northern France. A.W. Skempton Memorial Conference vol. 1: 417428. Cui, Y.J., Tang, A., Mantho, A. & De Laure, E. 2008. Monitoring Field Soil Suction Using a Miniature Tensiometer. Geotechnical Testing Journal 31(1): 95100. Delage, P., Cui, Y.J. & Antoine, P. 2005. Geotechnical Problems related with Loess deposits in Northern France. Proceedings of International Conference on Problematic Soils: 517540. Delage, P., Marcial, D., Cui Y.J. & Ruiz, X. 2006. Ageing effects in a compacted bentonite: a microstructure approach. Gotechnique 56(5): 291304. Delage, P., Romero, E. & Tarantino, S. 2008. Recent developments in the techniques of controlling and measuring suction in unsaturated soils. Keynote Lecture, Proc. 1st Eur. Conf. on Unsaturated Soils, Durham: CRC Press. 3352. Fawcett, R. & Collis-George, N. 1967. A Filter paper Method for Determining the Moisture Characteristics of Soil. Australian Journal of Experimental Agriculture and Animal Husbandry 7: 162167. Greacen, E.L., Walker, G.R. & Cook, P.G. 1987. Evaluation of the Filter Paper Method for Measuring Soil Water Suction. International Conference on Measurement of Soil and Plan Water Status: 137143. Hamblin, A.P. 1981. Filter Paper Method for Routine Measurement of Field Water Potential. J. Hydrol. 53: 355360. Harrison, B. & Blight, G. 1998. The Effect of Filter Paper and Psychrometer Calibration Techniques on Soil Suction Measurements. Proceedings of the Second International Conference on Unsaturated Soils 1, Beijing: International Academic Publishers: 362367. Karam, J.P. 2006. Etude de la rhologie des loess du Nord de la France. Application lvaluation de leur risque
de liqufaction. PhD thesis. cole Nationale des Ponts et Chausses. Paris. Leong, E., He, L. & Rahardjo, H. 2002. Factors Affecting the Filter Paper Method for Total and Matric Suction Measurements. Geotechnical Testing Journal 25(3): 322333. Mantho, A. 2005. Echanges solatmosphre: application la scheresse. PhD thesis. cole Nationale des Ponts et Chausses, Paris. Marinho, F. & Oliveira, O. 2006. The Filter Paper Method Revisited. Geotechnical Testing Journal 29(3): 250258. Parcevaux, P. 1980. tude microscopique et macroscopique du gonflement de sols argileux. PhD thesis. cole Nationale Suprieure des Mines de Paris. Ridley, A. 1995. Discussion on Laboratory Filter Paper Suction Measurements by Sandra L. Houston, William N. Houston, and Anne-Marie Wagner. Geotechnical Testing Journal 18(3): 391396. Ridley, A. & Burland, J. 1995. Measurement of suction in materials which swell. Applied Mechanics Reviews 48(9): 727732. Tarantino, A. 2004. Direct measurement of soil water tension. Unsaturated Soils. In Juc, de Campos & Marinho (eds). Lisse: Swets & Zeitlinger. ISBN 90 5809 371 9. Tarantino, A. & De Col, E. 2008. Compaction behaviour of clay. Gotechnique 58(3): 199213. Tarantino, A. & Mongiov, L. 2001. Experimental procedures and cavitation mechanisms in tensiometer measurements. Geotechnical and Geological Engineering 19: 189210. Yang, C., Cui, Y.J., Pereira, J.M. & Huang, M.S. 2008. A constitutive model for unsaturated cemented soils under cyclic loading. Computers and Geotechnics 35(6): 853859.
712
The use of filter paper method for field suction investigation

M.G.G. Nacinovic, C.F. Mahler, R.L.S. Izzo & P.F.S. Brandt
COPPE/UFRJ/PEC, Rio de Janeiro, RJ, Brazil
A.S. Avelar
UFRJ, Rio de Janeiro, RJ, Brazil
ABSTRACT: Studies related to erosive processes are an extremely important issue nowadays, due to its environmental and economical impacts. The erosive processes were related to suction and monitored in four different land management scenarios (olericulture with conventional tillage, pasture, forest recovery area and native forest) in an agricultural region in the mountainside of Rio de Janeiro, Brazil. Suction was measured using the filter paper method directly in situ, which confirmed to be simple, low-cost and efficient. Field work required a large number of filter papers, in which many were of different batches. For practical purposes, just one calibration curve (Chandler et al. 1992) was used, despite research work advice calibration for different batches of filter paper (Marinho & Oliveira 2006). Electron microscope scanning images of filter paper from the same and different batches were processed to observe its intrinsic properties. In the same batch different characteristics were observed using the electron microscope scanning, what did not produce significative differences in the behavior of the filter paper. Also, two different sizes of filter paper, smaller than the usual, adapted to the equipment developed to measure the total suction were used in the field experiments. The size of the filter paper used did not influence significantly the results. Total suction measurements were made in the four experimental plots during six months using the filter paper method in situ. This paper also presents the procedure used to measure the suction in the field. The results were well correlated with precipitation events data. Finally, characteristic curves were obtained using the filter paper method in laboratory to have better knowledge of the erosive and suction processes. The results were compared with the Fredlund SWCC Device showing good correlation. 1 INTRODUCTION acts attracting soil particles and reducing erosion. In order to evaluate this effect, both laboratory and field assays with filter paper were conducted along with the Fredlund Soil Water Characteristic Curve Device to obtain the characteristic curves (suction versus volumetric water content) of the hydro-erosive plots soils. The filter paper method is based upon the principle that when a soil under humid conditions enters in contact with a porous material, the water will pass from the soil to the porous material until equilibrium is reached. The water molecules need to overcome the matric suction of the soil and the osmotic suction in order to permit water vapor flow. The characteristic curves were obtained by the Van Genuchten (1980) equation using the filter paper assays values. The correlation of soil suction with rainfall events were investigated in 4 hydro-erosive plots with different types of cultivation, management and vegetation cover. Two manual and one digital rain gauges were installed to measure rainfall in the study site area.
The area of the research was composed of an agricultural site in the mountainside of Rio de Janeiro, Brazil. The local consists of variegated evergreen forest vegetation with closed canopy. The type of climate is humid mesothermal with little water deficit, with a well defined dry season during MayAugust. The soils of the region have a high susceptibility to erosion, which is due mainly to the sharp slope. The study of suction is essential for the knowledge of the effective normal stresses of the soil in understanding the erosive processes. The soils resistance to erosion is a function of the humidity and matric suction. The water is retained by capillary forces and surface forces appear due to sorption of water molecules by ions in the soil. When the air phase is continuous (S 85%), the contractile membrane interacts with the soil particles which influences the soils mechanical behavior. The moisture of the soil in the studied area varies throughout the year depending on rainfall. Suction
713
2 2.1
MATERIALS AND METHODS Filter paper assay in situ
The filter papers were arranged in PVC pipes in such a way that there was no direct contact with the inner surface of the pipe, so that the condensed water vapor in its surface would not be absorbed by the filter paper. The tube was composed of a 15 cm piece of 2 PVC solvent weld pipe attached to a 2 cap with an O ring and a disposable needle (30 0.8 mm) affixed to the base of the cap (Fig. 1). The material used followed the model proposed by Mahler & Oliveira (1997). The default size of the filter paper is 7 cm. This filter paper size makes the assays procedures more difficult. A thick spiral wire was needed to hold the 7 cm filter paper (Fig. 2). Three different sizes of filter paper were researched in each experiment of suction measurement 3.5 cm, 4.5 cm and 7 cm (Fig. 1)to simplify the filter paper assay in situ and make it more suitable. Twelve pipes were buried at the depths of 15, 30 and 45 cm nearby the 4 hydro-erosive plots to research
the suction phenomena at the first layers of the soil and its relation with precipitation during 7 months. The filter papers were taken to the field in selfsealing plastic bags (zip closure) to avoid humidity. They were taken out of the plastic bags and arranged in pipes. After 14 days, a new assay was realized with the exchange of the humid filter papers with a brand new set of filter papers. The humid papers were taken to the laboratory in sealed plastic bags within hermetically sealed glass containers. 2.2 Laboratory filter paper assay
Figure 1.
Pipe with 3 filter papers of different sizes.
Figure 2.
Setting of the 7 cm filter paper in the pipe.
Two undisturbed soil samples were taken at each depth of 30, 60 and 90 cm, nearby the 4 hydro-erosive plots (olericulture, pasture, forest recovery area and forest), summing 24 samples to perform the filter paper assay to obtain the retention curves. The 24 soil samples were saturated and gradually air dried to permit assays at regular diminishing humidity levels. Two types of filter paper assays were carried out: one in which the soil sample had direct contact with the filter paper to measure the matric suction; and another without a direct contact to measure total suction. The filter paper assay followed the ASTM D 529803 standards. The first assay measured the matric suction. Three filter papers were displayed in direct contact with each side of the soil sample. Just the middle filter paper was weighed. The 2 filter papers on the outside served to shield the middle filter paper from dust contamination. The 2 sets of filter papers with the soil sample together with the sampler ring were wrapped 3 times around in plastic protection film and twice in aluminum foil. The filter papers were handled with surgical gloves and tweezers to avoid grease contamination by the operators hands. The soil samples sealed and identified were put in a humid chamber at 25C for a period of 21 to 28 days to occur equalization of the filter paper with the suction of the soil sample, without influence of changes in air temperature and humidity. After the period of equalization, the samples were taken from the chamber. The middle filter paper was rapidly retrieved from the sample and put in a self-sealed plastic bag to avoid any loss of humidity. The filter paper and the sealed plastic bag were weighed in a balance with 0.1 mg of precision (Mettler H10T). When necessary, the dust in the filter paper was brushed off. Afterwards, the filter paper was taken out of the self-sealed plastic bag and dried in a stove at a temperature of 105C for 1 day. The stove-dried filter paper was retrieved from the stove and immediately put in the sealed plastic bag to avoid gain of humidity. The filter paper and the plastic bag were weighed together. The plastic bag was weighed separately to obtain the tare weight. The humidity of the filter paper
714
was calculated by the relation between the weight of the water and the weight of the dry filter paper. The second type of filter paper assay to obtain total suction was conducted without direct contact with the soil sample by means of a 1.5 mm thick perforated plastic spherical plate. The perforated plastic sphere plate permitted exchange of humidity between the soil sample and the filter paper. 2.3 Retention curve determination assay
The Fredlund Soil Water Characteristic Curve Device is based upon axis translation. It is a simple method which enables the determination of suction values up to 1500 kPa. Successive positive air pressures were applied to obtain 9 (suction versus volumetric water content) values. 2.4 Electron microscopic scanning of filter paper
The structure of the filter paper was analyzed using a scanning electron microscope model JSM6460 LV JEOL with a tungsten filament, a voltage acceleration of 20 kV and secondary electrons. Four samples of filter paper were analyzed in pairs, in which one pair belonged to the same batch and the other pair was of different batches. The images of the filter paper samples were processed to obtain the proportions of the larger fibers. 3 3.1 RESULTS AND DISCUSSIONS Results of the filter paper in situ assay
Figure 3. Graphic suction versus time versus precipitation of the soils of the 4 plots at a depth of 15 cm.
The results of the filter paper in situ assays are presented in Figures 35. 3.1.1 Discussions of the filter paper in situ assay results The results were well correlated with precipitation during the 7 months of the assays. The analysis of variance of the filter paper in situ assays suction results gave a F value of 2.18 which is less than the F 0.05 value of 3.09. The null hypothesis could not be rejected, and it can be affirmed in a significant level of 0.05 that there is no influence in the size of the filter paper in filter paper in situ suction measurement (Table 1). The forest soils had the lowest suction values, probably due to its higher water retention capacity related to the higher organic matter content. The soils of the olericulture hydro-erosive plot had often the lower suction values compared to the soils of the pasture and forest recovery area plots. This fact was due to the constant irrigation in the olericulture area, which maintained the soil humid during the cultivation period. This can be noticed by the lower suction values of the olericulture soils
Figure 4. Graphic suction versus time versus precipitation of the soils of the 4 plots at a depth of 30 cm.
715
Figure 6. (30 cm).
Characteristic curveolericulture plot soil
Figure 5. Graphic suction versus time versus precipitation of the soils of the 4 plots at a depth of 45 cm. Table 1. Analysis of variance of the filter paper in situ assays results. General Factorial Residual 104 2 102 Dispersion 272,640,808 11,165,413 261,475,394 Variance 2,621,546 5,582,706 2,563,484 F F0.05
2.18 3.09
in relation to the pasture and forest recovery area in the irrigation periods (18/8, 28/10 and 22/12). Even though the soils of the pasture and forest recovery area are more clayey and have a greater water retention capacity, they presented a quicker variation of suction responses to the rainfall events. This was due to the higher permeability of these soils. The suction responses in the olericulture soils in behalf of the rainfall events were slower than in the soils of the other plots due to its lower permeability. This can be noticed in the suction data of 9/2 in Figures 4 and 5. The pasture and forest recovery area soils had similar suction behaviors mainly because of their similar textures, permeability values and soil class. 3.2 Results of the filter paper assayretention curves
Figure 7. (30 cm).
Characteristic curvepasture plot soil
The retention curves were obtained by the Van Genutchen (1980) equation using the filter paper assays values (Figs. 613).
Figure 8. Characteristic curveforest recovery plot soil (30 cm).
716
Figure 9. Characteristic curveforest plot soil (30 cm).
Figure 11. Characteristic curvespasture plot.
Figure 10.
Characteristic curvesolericulture plot.
Figure 12. plot.
Characteristic curvesforest recovery area
Discussions of the results of the filter paper assaysretention curves The filter paper method was efficient in obtaining the retention curves for the soils in the plots of the study area. This can be stated by the semblance of the retention curves of the soil sample from the olericulture plot obtained by the filter paper assay and the axis translation method (Fredlund Soil Water Characteristic Curve Device). It is possible to obtain a soil retention curve with 5 data from filter paper assays provided that these points (suction versus volumetric water content) be uniformly distributed (Fig. 9). The use of different batches of filter paper didnt mischaracterize the assays. There was no need to calibrate each batch of filter paper to obtain the
3.2.1
retention curves. The filter paper calibration curve proposed by Chandler et al. (1992) was satisfactory, and the data (suction versus volumetric water content) were consistent. The characteristic curve of the olericulture soil sample at 30 cm has a lower suction value than the other retention curves at 60 and 90 cm (Fig. 10). This behavior may be due to a higher concentration of calcium carbonate at the depth of 30 cm, which is used to decrease the soils acidity. Calcium carbonate stimulates soil aggregation and therefore increases porosity that gives better hydraulic conductivity. The retention curve of the pasture soil sample at 30 cm has a higher suction value than the other curves at 60 and 90 cm (Fig. 11). Probably this occurred because of the soil compaction by
717
Figure 15. Figure 13. Characteristic curvesforest plot.
Sample B of filter paper.
Figure 16.
Sample C of filter paper.
Figure 14.
Sample A of filter paper.
pasture cattle grazing, diminishing porosity size and distribution. In general, the smaller the pore dimensions, the greater the water-retention capacity will be, along with higher values of suction. The evapotranspiration of the dense forest vegetation extracts humidity from the soil, which results in an increase in soil suction, promoting aggregate stability and giving greater resistance to erosion. Chandler (1992) stated that root water uptake promotes a desaturation zone where suction rises and pore-water pressure diminishes, and consequently the effective normal stress increases. 3.3 Results and discussions of the electron microscope scanning and image processing of filter paper
Figure 17.
Sample D of filter paper.
The images of the electron microscope scanning are in Figures 1417, and the results of their processing are in Table 2.
The image processing permitted to perceive that either filter papers from the same batch or from different batches have distinctions in fiber composition (Table 1). It was noted in the filter paper samples that their characteristics are singular and specific regardless of being of the same batch or not.
718
Table 2. paper. Sample
Results of the image processing of filter N of Pixels Observed 101.418 149.153 47.735 88.832 149.121 60.289 % of Larger Fibers 8,3 12,1 3,8 7,2 12,1 4,9
A B Difference (AB) (same batch) C D Difference (CD) (different batches)
A higher concentration of calcium carbonate in a soil layer can improve soil aggregation which increases porosity that gives better hydraulic conductivity. The soil compaction by pasture cattle grazing, diminishes porosity size and distribution, which gives higher suction for air-entry values that leads to the advancement of the air/water interface (contractile membrane). REFERENCES
ASTM Standards. D529803, Standard Test method for Measurement of Soil Potential (Suction) Using Filter Paper, West Conshohocken, USA, 2003. Chandler, R.J., Crilly, M.S. & Montgomery-Smith, G. 1992. A Low Cost Method of Assessing Clay Desiccation for Low-Rise Buildings. Proceedings of the Institute of Civil Engineering, n. 2, pp. 8289. Fredlund, D.G. & Rahardjo, H. 1993, Soil Mechanics for Unsaturated Soils, New York, USA, John Willey e Sons, Inc. Fredlund, D.G. & Xing, A. 1994. Equation for SoilWater Characteristic Curve. Canadian Geotechnical Journal. n. 37, pp. 521532. Mahler, C.F. & Oliveira, L.C.D. 1997. Determinao da Suco Total in situ de Solo Poroso de So Paulo Atravs do Mtodo do Papel Filtro, In: 30 Simpsio Brasileiro de Solos No Saturados, pp. 551556, Rio de Janeiro, RJ (in Portuguese). Marinho, F.A.M. & Oliveira, O.M. 2006. The Filter Paper Method Revisited, Geotechnical Testing Journal, v. 29, n. 3 (Fev), pp. 250258. Van Genuchten, M.Th. 1980. A Closed-form Equation for Predicting the Hydraulic Conductivity of Unsaturated Soils, Soil Science Society of American Journal, v. 44, pp. 892898.
CONCLUSIONS
There was no statistically significant difference in the results of suction measured in situ by the 3 different sizes of filter paper. The use of a filter paper with a diameter of 3.5 cm makes the filter paper assay more simple and suitable. Even though it is not precise, the filter paper assay in situ is simple and economical, and gives a good idea of the total soil suction. The results of the filter paper assay in situ were well correlated with precipitation during the 7 months of the experiment. The characteristic curve of a soil can be obtained by only 5 filter paper assays as long as the volumetric water content is well distributed. The use of different batches of filter paper didnt mischaracterize the assays. There was no need to calibrate each batch of filter paper to obtain the retention curves. The filter paper calibration curve proposed by Chandler et al. (1992) was satisfactory. The filter paper method was efficient in obtaining the retention curves for the soils in the plots of the study area.
719
The challenges of performing laboratory controlled cone penetration tests in unsaturated soils
M. Pournaghiazar, A.R. Russell & N. Khalili
Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering, University of New South Wales, Sydney, Australia
ABSTRACT: A suction and stress controlled calibration chamber has been developed and is used to conduct a series of cone penetration tests in unsaturated soils. The challenges associated with the use of the new equipment are discussed in detail including the preparation of samples with repeatable properties, and application of uniform suction throughout samples. Results of select cone penetration tests and the post test analyses of the results imply the success of the testing system. Interpretation of the test data is presented highlighting the effect of suction on penetration resistance. 1 INTRODUCTION are discussed and a method for preparation of uniform unsaturated samples of repeatable properties is presented. Results of select cone penetration tests are presented along with data interpretation, highlighting the effect of suction on penetration resistance and initial effective stress. 2 THE CALIBRATION CHAMBER
Unsaturated soils are wide spread throughout the world and need to be dealt with in many engineering problems including foundations, fills and embankment dams, pavements and airfields as well as natural and made slopes. The stress-strain behavior of unsaturated soils is complex and influenced by many factors including soil type, structure, porescale interactions and suction arising from surface tension effects across air-water interfaces within the pore space (Loret & Khalili 2000, 2002). The characterization of unsaturated soils requires expensive and time consuming site investigation, undisturbed sampling and laboratory testing to obtain even the most basic information on the suction-dependant strength and soil stiffness properties. Performing the cone penetration test (CPT) in unsaturated soils, on the other hand, may enable less costly and more rapid soil characterization in situ. Calibration chambers have been used in the past to develop correlations under laboratory controlled conditions for saturated or dry soils (e.g. Holden 1971, Bellotti et al. 1982, Huang et al. 1988). More recently a calibration chamber for performing CPTs in unsaturated soils was designed and built at the University of Oklahoma (Miller et al. 2002). In that chamber specimens were prepared by static compaction inside a specimen former, although difficulties were encountered due to a low former rigidity (Tan 2005). This paper presents details of a new suction and stress controlled calibration chamber for conducting cone penetration tests in unsaturated soils. Difficulties encountered during sample formation
The new calibration chamber accommodates a cylindrical soil specimen that is 460 mm in diameter and 800 mm in height. The chamber is designed to allow combinations of axially symmetric lateral and vertical pressures to be applied independently to the external walls of the specimen. Lateral stresses are applied by water pressure acting on a rubber membrane enclosing the soil specimen. Vertical stress is applied by a hydraulic loading ram pushing on the chamber piston connected to the base of the specimen. A cross section of the chamber is shown in Figure 1 (for a detailed description of the chamber segments, see Pournaghiazar et al. 2009). Suction within the specimen is controlled independently using the axis translation technique. In an unsaturated test air pressure is applied at the top of the chamber via fittings on the top cap and holes cut through the top plate. Pore water pressure is applied through eight high air entry value porous disks embedded in the specimen bottom plate. A pore air pressure (ua) that is larger than pore water pressure (uw) induces suction (s ua uw) in the soil specimen. High air entry value porous disks embedded in the bottom plate are shown in Figure 2.
721
Figure 1.
Calibration chamber cross section.
the chamber is assembled and a confining cell pressure is applied. The quarters of the former are then be pulled away from the specimen to allow a constant stress boundary condition. A cross section of moveable former is shown in Figure 1. The calibration chamber is fitted with a number of control and measurement units including a cell pressure supply system, a loading ram for controlling the vertical stresses acting on the specimen, and separate pore air and pore water pressure systems for controlling suction. The cell pressure is supplied using an air water interface cylinder and is measured using a analogue pressure gauge. Cell water volume change is determined by recording the water level in the air water interface cylinder and a glass burette volume change unit. The chamber has two separate lines to supply pore water to the base of a specimen. One line is connected to a set of perforated copper tubes at the bottom plate (Fig. 2) and is used to flood the specimen with water. The other set is connected to the ceramic disks embedded in the base plate and controls pore water pressure within an unsaturated specimen. Pore water volume changes are measured using separate volume change units comprising glass burettes and water oil interfaces. Pore air pressure supplied to the top of the specimen is controlled and measured using a pressure regulator and pressure gauge. An electric micro pump is used to operate the loading ram. A linear varying displacement transducer (LVDT) mounted below the piston captures the vertical displacement of the ram, which is equal to the vertical deformation of the specimen. 3 TESTING PROCEDURE
Figure 2. High air entry value porous disks embedded in the bottom plate and copper tubing for saturation of specimen.
A novel specimen formation system is introduced to enable creation of specimens of repeatable properties from a variety of soil types. The system consists of four cylinder quarters built into the chamber. Two handles are attached to each cylinder quarter near their top and bottom which enable them to be moved manually toward and away from the center of the chamber. During specimen preparation, the former quarters are pushed together and locked into position to form a rigid cylindrical mould. Vertical compaction of soil can be performed within the mould without experiencing bulging or distortion. After specimen preparation
The procedure for performing a CPT in a dry, saturated or unsaturated sand specimen is outlined in this section. The specimen formation method is presented and other less successful specimen formation procedures are discussed. 3.1 Formation of sand specimens
Dry sand specimens are prepared by the pluvial deposition technique. Sand is placed in a large hopper positioned above the calibration chamber. The sand is then allowed to flow through the base of the hopper through an opening. The flow rate is controlled by adjusting the size of the opening using a movable flap. The sand is passed through a diffuser consisting of two sieves aligned and fastened to a pulley system before it drops into the chamber. A uniform flow rate of sand through
722
the diffuser and a uniform drop height (distance between diffuser bottom and placed sand surface) enables preparation of a homogenous specimen. The system is designed so that the diffuser is moved upward during the pluviation process using the pulley system to maintain a constant drop height. Specimen dry density depends on the drop height and flow rate (Rad & Tumay 1987). A 50 mm thick gravel layer is placed on top of the specimen to enable even spread of air/water pressure. The top plate is then placed in position and the rubber membrane is clamped to the top plate. The chamber top cap is fastened to the top plate and chamber shell. A small confining pressure is applied to the specimen and the moveable sides of the former are pulled away from the specimen towards the chamber shell. The cell pressure and axial pressure are then increased to the desired values and the specimen is allowed to consolidate. The volume change of the sample is measured during this process to determine when consolidation is complete. Specimen saturation is achieved by passing de-aired distilled water through the perforated copper tubes located at the bottom of the specimen while applying a 10 kPa vacuum to the top of the specimen. Unsaturated specimens are formed by first saturating the specimens and then letting the moisture content reduce to a target suction. Initially, water is allowed to drain freely under its self weight by opening the valves connected to the copper tubes at the base. Once the drainage of water is completed, the copper tubing is sealed, and pore air pressure and pore water pressure are applied via the top and through the high air entry ceramic discs at the base, respectively. The pore water volume change is monitored during this process until moisture equilibrium is reached within the specimen. 3.2 Cone penetration
Figure 3. chamber.
Penetrometer mounted on the top of the
high air entry disks. In trial tests, it was observed that connecting laboratory air pressure directly to the top cap fittings resulted in local drying of the specimen surface, which can extend as far as 0.2 m below the top. This was caused by pore water evaporating into the dry air. Therefore, to prevent localized drying, the laboratory air is passed through a water bath cylinder before entering the chamber as shown in Figure 3. 4 TEST RESULTS
Cone penetration tests are conducted using a miniature electrical cone, model ELC2, manufactured by A.P van den Berg. The cone is 16 mm in diameter and the cone tip area and friction sleeve area are 2 cm2 and 30 cm2, respectively. The cone is pushed by a HYSON single cylinder static cone penetrometer. Figure 3 shows the penetrometer mounted on the top of the chamber. A hollow bush cylinder creates a seal around the cone and the centre hole of the chamber top cap as shown in Figure 3. The cone is pushed into the soil specimen with an insertion rate of 2 cm/s. As mentioned previously, for unsaturated testing, matric suction is applied to the specimens by increasing the air pressure connected to the chamber top cap while maintaining constant pore water pressure applied at the specimen base through
Results of CPTs conducted in saturated and unsaturated Sydney sand are shown in Figure 4. Sydney sand, a predominantly quartz sand, is classified as SP with a D50 of 0.3 mm. The relative densities of the specimens were about 61%. The saturated specimen (S1) was subjected to an isotropic effective stress of 100 kPa (lateral, vertical and pore water pressures of 350 kPa, 350 kPa and 250 kPa, respectively). The first unsaturated specimen (U1) was subjected to an isotropic net confining stress of 100 kPa and a suction of 50 kPa. The second unsaturated specimen (U2) was subjected to an isotropic net confining stress of 100 kPa and a suction of 200 kPa. The most useful portion of the test records is between depths of 0.3 m and 0.55 m where the qc
723
layer at the top of the specimen. The target suction was then applied by increasing the pore air pressure at the top of the specimen and maintaining a constant water pressure at the base of the specimen. It was expected that this procedure would result in a uniform distribution of moisture throughout the sample; however, post equilibrium analysis of the specimen indicated that the procedure resulted in a preferential flow of water within the specimen leading to zones of high saturation separated by zones of relatively dry sand. The other sample preparation method tested involved moist tamping of equal layers of loosely deposited sand. The sand in each layer had a moisture content related to the target suction. This method resulted in a relatively uniform distribution of moisture within the sample. However, CPT results indicated non-uniform densities within each layer due to the compaction procedure adopted. 6 INTERPRETATION OF RESULTS
Figure 4. CPT results in saturated and unsaturated sand specimens prepared by the pluviation deposition method (Dr 61%).
values are approximately constant and free of the influences of the top and bottom boundaries. The values of the cone tip resistance recorded between about 0.1 m and 0.25 m are slightly higher, caused by interaction between the top of the specimens and the top plate. From about 0.55 m the cone tip resistance values show a modest increase due to interaction with the rigid base boundary so tests are terminated around depths of 0.2 m from the bottom of the specimens. Between 0.3 m and 0.55 m depths the cone tip resistance averages about 10.7 MPa, 11.7 MPa and 13.6 MPa for S1 (saturated), U1 (unsaturated, s 50 kPa) and U2 (unsaturated, s 200 kPa) specimens, respectively. The suction noticeably increases the penetration resistance and clearly shows the importance of considering suction in interpreting CPT results. 5 LESS SUCCESSFUL SPECIMEN FORMATION PROCEDURES
Suction causes the particle contact forces of an unsaturated soil to increase above their saturated or dry values. It is usually assumed that suction is an isotropic phenomenon and the potential macroscopic effects of this are an increase in the stiffness of the soil skeletal response and the mean effective stress p. Following the work of Bishop (1959), the mean effective stress is quantified as: p pn s (1)
where pn is the mean net stress or the mean total stress in excess of pore air pressure ua; s is the difference between the pore-air and the pore-water pressures (ua uw), also referred to as matric suction; and is the effective stress parameter and has a value of 1 for saturated soils and 0 for dry soils. One of the most appealing features of the effective stress approach in unsaturated soil mechanics is that many behavioral characteristics may be described using properties and relationships well established for saturated soils. For example, the shear strength at the critical state may be defined as: qcs Mcs pcs (2)
Two other methods of forming unsaturated sand specimens were tested but were found to be less successful than the method described above. In one test a dry sand specimen was formed using the pluviation method and then de-aired water was added to achieve a moisture content 2% higher than the target moisture content (obtained from the soil water characteristic curve for the target suction). The water was poured uniformly across the gravel
where Mcs is the usual function of the critical state friction angle and is a material constant relevant to both saturated and unsaturated conditions. Furthermore, the complex phenomena of suction hardening and volumetric collapse upon wetting can be explained using the effective stress (Loret & Khalili 2002).
724
Khalili et al. (2004) and Russell & Khalili (2006b) found that may be expressed in the nondimensional form: 1 0.55 s s for 1 25 se se 1 0.45 s s 25 for 25 s s e e s for 1 se
c 16.3 18.6 ln Dr exp 2.1Dr p0 0.7
(4)
Slightly different and more accurate relationships may be observed for small ranges of p0, but Equation 4 is more general and places no restriction on the ranges of p 0 values to which it applies. The advantage of having this type of proportionality is that the following equation also applies: p c2 02 c1 p01
0.7
(3)
(5)
where se is the suction value separating saturated from unsaturated states. This relationship was determined following consideration of laboratory strength and volume change data for a quartz sand containing no fines and values of s/se less than 70 as well as fine grained soils having values of s/se less than 12. A lack of experimental data has prevented it being tested for fine grained soils with s/se larger than 12 (e.g. see Khalili and Khabbaz 1998). In many engineering problems pn can be readily quantified as it is related to the externally applied stresses on a soil element and therefore overburden pressure. However, the s requires careful determination of se and the suction in the ground. This usually involves suction determination tests conducted on representative samples of the soil in the laboratory or measuring the moisture content in the ground and indirectly obtaining suction from the soil-water characteristic curve. These methods are time consuming and expensive. There is need for a more rapid estimation of s, for example using the CPT. Suction has been shown to have a distinct influence on cone penetration resistance and cavity expansion pressure (Russell & Khalili 2006a). For sands cone penetration resistance (or cavity expansion pressure) is proportional to initial mean effective stress in the ground. Cone penetration or cavity expansion occurs under drained conditions in saturated sands. In unsaturated sands, although perfect drainage is less likely as unsaturated sands have a much lower permeability than saturated sands, the cavity expansion analysis of Russell & Khalili (2006a) has shown that drained and undrained expansions give almost identical results unless the initial degree of saturation is slightly less than unity. The assumption that drained conditions prevail around a cone penetrometer in unsaturated sands is therefore applied here. Cavity expansion pressure may be related to initial mean effective stress p0 according to a power law (Russell 2004):
where c1 and c2 are the cavity expansion pressures determined for initial stress conditions p01 and p02 , respectively, when the initial density of the sand is the same for each. It follows that if a cavity is expanded in a sand when saturated or dry (such that p0 pn0 uw, uw is the pore water pressure and is zero when the sand is dry) and again in the same sand when unsaturated (such that p0 pn0 s), the ratio between the cavity limit pressures for unsaturated and saturated states is:
c s c 0
p s n0 pn 0 uw
0.7
(6)
There are similarities between cavity expansion results and CPT results. Cone resistance is also proportional to p0 according to a power law (for example Baldi et al. 1986; Houlsby & Hitchman 1988). A relationship similar to Equation 6 therefore also applies to cone penetration resistance: qc 2 p02 qc 1 p01
0.7
(7)
It follows that if a CPT is performed in a saturated sand and again in the same sand when unsaturated, the ratio between the total cone resistances qc in the unsaturated state qcs and the saturated or dry state qc0 is: qc s qc 0 u p s 1 w n 0 q c 0 pn 0 uw
0.7
s qc 0
(8)
Realizing that qc0 |uw| and |s|, Equation 8 may be simplified to: qc s qc 0 p s n0 pn 0 uw
0.7
(9)
725
Equation 9 may be used to convert a CPT profile in an unsaturated sand at a site to an equivalent profile in a saturated sand at the same site as long as the variation of s and with depth is known (assuming of course that the uw profile for saturated conditions is known). Alternatively, if CPT profiles at a particular site are available for saturated and unsaturated conditions then the s profile can be back-calculated for the unsaturated condition. Applying Equation 9 to the CPT results shown in Figure 4 enables values of s 14 kPa and s 41 kPa to be determined for U1 and U2 specimens, respectively. Alternatively, as suction values were known to be 50 kPa and 200 kPa immediately prior to testing, Equation 1 can be applied, along with se 7 kPa for the sand tested, to obtain s 17 kPa and s 30 kPa for U1 and U2 specimens, respectively. The two methods for evaluating s give broadly similar results. It seems as though the power law of Equation 9 is an appropriate and simple tool for back calculation of s, with errors being less than 11 kPa. 7 CONCLUSIONS
Initiative and Australian Research Council Discovery Project is gratefully acknowledged. REFERENCES
Baldi, G., Bellotti, R., Ghionna, V., Jamiolkowski, M. & Pasqualini, E. 1986. Interpretation of CPTs and CPTUs; Part 2: drained penetration of sands. In Proc. 4th Int. Geotech. Seminar, Singapore: 143156. Bellotti, R., Bizzi, G. & Ghionna, V. 1982. Design, construction and use of a calibration chamber. In Proceedings of the second European symposium on penetration testing, Amsterdam, 2427 May 1982. Vol. 2: 439446. Bishop, A.W. 1959. The principle of effective stress. Teknisk Ukeblad, 106(39): 859863. Holden, J. 1971. Research on performance of soil penetrometers. University of Florida Dept. of Civil Engineering Internal Report, CE-SM-711. Houlsby, G.T. & Hitchman, R. 1988. Calibration chamber tests of a cone penetrometer in sand. Geotechnique, 38(1): 3944. Huang, A.B., Holtz, R.D. & Chameau, J.L. 1988. A calibration chamber for cohesive soils. Geotechnical Testing Journal, 11(1): 3035. Khalili, N., Geiser, F. & Blight, G.E. 2004. Effective stress in unsaturated soils, a review with new evidence. Int. J. Geomech, 4(2): 115126. Khalili, N. and Khabbaz, M.H., 1998, A Unique relationship for for the determination of the shear strength of unsaturated soils, Geotechnique, 48(5) pp. 681687. Loret, B. & Khalili, N. 2000. A three-phase model for unsaturated soils. International Journal for Numerical and Analytical Methods in Geomechanics, 24: 893927. Loret, B. & Khalili, N. 2002. An effective stress elastic-plastic model for unsaturated porous media. Mech. Materials, 34: 97116. Miller, G.A., Muraleetharan, K.K., Tan, N.K. & Lauder, D.R. 2002. A calibration chamber for unsaturated soil testing. In Proceedings of the 3rd International Conference on Unsaturated Soils, UNSAT 2002, Vol. 2, Balkema, Lisse, pp. 453457. Pournaghiazar, M., Russell, A.R. & Khalili, N. 2009. A calibration chamber for unsaturated soils. In Proc. 4th Asia-Pacific Conference on Unsaturated Soils, 2325 November 2009, Newcastcle, Australia. Rad, N.S. & Tumay, M.T. 1987. Factors affecting sand specimen preparation by raining. Geotechnical Testing Journal, 10(1): 3137. Russell, A.R. 2004. Cavity expansion in unsaturated soils. PhD thesis, The University of New South Wales, Australia. Russell, A.R., & Khalili, N. 2006a. On the problem of cavity expansion in unsaturated soils. Computational Mechanics. 37(4): 311330. Russell, A.R., & Khalili, N. 2006b. A unified bounding surface plasticity model for unsaturated soils. Int. J. Num. Anal. Meth. Geomech. 30(3): 181212. Tan, N.K. 2005. Pressuremeter and cone penetrometer testing in a calibration chamber with unsaturated minco silt. PhD theses, school of civil engineering and environmental science, university of Oklahoma, USA.
It is shown that a uniform unsaturated sand specimen is best created by initially saturating a dry specimen, then allowing the pore water to drain from the specimen under its own weight, followed by the application of suction using the axis translation technique for moisture equilibrium. It is also shown that localized drying near the top of the specimen can be prevented by passing the pore air through a water bath prior to connecting it to the chamber. Results of cone penetration tests carried out in saturated and unsaturated sands show that suction noticeably increases average penetration resistance. Correlations developed for saturated (or dry) conditions cannot be used for interpretation of cone penetration test results conducted in unsaturated soils. New correlations considering suction need to be developed. Similar power law proportionalities between initial mean effective stress and cavity expansion pressure or cone penetration resistance exist. From these a simple equation was developed enabling the contribution of suction to the effective stress - s in an unsaturated soil to be estimated. Values of s determined were in broad agreement with those prevailing in controlled calibration chamber tests. ACKNOWLEDGMENTS Financial support provided through the UNSW Major Research Equipment and Infrastructure
726
Characterization of unsaturated zone susceptible to landslides in tropical volcanic residual soil with in-situ tests
P.P. Rahardjo
Parahyangan Catholic University Bandung, Indonesia
R. Karlinasari
Sultan Agung Islamic University Semarang, Indonesia
ABSTRACT: Tropical residual soil has unique characteristics due to the influence of weather. The intensity of the rainfall is much high, and landslide become a serious threat. The number of landslides in West Java is numerous, more than 100 events yearly. Landslide is produced if the unsaturated zone on the upper side of a slope becomes saturated, hence it is necessary to characterize the unsaturated zone phrone to landslide on a slope. Three location of tests were selected, i.e. Cijengkol Slope, Neglajaya Slope and Kalijati. A research of weathering zone on residual soil were conducted with Scan Electron Microscopy (SEM) and X-Ray Diffraction (XRD) methods to gain the physical and geo-chemical characteristic of slopes, then a suction measurement of Filter Paper Method were conducted. This resulted in a conclusion that the weathering zone is a major reason to define the depth of unsaturated zone. A zone called zone 4 (the term from Little,1969) is defined as a zone where weathering process change between disintegration of base rock to lateritation process. The profile above zone 4 is a zone where unsaturated zone highly influenced by weather and potential to slide. A series of in-situ tests (CPT-u and Dilatometer Test) and drilling for SPT and undisturbed samples and laboratory tests were conducted to confirm the characteristic of this zone. The in-situ test were proved to be effective to define and characterize this zone, and consequently proved to be successful to define the unsaturated zone for landslide analysis. Keywords: Unsaturated Zone, Tropical residual soil, weathering zone, SEM, XRD, Filter Paper, CPT-u, Dilatometer 1 INTRODUCTION The water table lay deep below ground surface, this means that soil is in unsaturated condition. Besides due to the weight of infiltrated water, landslide occurs as a result of decreasing the shear strength of soils when soils become saturated from unsaturated condition. As an initial effort to define and anticipate the landslide, it is important to first define the characteristic of the unsaturated zone. Weathering process from base rock to residual soils is the independent variable to define another residual soils characteristics. Therefore as a first step, it is necessary to define the mineralogy of the residual soils. Weathering process than become the main influence factor to define physical properties characteristics, shear strength, compressibility, insitu stress, and slope hidrology characteristics of residual soils. In dry condition, a slope of tropical volcanic residual soils has a high shear strength characteristics. The slopes commonly stand up with steep slope, the angle of slope is higher than its friction angle. This is commonly the effect of two factors,
Indonesian territory has a specific characteristic geology, it is llocated on tropical zone and influenced by tectonic or volcanic forces. This condition is responsible for the formation of soil characteristics, and categorized the soils as tropical volcanic residual soil. Residual soils is a formed by weathering process of rock, as a result of exposure to weather and still lay above its base rock. Tropical soil is a soil located between Tropic of Cancer (23.5N) in North and Tropic of Capricorn (23.5S) in South. Volcanic soils is a soils derived from volcanic activity, directed from volcanic ash or as a weathering product of volcanic base rock (magma). Tropical volcanic residual soil of West Java, has a weathering process with high temperature, high rain fall and good drainage condition. Landslide in West Java is numerous, more than 100 cases each year. Almost all landslides cases are induced by rainfall (rainfall induced lanslides).
727
the soil structure and the contribution of matrix suction to shear strength (Fredlund & Rahardjo, 1993; Melinda, 2004). How this two factors work on slope is a problem that need more study. Tuncer & Lohnes (1977) published a profile of mineralogy, physical properties and shear strength parameters of Hawaiian and Puerto Rico tropical volcanic residual soil. This profile is very interesting. It shows a different condition over weathering zone. Using the same method of profiling, the authors intent to study further, to have a depth knowledge and define the mechanism of the two factors, the soil structure and the matrix suction controlling the shear strength of slope. 2 MINERALOGY OF TROPICAL VOLCANIC RESIDUAL SOILS
b. The Washing of Carbonat and Chlorite Process. In this stage a wash of carbonat and chlorite mineral is occuring. This washing process causes the soil to behave dispersively. The soil become lighter in colour, it is white or yellowish. c. Oxidation Process. In this stage, an oxidation process is occuring. Mineral from oxide and hydroxide group produced by an oxidation process of iron were found, such as Goethite (FeO(OH)) and Hematite (Fe2O3). The soil become reddish in present of iron mineral. In the influence of drying-wetting weather condition, the soil become flocculate, the mineral grain clump together into flocks, with a pararel structure. Based on these three specific process the profile of tropical volcanic residual soils, then can be divided into three different zone. The three specific zone are The Beginning of Weathering Zone, The Washing of Carbonat and Chlorit Zone, and Oxidation Zone. Based on Tuncer & Lohness,1977 profile, these three zones categorized into a four zones as follow: a. Zone 2, as the beginning of weathering zone b. Zona 3, as the washing of carbonat and chlorite zone c. Zona 4, as the beginning of oxidation zone d. Zona 5, as the end of oxidation zone
Based on mineralogy of tropical volcanic residual soils from Older Volcanic Products Formation (Qob) it is known, that the weathering process occurs in stages as follow (Figure 1): a. The Beginning of Weathering Process. In this stage a disintegration of base rock is occuring. In the layer of soils a feldspar minerals were found in a significant number. Feldspar mineral were found as a common mineral in base rock. In microscopic (from SEM), it is observed that the grain of the soil is still in some chrystal shapes and in a non pararel structure. Feldspar mineral found in this sample is Oligoclase, Na0.8Ca0.2Al1.2Si2.8O8, and Orthoclase KAlSi3O8. Feldspar mineral then broke by rainfalls water and composed a carbonat and chlorite mineral.
PHYSICAL PROPERTIES CHARACTERISTIC OF TROPICAL VOLCANIC RESIDUAL SOILS
Based on Tuncer & Lohnes diagram, the physical properties of tropical volcanic residual soils evolved as follows: Stage 1 and stage 2 are categorized as
Figure 1.
Weathering zones of tropical volcanic residual soils.
728
Figure 2. The variation of physical properties of Basalt rock in weathering to laterite (Tuncer and Lohnes, 1977).
Figure 3.
Specific gravity of minerals and Gs profile.
the beginning of weathering zone (where the base rock start to break), stage 3 is categorized as the washing of alkaline mineral zone where clay percentage start to increase, and stage 4 -5 categorized as the oxidation zone, where the sesquioxides (Fe2O3 dan Al2O3) are present. In the stage 3 the broken mineral, the alkalin, is a dispersive agent, it makes the soil grain become finer. As result the void ratio reduce and soils permeability decrease. In stage 4, the increase of sesquioxides (Fe2O3 dan Al2O3) means the increase on specific gravity because the iron mineral have a relatively high specific gravity. The clay particle cemented in flocculation process of sesquioxides. The void ratio increase because higher specific gravity means decrease on solid volume (Vs). In stage 5 a decrease on degree of saturation (Sr) occured, the soil become unsaturated. In unsaturated soil the index relation becomes as follow:
Figure 4.
Void diameter and void ratio profile.
Gs Se 1 e
w
(1)
In equation 1, if an increasing density and specific gravity occured, the decrease of degree of saturation will result on decrease of void ratio. Specific gravity and void ratio become a main parameters in controlling other soil parameters. In conclusion, these two parameters profile are describe as follow: Specific Gravity. A difference on type of minerals in each zone results on a difference of specific gravity, as an example the minerals resulted from oxidation process as Geothite and Hematite has a higher specific gravity of 3.33.5 and 4.95.3. So the profile of specific gravity has a unique shape as shown in Figure 3. Void ratio. The difference of minerals behaviour as dispersive in zone 3 and flocculative in zone 4 and 5 resulted on the difference on void
ratio (Figure 4). Dispersive behaviour resulted on a decreasing of void ratio, meanwhile flocculative behaviour resulted on increasing of macro void ratio and decreasing of micro void ratio. The void ratio increase in zone 2, decreasing in zone 3 and increasing back in zone 4. It achieves a maximum value at the border of zone 45. Void ratio decrease again on zone 5 because of the relatively high decreasing of degree of saturation as described in equation 1. The high magnitude of void ratio in zone 4 towards the border of zone 45 consequently cause higher permeability, this zone become a collecting place of water (water trap zone). The degree of saturation and water content in this zone become higher. 3.1 Matrix suction
To define the matrix suction profile, an instrumentation at Cijengkol Slope (a slope of tropical volcanic residual soil) was conducted with the instalation of 1 set Jetfill Tensiometer for monitoring matrix suction and 1 set ADR Thetaprobe for monitoring volumetric water content. The depth of instalation are 0.6 m, 1.2 m and 2.1 m. Meanwhile a Filter Paper method of measuring matrix suction was conducted at laboratory. Figure 5 shows all the
729
measuring matrix suction profile. In conclusion, the profile of matrix suction was described as follow: suctions on soils consist of two type of suction, matrix suction and osmotic suction. Osmotic suction is a suction related to a specific mineral on soils, as an example a soil with high carbonat and chlorite has a high value of osmotic suction. Matrix suction is a suction related only to soil structure, the value of void ratio between soil grain, micro or macro void. The magnitude of macro and micro soils are dependent to the type of mineral in soils, so the matrix suction is also influenced by the type of mineral in soils indirectly. In this research the matrix suction profile is defined as follow: on zone 3 matrix suction is low because the grain size
of soil categorized as silty or sandy clay. On zone 5 matrix suction relatively higher because the grain size of soil categorized as clay. The condition of matrix suction on site is as follows, matrix suction at zone 4 is the lowest because the water trap condition, on zone 5 matrix suction varied over degree of saturation, in general because of the high rainfall intensity, matrix suction fell below the soils air entry value. 4 IN-SITU CHARACTERISTICS OF TROPICAL VOLCANIC RESIDUAL SOILS
4.1 CPT-u test CPT-u tests were conducted at residual soil of Kalijati, West Java. This residual soils is a weathered result from a base rock of Qos formation, a sediment formation of Older Volcanic Product (Qob). This area is located at a lower elevation than Cijengkol slope, so the weathered zone is thicker. The soil profile is homogenus from silty clay to clay, without any other insertion of different type of soil (sand). A Sta 109 500 (Figure 6) zone 4 located at 10 to 16 m, it is marked by a value of FR from 2 to 4. Besides the value of FR, the value of u2 and the value of Bq could also be as an indicator of zone 4. The value of u2 at zone 4 is between 0.2 to 0.6 MPa,
Table 1. Values of FR, u2 and Bq from CPTu test at Kalijati residual slope. Sta Zone 4 (depth) FR u2 (MPa) Bq (%) 24 0.20.6 26 0.20.4 26 0.00.4 0.20.6 0.20.4 0.00.4
Figure 5. Slope.
Matrix suction profile of BH02 Cijengkol
Sta 109 500 106 m Sta 113 650 1318.5 m Sta 116 900 6.59.5 m
Figure 6.
Parameters profile from CPT-u test.
730
Figure 7.
Parameters profiles from dilatometer tests. Table 2. The values of maximum p0 and p1 from dilatometer tests. Location DMT01 BH-02 Neg DMT02 BH-03 Neg DMT03 BH-04 Neg DMT04 BH-05 Neg DMT01 BH-03 Cij Depth (m) p0 max p1 max (kPa) 12 10 13 12 10.6 1100 1000 1200 600 460 1600 1600 1500 1000 1600
while the value of Bq is around 0.2 to 0.6%. The same range of values were obtain from other bore holes as shown on table 1. In conclusion, zone 4 is a position where positive water pressure present (u2 present). Schneider,Peuchen & Mayne (2001) in their paper of Piezocone Profiling of Residual Soils, concluded that positive water pressure u2 profile indicate clearly the difference in stratigraphy of residual soils that related to weathering profile. As an example in Schneider profile of CPT-u test result on Singapore residual soils as follow: zone 4 is indicated at 1213 m depth, because u2 is at 0.2 0.4 MPa and FR is around 2 to 4%. These same conclusions were produced from this research, moreover this conclusion also confirm by the results of characteristics profile of mineral and physical properties. With knowledge to define zone 4 location, then it will be possible to predict the unsaturated behaviour of the soil profile, because the residual soil undergo the same weathering process. 4.2 Dilatometer test
Table 3. zone 4. Location
Parameters value from Dilatometer Test at p0 (kPa) 100500 400500 200500 200400 200400 ED (MPa) p1 (kPa) 7001000 7001000 5001000 4001000 500900 Ko 0.5 0.5 0.50.8 0.5 0.50.8
ID
KD
D01-2 Neg D02-3 Neg D03-4 Neg DMT04 BH-05 DMT01B-03 Cj Location
13.75 (sand) 24 0.61.6 (silt) 24 0.61.8 (silt) 24 0.61.0 (silt) 12 14 (sand) 24 vOCR 12 2 12.5 1 12.5 MDMT (MPa) 1025 1822 1030 213 1040
Insitu test with dilatometer were conducted at four location at Neglajaya slope and one location at Cijengkol slope. One of the results was shown in Figure 7 below. Based on corrected first reading (p0) and coreccted second reading (p1), an intermediate parameters of ID (material index), KD (horisontal stress index) and ED (dilatometer modulus) were obtained. From this intermediate parameters then derived by empirical formula another parameters: Ko, OCR and MDMT (Vertical Drained Constrained Modulus). Different from sediment soils, OCR value obtained from this dilatometer test was not correlated with the thickness of sediment in past, but more correlated to the presence or not a structures in soils. Because of the different perception, Fonseca et al (2008) proposed a change of term to vOCR or virtual OCR.
D01-2 Neg 1020 D02-3 Neg 1020 D03-4 Neg 1022 DMT04 BH-05 513 DMT01 B-03 Cj 1030
From the research, in general p1 profile is almost the same as p0 profile with a maximum value as follow: KD, Ko and vOCR profile in general is almost in the same shapes. The dilatometer parameters in each zone were discussed in order as follow. 4.2.1 Dilatometer parameters of zone 4 The value of KD at this zone shown a relatively the same value between 2 to 4. Based on characteristics of mineral and physical properties, it is concluded
731
Table 4. Location
Parameters from Dilatometer tests of zone 5. po(kPa) 0100 200800 200400 100350 100200 Ko 0.51.6 13 0.52 0.50.7 0.52 p1(kPa) 0300 5001200 250750 400650 250500 vOCR 28 420 110 11.5 110 ID KD ED (MPa) 010 525 1022 514 715
D01-2 Neg D02-3 Neg D03-4 Neg DMT04 BH-05 DMT01 B-03 Cj Location DMT01 BH-02 Neg DMT02 BH-03 Neg DMT03 BH-04 Neg DMT04 BH-05 Neg DMT01 BH-03 Cij
0.62.8 (Silt to sand) 28 0.34.0 (Clay to sand) 420 0.51.8 (Silt) 210 0.73.0 (Silt to sand) 1.52.5 0.84.0 (Silt to sand) 28 MDMT(MPa) 010 1080 830 2.515 830
that KD 4 as an indicator of lost of structures soils zone. The value of Ko is around 0.5, a common value for NC soil. The value of vOCR is between 1 to 2, indicated the soil as an NC to lightly Over Consolidated (OC). The ID parameter shown uncorelated to the result of index properties test at laboratory. ID parameter is a function of po, p1 and u0 as folow: ID ( p1 p0 ) ( p0 u0 ) (2)
and CPT-u test are among the best insitu tets to define the weathering zone. The insitu parameters from each different zone of weathering, have its specific range of values. The Index Material parameter ID from Dilatometer test is required to redefine because it is not yet included the effect of matrix suction on soils. REFERENCES
Blight, G.E., 1988, Keynote Paper: Construction in Tropical Soils, Proceedings Geomechanics in Tropical Soils Seminar, NTU-ISSMFE-SEAGS, Singapore. Fonseca, A.V, Coutinho, R.Q., 2008, Characterization of residual soils, Geotechnical and Geophysical Site Characterization Huang & Mayne (eds), Taylor & Francis Group, London. Fredlund, D.G. and Rahardjo, H.,1993, Soil Mechanics for Unsaturated Soils, New York: John Wiley & Sons, Inc. Karlinasari, R., 2009, Study of Characteristics of Tropical Volcanic Residual Soil From Older Volcanic Products Formation (Qob), West Java, PhD Dissertation at Parahyangan Catholic University, Bandung. Marchetti, Monaco, DMT Course Notes (2001), 77 p, Insitu 2001, Bali. Melinda, F., Rahardjo, H., Han, K.K., and Leong, E.C., 2004, Shear Strength of Compacted Soil under Infiltration Condition, Journal of Geotechnical and Geoenvironmental Engineering, Vol. 130, No. 8, August 1, 2004,ASCE. Schneider, J.A., Peuchen, J., Mayne, J., and McGillivray, A.V. (2001) Piezocone profiling of residual soils, Proceedings, International Conference on In Situ Measurements of Soil Properties and Case Histories, Bali, Indonesia, May 2124, 2001, pp. 593598, GEC, Bandung.
It is shown that the ID value is higher than the ID value of the index properties type of soils, which is clay and silt. This is possible because of the unsaturated condition. If the water pressure become negative (in unsaturated soil), therefore the ID will be lower and in suitable range of value, of clay and silt type of soils. 4.2.2 Dilatometer parameters of zone 5 The parameters from dilatometer test of zone 5 is shown in the table 4 below. 5 CONCLUSION
Zone 4 of tropical residual soils, where water trap condition occured, becomes an indicator zone to divide the tropical volcanic residual soil into weathering zones. With knowledge to define the zone 4 location, then it will be possible to predict the unsaturated behaviour of the soil profile, because the residual soil undergo the same weathering process. This research proved the effectiveness of insitu tests to define the zone 4 location. The Dilatometer
732
Analysis of a drawdown test displaying the use of transparent soil in unsaturated flow applications
G. Siemens & S. Peters
GeoEngineering Centre at Queens-RMC, Department of Civil Engineering, Royal Military College, Ontario, Canada
W.A. Take
GeoEngineering Centre at Queens-RMC, Department of Civil Engineering, Queens University, Ontario, Canada
ABSTRACT: Seasonal fluctuations within the unsaturated zone result from drying and wetting processes. Recently several column and two-dimensional apparatuses have been developed to study infiltration and drainage behaviour in near-surface settings. Traditional laboratory testing of unsaturated soil systems have been limited by their inability to visually observe air-fluid front migration and relied solely on instrumentation. At times the instrumentation gives confusing results that may be due to fingering or other mechanisms. To overcome this challenge a transparent soil was formed by matching the refractive indexes of a granular material and a fluid. At 100% saturation the soil-fluid combination is transparent. The transparent soil was used with the aid of digital photography to map the progression of the air-fluid interface during a drawdown test. In this paper, the transparent soil and test apparatus is described and the results from a drawdown soil are presented to display the use of transparent soil in unsaturated hydraulic applications. 1 INTRODUCTION men preparation techniques or other mechanisms occurring within the soil mass. To overcome this challenge a transparent soil was formed by matching the refractive indexes of a granular material and a fluid. As seen in Figure 1, at 100% saturation the soil-fluid combination is transparent. At degrees of saturation less than 100%, the air-fluid interfaces as well as the soil grains become visible. The transparent soil has been used with the aid of digital photography to map the progression of the air-fluid interface during drawdown and infiltration tests under a variety of configurations
Movements of moisture within the unsaturated zone result from a variety of natural and manmade processes at surface as well as at depth. The effect of moisture movement within the unsaturated zone plays a key role in performance of engineering applications founded in this zone as well as natural features such as slopes. Conventional modeling of moisture movements employs averaging constitutive models using Darcys law assumptions. However, with ever expanding computing power, representation of flow behaviour using networks of capillary tubes connected at nodes has gained further interest. (Siemens et al. 2007). Calibration and validation of these types of models would gain significantly from physical testing methods in which the pore structure can be viewed directly. Recently several column and two-dimensional apparatuses have been developed to study infiltration and drainage behaviour in near-surface settings (Ho 2000, Bathurst et al. 2007, Yang et al. 2004, Nahlawi et al. 2007). Traditional laboratory testing of unsaturated soil systems have relied on instrumentation for interpretation of moisture movements. Unfortunately the air-water interface cannot be viewed directly. At times the instrumentation has given confusing readings that may be due to fingering, embedded issues due to speci-
Figure 1. Transparent soil.
733
and boundary conditions. Transparent soil placed in a one-dimensional (1-D) column model constructed of Perspex allows visual interpretation of the developing air-fluid interfaces throughout the profile. In this paper results from a drawdown test are presented to demonstrate the use of the transparent soil for near surface moisture movements. In order to identify the air-fluid interface throughout the test, an algorithm was developed to track the position of the change from transparent to opaque soil. Results show the benefit of using transparent soil to accurately identify the air-fluid interface during drawdown. In addition, dramatic fingering mechanisms can be observed visually from within the column and measured quantitatively using the tracking algorithm. Results are analyzed to determine the conductivity of the soil during drawdown using a falling head test methodology. 2 TRANSPARENT SOIL
photograph is taken, the larger pores have drained first and will be followed by draining of the smaller sizes of the pore structure. The combination of soil and fluid used in the test reported in this paper is a fused quartz and a mineral oil. The fused quartz is classified as a uniform sand with the particle size distribution described in Table 1. It was placed at a dry density of 1130 kg/m3 and a corresponding porosity of 0.49 and void ratio of 1.0. The fluid was formed by mixing two white mineral oils to form a refractive index of 1.458 to match the fused quartz. The resulting mixture has a fluid density of 820 kg/m3. 3 TEST APPARATUS AND METHODOLOGY
Transparent soil is formed by combining a granular material with a fluid of the same refractive index. The advantage with using transparent soil is that the air-fluid interface can be observed visually and flow behaviour can be compared directed with instrumentation placed within the soil. Figure 1 is a close-up view of the column during a drainage test. The air front is progressing downward through the transparent soil. Below the air front, the soil is transparent and the black background is visible. Within the saturated transparent soil even very small bubbles and slight imperfections can be observed. Although they are expected to play a small role in the drawdown test, these non-uniformities are only visible because the saturated soil is transparent. Above the air interface the soil particles and air bubbles are visible and appear white on the photograph giving a striking difference to the saturated soil. In advance of the air front fingering mechanisms can be observed. As the fluid drains, the air is drawn downward and follows the path of least resistance. At the time the
Table 1. Index properties of transparent soil. Value 0.73 mm 1.1 mm 1.8 mm 2.4 1.0
The column test apparatus (Peters et al. 2009) developed to apply the use of transparent soil is shown schematically in Figure 2. The column is constructed of 17.5 mm thick Perspex. The length of soil that can be tested is 1300 mm tall with cross-sectional dimensions of 45 mm 45 mm (2025 mm2). Along the length of the column are instrumentation ports for pressure transducers to define the porewater pressure regime along the length of the column. Outflow from the bottom of the column was measured using a digital scale. Along the front of the column three digital single lens reflex (SLR) cameras and a high speed camera
Parameter Particle size distribution D10 D30 D60 Coefficient of uniformity Coefficient of curvature
Notes: D10, D30 and D60 particle diameter corresponding to 10, 30 and 60% by mass of finer particles, respectively.
Figure 2. Transparent soil column apparatus schematic.
734
are located to visually record the location of the air-fluid interface. Digital photographs are taken simultaneously every six seconds during the test with the SLR cameras. The focal area of adjacent cameras is set so that they overlap and the readings from adjacent cameras can be compared. In addition the digital readout from the scale faces the lower camera so that outflow is also recorded digitally. The high speed camera is placed to record specific areas of interest in individual tests. This combination of cameras allows tracking of the air-fluid interface along the complete length of the column. The column apparatus is configured to allow a wide range of potential test conditions, specimen preparation and soil saturation techniques. The boundary conditions that can be applied to the soil are constant head (positive or a vacuum) or constant flow at the surface through use of the overflow or a Mariotte bottle respectively. At the base, constant head or no flow conditions can be applied using the overflow or by closing the valve below the column. The configuration of the column apparatus also allows evaluation of the effect of specimen preparation techniques on measured behavior. In this particular test wet pluviation is implemented for specimen preparation. The column is filled with oil and the fused quartz is pluviated from the top of the column in 114 g increments to form layers approximately 50 mm thick. A total of 26 layers are required to fill the column to 1300 mm. Seed particles were placed at the top of each lift boundary to denote their height and potential displacement during tests. This preparation technique results in an initially clear medium to display the benefits of using transparent soil. Following specimen preparation, the test protocol includes completion of a constant head test and a falling head test. This allows measurement of the conductivity of the soil. Finally a drainage test is completed where the fluid is drawn down to the constant head boundary at the base (Figure 2). 4 AIR-FLUID INTERFACE INTERPRETATION
This section describes the process used for interpretation of the air-fluid interface in transparent soil during unsaturated flow. Figure 3a and 3b are close-up photographs of the column taken 12 seconds apart during the drawdown test over the elevation of 10601185 mm. Through the saturated portion of the column, the black background is visible while at the unsaturated portion the soil particles and air bubbles are observed. Within the saturated portion a black seed particle is noted. In Figure 3a, the air-fluid interface is relatively flat with some small fingering mechanisms
apparent. Twelve seconds later in Figure 3b, the air front has moved downwards and dramatic fingering is noticeable ahead of the main air-fluid interface. At the upper lift boundary shown in Figure 3b pockets of pore fluid are also seen. The mechanisms displayed in this photograph are a function of the specimen preparation procedure. Wet pluviation of the soil particles results in the larger particles falling faster than the smaller particles. Therefore within each lift there is inherent heterogeneity with grain-size decreasing vertically through the lift profile. At the upper lift boundary, the pocket of pore fluid is due to fluid being held temporarily by capillary forces in the relatively finer grained material. At the lower lift boundary the air-fluid interface is temporarily held up as it encounters the relatively finer-grained soil at the top of the next lift. The air fingering is the first sign that this small capillary break has been breached. These mechanisms are only visible through the use of transparent soil. In order to interpret the location of the air-fluid interface along the length of the column, digital photographs are taken simultaneously every six seconds with three digital SLR cameras (Figure 2). The dramatic difference between the black background visible through the clear saturated soil and the white unsaturated soil is exploited in order to define the interface. In terms of pixel intensity the range from black to white is 0255. At each elevation along the length of the column, the average pixel intensity is calculated and defined as IR. The mechanisms illustrated visually in the photographs are now quantified. Figure 3c displays the average pixel intensity, IR, of the photographs shown in Figure 3a and 3b (over the elevation 10601185 mm). Looking at Figure 3c, at 96 s the pixel intensity is 65 below 1164 mm and then increases to the top of the graph. The jump in pixel intensity is the change in shade from black to white at the air-fluid interface. At 108 s, pixel intensity is still 60 from 10601084 mm and then a small increase in pixel intensity to 100 is observed over the range of 10841102 mm. Finally the pixel intensity increases to approximately 195 for the rest of the plot except for a dip at 1150 mm. The air-fluid interface is clearly the dramatic change in pixel intensity. The fingering mechanism shown in Figure 3b is also visible on the graph as the jagged pixel intensity over from 10841102 mm elevation. Finally the pocket of pore fluid caught in the finer-grained soil at the top of the lift is also quantified as the dip in pixel intensity at 1150 mm elevation. In order to compare pixel intensity taken from adjacent cameras and to account for the change in light quality along the length of the column normalization of pixel intensity (IR) is required. As photographs are taken farther away from the light
735
Pore fluid pocket
Lift boundary seed particle
Lift boundary
Air fingering Time = 96 s Black intensity and location reference White intensity reference Time = 108 s Instrumentation ports
Figure 3. Transition from transparent to non-transparent soil, illustrating a) advancing air front approaching lift boundary (t 96 s), b) air fingering with pocket of fluid trapped at interface of lift boundary (t 108 s), c) measured pixel intensity over section of column shown in a-b with measured variation in black and white intensity values, and d) calculated normalized pixel intensity.
source in the ceiling the soil appears darker. This effect is increased in the overlapping camera areas since one camera is viewing the same area from below compared with the view from above from the adjacent camera. Normalization of pixel intensity is completed by using black and white references on the patches attached to the front face of the column (Figure 3a and 3b). The identical patches provide reference values for black and white intensity to normalize to along the length of the column. They provide an independent measure of the change in light quality. Also plotted in Figure 3b is the black and white references (RB and RW). The black and white references are measured at each patch placed every
50 mm along the length of the column and linearly interpolated between patches. Normalized intensity, IN, is then calculated as IN I R RB R W RB (1)
with the terms defined above and plotted in Figure 3c. In this framework reference values RB and RW are equal to 0 and 1 respectively. Normalized intensity over the length of the photographs is plotted in Figure 3d. The mechanisms visible in the photographs and the raw pixel intensity plot are still visible. However, normalizing the pixel intensity
736
allows interpretation along the entire length of the column to be standardized. The criterion used to define the air front in this paper is IN 0.3. This value was selected as it normally falls on the sharp increase in normalized intensity during column tests. 5 TEST RESULTS
The test results displaying the use of transparent soil in a hydraulic application are described in this section. The test followed the procedure described in Section 3, which included wet pluviation, a constant head test, a falling head test and drawdown to the lower constant head boundary. Results are plotted in Figure 4a-c as air front elevation versus time, flow rate and conductivity. Prior to drawdown a constant head conductivity test was completed. The measured conductivity is plotted at the top of Figure 4c as 1.50 103 m/s. Then addition of fluid to the top of the column was discontinued and the fluid was allowed to drain through the column. The fluid height drops at a constant rate as plotted in Figure 4a and the outflow rate is shown in Figure 4b. The time period before the fluid reached the top of the soil provided an opportunity to measure the conductivity through this falling head test. The average interpreted value for conductivity was again measured as 1.52 103 m/s. Finally the air front was allowed to progress through the soil. From Figure 4a, the air front moves downward linearly with time until it asymptotically
approaches the constant head boundary at the base. Some slight jaggedness is observed in the air front data. This is interpreted as the air front speeding up and slowing down as it encounters the lift boundaries. As shown on Figure 4a, the air front does not actually reach the constant head boundary due to capillary action at that elevation as well as the decreasing gradient for flow as the drawdown test nears the end. The drawdown test was monitored for a total of 188 minutes with the main data of interest plotted in Figure 4a. Outflow from the bottom of the column (Figure 4b) shows a slightly decreasing flow rate is measured during the falling head test compared with the constant head. At the surface of the soil the flow rate is 2.4 g/s. Once the air front penetrates the top of the soil the flow rate is relatively constant until it asymptotically approaches zero as the air front approaches the constant head boundary. To interpret the conductivity of the soil during drawdown a falling head methodology was adopted. The equation for interpreting conductivity in a falling head test with the bottom boundary condition held constant (not increasing) is K aL h1 ln h2 At (2)
where K conductivity, a area of the burette, L length of the soil, A cross-sectional area of
Figure 4. a) Elevation of air front migration through soil column, b) resulting mass outflow rate of fluid exiting the soil column, and c) calculated falling head conductivity with air front migration.
737
the soil, t length of a time increment between head measurements and h1, h2 head loss across the specimen at the beginning and end of the time increment. During the falling head test, a and A correspond to the internal area of the column, L is 1300 mm, h1 and h2 are the difference between the fluid height and the bottom boundary and t corresponds to the time between photographs which is six seconds in these tests. Once the air front penetrates the surface of the soil, interpretation of conductivity requires some assumptions to be made. First, L is the height of the soil below the air-fluid interface at the end of the time increment. Second the a has to be corrected for the area of pore space that is drained during the time increment. The exact distance the air front moves downward is known from Figure 4a. Flow volume during the time increment is measured with the digital scale and recorded with the lower camera. As in ASTM D 5084, flow in the column is assumed to be equal to flow out of the column. Therefore dividing the volume of outflow by the distance the air front has moved gives the area of the pore space that has been drained. This is set equal to a in Equation 2. Initially the porosity of the soil was used for a in Equation 2, however, this was found to overestimate the pore space drained and the calculated conductivity. This overestimation is a result of not all the pore fluid draining behind the air front as shown in Figure 3b. Finally the pressure (or suction) head at the air-fluid interface must be determined in order to properly calculate the head loss over the length of the soil. Through a small parametric analysis the conductivity for this particular soil was found to be insensitive to changes in the suction head at the air-fluid interface. To generate Figure 4c a positive pressure head of 5 mm was used. For finer-grained soils this insensitivity to pressure at the air front is likely not to be the case. The calculated conductivity of the soil using the assumptions explained in the previous paragraph is plotted in Figure 4c. The average value of 1.44 103 m/s agrees well with the previously measured values from the constant head and falling head tests. 6 CONCLUSIONS
interface, observation of fingering mechanisms, allows the possibility to display potential issues from installed instrumentation and provides data for modeling results using averaging techniques or capillary tube models. In this paper, a test methodology and column apparatus is described for testing unsaturated flow behavior using transparent soil. The column configuration allows a wide range of test boundary conditions, specimen preparation techniques. The parameters of interest including pore pressure and outflow can also be measured and used to interpret the air front movement. The air-fluid interface is tracked during a drawdown test through the use of digital photography. The test results were analyzed and a falling head methodology was used to interpret conductivity during drawdown which was consistent with earlier constant head and falling head tests. REFERENCES
ASTM D 5084, Standard Test Method for Measurement of Hydraulic Conductivity of Saturated Porous Materials Using a Flexible Wall Permeameter, American Society for Testing and Materials, West Conshohocken, Pennsylvania, USA. Bathurst, R.J., Ho, A.F., and Siemens, G. 2007. A Column Apparatus for Investigation of 1D UnsaturatedSaturated Response of Sand-Geotextile Systems. Geotechnical Testing Journal, 30(6): 19. Ho, A. 2000. Experimental and Numerical Investigation of Infiltration Ponding in One-Dimensional SandGeotextile Columns. M.Sc thesis, Department of Civil Engineering, Queens University, Kingston, ON. Nahlawi, H., Bouazza, A., and Kodikara, J. 2007. Surface Water Infiltration in a 1-Dimensional Soil-Geotextile Column. Common Ground Proceedings, 10th Australia New Zealand Conference on Geomechanics, Brisbane: 368373. Peters, S.B., Siemens, G.A., and Take, W.A. 2009. A transparent medium to provide visual interpretation of saturated/unsaturated hydraulic behviour. 62nd Canadian Geotechnical Conference, Halifax, NS, 2123 September 2009, 5966. Siemens, G.A., Blatz, J.A., and Ruth, D.G. 2007. A capillary tube model for transient two-phase flow through swelling soil. Canadian Geotechnical Journal. 44(12): 14461461. Yang, H., Rahardjo, H., Wibawa, B., and Leong, E.C. 2004. A Soil Column Apparatus for Laboratory Infiltration Study. Geotechnical Testing Journal, 27(4): 19.
Use of transparent soil in unsaturated flow applications allows direct interpretation of the air-fluid
738
Calibration of Smart Irrigation Sensor (SIS-UMS) for the blanket layer soil from old railway lines
V.N. Trinh, A.-M. Tang, Y.-J. Cui, J.C. Dupla & J. Canou
Ecole des PontsParisTech, France
N. Calon, A. Robinet & L. Lambert

Direction of EngineeringSNCF, France
O. Schoen
Railway network of FranceRFF, France
ABSTRACT: The sensor SIS (Smart Irrigation SensorUMS) was selected to be used in monitoring the soil matric suction changes in an old railway line. The sensor was calibrated using a soil taken from the blanket layer. This layer was created mainly by the interpenetration of the subgrade soil into the ballast layer under the trains actions. The sensitivity of its mechanical behaviour to water moisture changes is a key point when evaluating the maintenance of the railway line involved. Obviously, monitoring the suction helps for this evaluation. In the present work, a SIS sensor was installed beside a T8 tensiometerUMS and a high-capacity tensiometer in a plastic box of 240-mm in diameter and 240-mm in height. The readings of the high capacity tensiometer were recorded automatically in voltage and were converted in kPa using the calibration curve carried out in laboratory. The values of the SIS sensor and the T8 tensiometer were measured by a box INFIELD 7 with manual reading based on the calibration of the supplier. The comparison of the matric suction between three sensors shows that the measurements of two tensiometers are similar. Values of approximate of 2.4 times lower were measured by the SIS sensor. 1 INTRODUCTION laboratory testing with hydraulic conductivity measurement or field monitoring with suction or water content measurement. Obviously, laboratory testing allows a good test conditions in terms of controls of environmental conditions and physic measurements. However, the soil samples tested are usually prepared by compaction for a simple reason of commodity. The microstructure of these compacted samples can be very different from that of the soil in field conditions. This justifies the field testing. In this study, to monitor the soil moisture changes in field conditions, an old railway platform has been instrumented using the SIS sensor (Smart Irrigation SensorUMS). A specific calibration work was carried out in laboratory prior the installation in situ. For many years, standard tensiometers have been used to measure soil matrix suction for the purpose of irrigation scheduling in the field of agriculture or other research studies (Campbell & Mulla 1990). The standard tensiometers are certainly accurate but they have a limited range for suction measurement (80 kPa) and require specific preparation and regular maintenance. On the other hand, the Watermark, an electrical resistance based
A blanket layer is referred to as a layer in track foundation, between the ballast and the subgrade (UIC 719R 2003). For the new lines in France for high speed trains, the blanket layer was designed and constructed according to the technical specification of French railway company (SNCF, IN3278 2006). In the structure of the old railway platform in France, a blanket layer was created mainly by the interpenetration of subgrade soil into the ballast layer under train circulation. Its formation and composition differs from that for the blanket layer of new lines. Note that the blanket layer in this case is more commonly called sub-ballast. One of the main functions of this layer is ensuring drainage of the sub-structure. Water can come from rainfalls through the ballast layer; water can also come from the subgrade by underground water table growth. Moreover, under cyclic loading by train circulation, excess pore pressure can be built-up in both the blanket layer and the subgrade, leading to water flow in the blanket layer during the water pressure dissipation. The evaluation of the drainage function can be achieved by either
739
sensor, has been used since 1978. The Watermark operates following the same principle as other electrical resistance-based sensors. Water conditions inside the Watermark sensor change with the corresponding variations of the water conditions in the surrounded soil. These changes are reflected by the difference in electrical resistance between two electrodes embedded in the sensor. This resistance difference decreases with increasing soil water content (Shock 2004, Irmak et al. 2006). Calibration is required to allow conversion of electrical resistance, measured with the Watermark sensor, to soil suction. Thomson and Armstrong (1987) proposed a calibration equation for Watermark 200 giving the relationship between the soil suction, the soil temperature and the electrical resistance within the range from 10 to 100 kPa. Spaans and Baker (1992) calibrated Watermark 200 sensor using tensiometers and thermocouples for soil suction and temperature measurement respectively, and observed that the different sensors provided similar results in terms of temperature, but different results in terms of suction. There is another model of Watermark sensor 200SS by Irrometer Co. Riverside, CA, USA. It involves a granular matrix sensor that is surrounded by a synthetic membrane for protection against deterioration (Shock 2004, Irmak et al. 2006). Shock et al. (1998) compared the calibration equations for Watermark 200 and 200SS in laboratory conditions within a suction range from 10 to 75 kPa, and they observed that the calibration curve is linear for model 200 but curvilinear for model 200SS. The SIS-UMS is a Watermark sensor of 200SS type with DC-voltage output signal. In the present work, the SIS sensor is calibrated using a highcapacity tensiometer and a T8 tensiometer (UMS) that is equipped with a temperature sensor also. The calibration was performed on the fraction passing through 2 mm of a blanket layer soil from an old railway line. 2 2.1 SOIL STUDIED, DESCRIPTION OF THE DEVICE AND METHODOLOGY Soil studied
Table 1. Geotechnical properties of the fraction passing through 2 mm sieve. wL (%) 35.6 wP (%) 25.5 IP (%) 10.1 VBS (g) 1.97 s 2 m (Mg/m3) (%) 2.67 20 80 m (%) 40
The water retention curve was determined on compacted samples at a dry density of 1.81 Mg/m3 using filter paper method. The results obtained are presented in Figure 1 where volumetric water content and degree of saturation are plotted versus soil suction. The experimental data were also fitted with the function proposed by Van Genuchten (1980):
1 r ( s r ) 1 ( hw )1
(1)
where , s, r are volumetric water content, saturated volumetric water content and residual volumetric water content, respectively; hw is soil suction (in cm); and are fitting parameters. The following values were obtained for the studied soil: s 0.32; r 0.012; 0.23 and 0.0007. The air entry value is not clear from the experimental data and it can only be noted from the fitting curve using equation 1: the soil remained in saturated state for suctions lower than 40 kPa. 2.2 Description of the device
The soil studied was taken from the blanket layer of an old railway platform at Snissiat (near Lyon, South-East of France). This layer was created mainly by the interpenetration of the subgrade soil into the ballast layer under trains actions. The soil studied contains 25% of fine grains smaller than 2 mm. The fine fraction is composed of subgrade soil (clay), sands, degraded ballast and polluted materials (hydrocarbon, product wear, etc.). The geotechnical properties of the soil studied are presented in Table 1.
The SIS-UMS is a combination of the modified Watermark granular matrix soil moisture sensor of type 200SS and a special amplifier which provides a fully linearized and temperature compensated DC-voltage output signal. The amplifier uses a new technique allowing an accurate measurement. The linear voltage signal allows reducing the efforts devoted to the electronic measuring equipments and cable connections. The temperature compensation system directly considers the soil temperature. Thus, the problem related to the temperature difference between logger and sensor is avoided (SIS-UMS 2006). The technical specifications of the sensor are shown in Table 2. The T8 tensiometerUMS measures directly the matrix suction in the soil with a satisfactory accuracy. Indeed, this tensiometer works in the range from 100 kPa (water pressure) to 85 kPa (suction/ soil water tension) with an accuracy of 0.5 kPa (T8-UMS, 2008). The values of the SIS sensor and the T8 tensiometer were measured by a box of manual reading INFIELD 7 (INFIELD7-UMS 2008) according to the calibration of the supplier.
740

6 0 20 40 60 80 100 120 5 10
SIS
HCT
T8 tensiometer
10
Suction (Log h(cm))
10
10
Suction (kPa)
Experimental data Van Genuchten (1980) model

0 0.1 0.2 0.3
10
Soil
Volumetric water content (cm3/cm3)
0.4
10
Figure 2.
Schematic view of the experimental setup.
Figure 1. Water retention curve of blanket layer soil sieved at 2 mm. Table 2. Technical specifications of SIS-UMS. 50 2000 mV corresponds to 5.0 kPa 200.0 kPa 0.2 kPa (5.0 70.0 kPa) 6 20 VDC 2.5 mA Watermark 200SS Diameter: 25 mm Length: 116 mm
SIS sensor, while in the second test, all the three sensors were installed inside the box. 3 RESULTS
Soil water tension Measuring range Resolution Power supply Current consumption Gypsum block Dimensions
The high capacity tensiometer (HCT) is composed of a high air entry value ceramic stone of 1.5 MPa that is stuck on the tensiometer body in stainless with epoxy glue. The strain gauge that are stuck on the diaphragm allowed the monitoring of water pressure based on a preliminary calibration (Cui et al. 2008). The readings of the HCT were recorded automatically in voltage and were converted to kPa using the calibration curve. 2.3 Methodology
In order to calibrate the SIS sensor, the blanket soil layer sieved at 2 mm was first wetted to have an initial water content of 12%. Then, the soil was compacted manually in layers in a plastic box having a diameter of 240 mm and a height of 240 mm. The SIS sensor and T8 tensiometer were placed at the same depth of 100 mm. The HCT was installed at 20 mm depth. A schematic view of the experimental setup is shown in Figure 2. Two calibration tests were performed in this study. The first test used the HCT to calibrate the
Results of the first test are shown in Figure 3 where the soil suction is plotted versus the elapsed time. Note equally that the soil suction measured by the HCT was recorded automatically by computer every 30 minutes while that of SIS sensor was recorded manually. After the installation of two sensors, the soil surface was let open for drying. The results show that the suction measured by the HCT increased quickly from 0 to 100 kPa after 35 h while that of the SIS sensor increased from 0 to 25 kPa only (Step 1). The increase of suction can be attributed to both the initial equilibrium phase after the installation of sensors and the drying from the soil surface. In order to better discriminate the two processes, from t 35160 h (Step 2), the soil surface was covered by a thin plastic film. The results show that the suction measured by the sensors continued to increase but at a lower rate. At the end of Step 2 (t 160 h), the suction measured by the HCT was 175 kPa while that measured by the SIS sensor was 75 kPa. For the Step 3 (from t 160), a layer of wax was put on the plastic film. It was observed that soil suction continued to increase. Wax was then added several times (t 168, 192, 240, 264, 336 h) in order to improve the water tightness. The results show that the soil suction measured by the SIS sensor continued to increase at the same rate before and after adding the wax. As far as the HCT is concerned, the measurement was significantly affected by the adding of wax on the soil surface. Indeed, the
741
0 50 100 150 200 250 300 0 48 96 144
High capacity Tensiometer SIS sensor SIS sensor * 2.4
Step 1
Step 2 Step 3
192 240 288 336 384
The soil surface was finally covered by a layer of wax at t 168 h. That induced a decrease followed by an increase of suction measured by the HCT. At the end of the test, the suction measured by the HCT was 70 kPa while that measured by the T8 tensiometer was 70 kPa. The final value of suction measured by the SIS sensor was 30 kPa. As in the case of the first test, a correction coefficient of 2.4 was used to correct the calibration of the SIS suction. The figure 3 shows that the corrected curve of the SIS sensor fits well with the suction measured by the T8 sensor. 4 DISCUSSION
Suction (kPa)
Time (h)
Figure 3.
First testCalibration of SIS by HCT.
suction measured by the HCT decreased abruptly after adding the wax and then increased slowly to reach the initial value after about one day. This phenomenon can be explained by the temperature effects on the HCT because the wax was heated up to 60 C before the application on the soil surface. The SIS sensor is temperature compensated and it was not affected by the temperature changes. In addition, the HCT was situated closer to the soil surface where the changes of temperature were more significant. The results obtained from the first test show a similar trend for the two sensors. Furthermore, the suction measured by the HCT was on the whole two times higher than that given by the SIS sensor. In Figure 3, the suction measured by the SIS sensor was then corrected by multiplying with a correction coefficient of 2.4. It can be observed that the corrected curve fits well with that of HCT. The results of the second test are shown in Figure 4. Firstly, only the SIS sensor was installed inside the box. The results show that the suction measured by this sensor increased from 0 to 20 kPa after 72 h. The tensiometer T8 was installed at t 72 h. This sensor allows equally the measurement of temperature. The results shows that the suction measured by the tensiometer increased immediately and reached 30 kPa few time after its installation. A small effect of temperature changes can be observed though the fluctuation of the measured suction. The HCT was installed at t 144 h. The results show that the suction measured by this sensor increased quickly after the installation, from 0 to 65 kPa; after a few hours. The value of 65 kPa was equally the suction measured by the T8 tensiometer.
The Figure 4 shows that the suction measured by two tensiometers is similar and stable at 70 kPa after a wax layer covered the soil surface. The correction coefficient of 2.4 of SIS sensor found in the first test (Figure 3) was confirmed in the second test with both the T8 sensor and HCT. Thomson et al. (2006) presented another relationship between the soil suction from Watermark 200SS sensor and that fromtensiometers sensor, determined in field condition in a sandy loam soil (equation 2): y = 1.35x 9.91
22 21 T (C) 20 19 18 Time (h)
0 0 48 96 144 192 240
0 48 96 144 192 240
(2)
20
Suction (kPa)
40
60
80
HCT SIS sensor SIS sensor * 2.4 T8 tensiometer_SMP T8 tensiometer_T
Figure 4. Second testCalibration of SIS by HCT and T8 tensiometer.
742
where y is the soil suction measured by tensiometers; x is the soil suction measured by Watermark 200SS sensor using the calibration equation by Shock et al. (1998). Equation 2 also shows a linear relationship but it is obviously unsuitable for blanket layer soil sieved at 2 mm, showing that the correlation relationship is probably soil nature dependent. It is also important to mention that the suppliers calibration of SIS sensor is different from the calibration equation proposed by Shock et al. (1998) and used in this work. Spaans and Baker (1992) also reported a poor correlation of the calibration for two different soils. Irmak and Haman (2001) compared the suctions measured by Watermark 200SS sensor using five different calibration equations found in the literature with the suctions measured by mercury manometer-tensiometer for two different soils, and they concluded that the performance of the calibration equations was poor. The temperature effect was observed in two tests. The suction measured decreased when the temperature increased. This result is in good agreement with the existing works (Spaans and Baker 1992, Shock et al.1996). In the first test, the SIS sensor was calibrated beside the HCT, until a suction of 115 kPa (the suction given by the HCT was 275 kPa). The final suction measured by SIS sensor, in the second test, was lower than in the first one (30 kPa) (70 kPa by T8 sensor and HCT). This is because the T8 tensiometer works only in the range from 100 kPa (water pressure) to 85 kPa (suction/soil water tension). The linear relationship between the responses of SIS sensor and two tensiometer types with a correction coefficient of 2.4 was validated in two cases. This means that the SIS sensor may be used to measure higher soil suctions than standard tensiometers. Several calibration equations were developed to calculate the response of Watermark 200SS sensor as a function of the electrical resistance and temperature (Shock et al. 1996, 1998, Thompson et al. 2006). However, in the works performed on SIS sensorUMS, the linear relationship between suction and the output voltage signal was not provided. It is therefore not possible to compare the correlation relationship obtained in this study with that from other works. 5 CONCLUSION
height) containing the soil passing through 2 mm sieve. Two tests were carried out and the following conclusions can be drawn. The value of suction obtained by the SIS sensor should be multiplied by a correction coefficient of 2.4 in order to have the same values obtained by the two other sensors. The sensibility of the sensors with the variations of temperature was examined: the tensiometers are much more sensitive than the SIS sensor mainly because the latter is temperature compensated. Few hours were need for the three sensors to reach the equilibrium state after their installation. It is difficult to compare the results of this study with the calibration equations found in the literature because all the calibration equations from the literature were achieved with Watermark sensor resistance while the response of the SIS sensor UMS obtained in this work was calculated from a linear relationship between soil suction and output voltage signal. Further works have been planned to monitor the soil suction in an old railway platform using the SIS sensors. REFERENCES
Campbell, G. & Mulla, D. 1990. Measurement of soil water content and potential. In: Stewart BA, Nielsen DR (eds) Irrigation of agricultural corps. American society of agronomy. Madison, USA, 127142. Cui, Y., Tang, A., Mantho, A. & De Laure, E. 2008. Monitoring Field Soil Suction Using a Miniature Tensiometer. Geotechnical Testing Journal 31(1), 95100. INFIELD7-UMS. 2008. INFIELDHandheld readout device. User Manual. 56 p. Irmak, S. & Haman, D.Z. 2001. Performance of the Watermark Granular Matrix Sensor in Sandy Soils. Applied engineering in agriculture 17(6), 787795. Irmak, S., Payero, J.O. & Eisenhauer, D.E. 2006. Watermark Granular Matrix Sensor to Measure Soil Matric potential for Irrigation Management. The Board of Regents of the University of Nebraska on behalf of the University of NebraskaLincoln Extension. Shock, C.C. 2004. Granular matrix sensors. http://www. cropinfo.net/granular.htm, 139146. Shock, C.C., Barnum, J.M. & Seddigh, M. 1998. Calibration of Watermark Soil Moisture Sensors for Irrigation Management. Malheur Experiment Station, Oregon State Univ., Ontario, Oreg. 139146. Shock, C.C., Fibert, E. & Saunders, M. 1996. Malheur Experiment Station Annual Report. Special Report 964, Oregon State Univ., Ontario, Oreg. SIS-UMS. 2006. SISSmart Irrigation Sensor. User Manual., 14 p. Spaans, E.J.A. & Baker, J.M. 1992. Calibration of watermark soil moisture sensor for soil matric potential and temperature. Plant and Soil 143, 213217.
The SIS sensor was calibrated for measuring the matric suction of a blanket layer soil taken from an old railway platform. A T8 tensiometer and a high-capacity tensiometer were used for the calibration. The three sensors were installed inside a plastic box (240 mm in diameter and 240 mm in
743
T8-UMS. 2008. T8Long-term Monitoring Tensiometer. User Manual. 56 p. Thomson, R.B., Gallardo, M., Agera, T., Valdez, L.C. & Fernandez, M. 2006. Evaluation of the watermark sensor for use with drip irrigated vegetable crops. Irrig. Sci. 24, 185202. Thomson, S.J. & Armstrong, C.F. 1987. Calibration of watermark model 200 soil moisture sensor. Appl. Eng. Agric. 3, 186189.
Van-Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, 892898. UIC 719R. 2003. Earthworks and track bed for railway lines. International Union of Railways.
744
Wave velocity analysis of state parameter changes in multi-phase granular materials

F. Wuttke & M. Asslan T. Schanz
Bauhaus-Universitt Weimar, Chair and Laboratory of Soil Mechanics, Germany Ruhr-Universitt Bochum, Chair of Foundation Engineering, Soil and Rock Mechanics, Germany
ABSTRACT: The determination of wave velocities in soils captures an important role for initials parameter in constitutive modelling as well as in monitoring processes as state indicator nowadays. The wave velocity differs substantially with the porosity, surrounding pressure, degree of saturation and other parameters in soils. To determine the wave velocity in soils, different methods in situ and ex situ exist. For studying the behaviour of waves under different conditions, the laboratory methods are indispensable, but the identification of the correct velocity can be a large challenge particularly if the perturbations are small. We discusses existing methods in time, frequency and time-frequency space and applying of the coda wave interferometry to determine the wave velocity in dependence of small soil perturbations. During measurements, the application shows uncertainties and mistakes in identification of wave velocities. In comparison to conventional methods, like first arrival measurement, the advantages of extended methods as wavelet transforms and time-frequency distributions are shown. To study the dependency of wave velocities caused by small perturbations, as hydro-mechanical changes, the coda wave interferometry was used. The laboratory tests were done on different type of sands with well known hydro-mechanical parameter. In result of the tests, the coda wave interferometry emphasized its large potential for the timelapse monitoring of soils. 1 1.1 INTRODUCTION Geotechnical motivation and current state Boulanger 1998). Both of them can be difficult to interpret due to inherent dispersion characteristics in granular material and often insufficient signal coherence. A further, more advanced method was presented by Greening & Nash 2004. Based on a sweep signal, the phase difference method between the transmitter and receiver signal was successfully applied over a wide frequency range. However from SASW method, where the phase difference method is a basis, the large uncertainties and assignment problems in the low frequency range are known (Nazarian, Yuan & Baker 1994, Yuan & Nazarian 1993). 1.2 Physical background
The determination of initial states, like initial shear stiffness, is a basic part in the description of strain-dependent material behaviour of soils. For the determination of initial shear stiffness, different in- and ex-situ methods are available. The performance of laboratory tests permits an extensive study under controlled conditions. The study of the acoustical characteristics of granular material in laboratory experiments is technically realized since many years (e.g. Lee & Santamarina 2004, Santamarina & Fratta 2005). Most of the literature is related to different interpretation of time-of-flight measurements (travel time difference) as base of initial shear modulus assessment (Arulnathan & Boulanger 1998, Jovicic, Coop & Simic 1996, Ismail & Rammah 2005, Youn, Choo & Kim 2008). Due to several superimpose effects on the wave path the determination of wave velocity is a non-trivial task. During the laboratory tests, straight forward methods as peak or cross-over point differences were applied (Blewett, Blewett & Woodward 2000) as well as the more suitable application of cross-correlation methods (Arulnathan &
From the point of view of granular materials, it is essential to interpret the signals in more general context to determine the velocities or to monitor the material perturbations. This paper tries to consider microscopical influences during the propagation of a wave train in a granular material. From numerical simulations (Lomax 1999, Somfai et al. 2005) it was ascertain that the transmitted acoustic signal can separated into an initial coherent part and a subsequent randomly or noisy part (coda)
745
influenced by multiple scattering through the granular media. The soil mechanics literature on wave propagation is chiefly concerned with the measurement of macroscopical elastic modules under specific soil conditions. In particular, under multiphase conditions (Alramahi et al. 2008). After extraction of the wave velocity from the coherent part of the signal, as discussed, the remaining part is discarded as scattering and other effects. It can be postulated that a granular assembly can be considered as an effective medium for the transmission of a wave train, if the tests were done on sufficiently large scales and pressures and if the focus is on the initial coherent part of the response. After this wave front follows the noisy part, which is sensitive to small changes in the structure and packing details. Any measurements which are dominated by this noisy part will be influenced by the granular structure and its constituents, like density, grain size, shape, water content and others and not by an average effective parameter of the media. 2 2.1 ASSIGNMENT OF ABSOLUTE VELOCITY LEVEL Conventional methods
Many measurements concerning the determination of the right time-of-flight (TOF) were done in the past. Most of them tried to find the right point of the incoming wave train, whereas all analyses were related to first coherent part of the wave train and the determination of the effective modules in the material. The objective of research was mostly the investigation of different impulse functions, different impulse shapes, lengths, frequency content and others. Greening and Nash (2004) used a spectrum analyser to investigate sweep signals for further phase velocity calculations. The results show differences between the TOF velocity and the phase velocity. In general the phase velocity as a consistent velocity linked with the wave length is much more suitable to explain the complex wave behaviour in granular materials as the sharp wave front (TOF). Additionally, the phase velocity is beside the group velocity connected to the energy transport, hence more sensitive to characterize a dispersive material and accordingly gives more realistic wave velocities. 2.2 Time-scale and Time-frequency signal transforms
tion. Besides, the phase difference method, timefrequency and time-scale transform methods were used in this study to analyse the time-dependent signal. By using these methods the group velocity is calculated. The application of the easy shortterm Fourier transform in time domain (Bath 1974) or similar methods, as multiple filter methods (Dziewonski, Bloch & Landisman 1969), fails because the resolution of this method is quite poor. The time scaled signal transform methods like continuous wavelet transform (Santamarina & Fratta 2005, Gucunski & Shokouhi 2004, Mallat 1998) were successfully applied to other tasks. Beside this signal transform methods, in material science the time-frequency signal transform methods or time-frequency distributions (Hammond & White 1996, Mallat 1998, Niethammer & Jacobs 2001, Narasimhan & Nayak 2003) are often used to analyse the wave fields and dispersion characteristics. In own laboratory tests, the continuous wavelet transform (CWT) was required to keep a sufficient level of redundancy in the transformed signal. Due to this demand the use of the CWT during running laboratory tests requires a further special mathematical treatment to avoid a huge calculation time effort. To apply the CWT in laboratory tests the CWT (eq. 1) was transformed into the frequency domain. As suitable mother wavelet for vibration analyses, the Morlet wavelet was used. Ws,
1 s
f (t )
*t
d F( Ws, ) s
(1)
The given variables are f(t) time history, wavelet function, s scaling factor, shift variable, * nominates the complex conjugated term and F the Fourier transformed function. By using the rules of the discrete Fourier Synthesis, the decomposed Wavelet transformed signal in time and frequency is given by eq. 2, see Torrence 2005. Ws,
N 1 0
F f t F * k , 0 , a,b, t e j bt

k
(2)
In following, the sweep-signal excitation was used to find opportunities to overcome the problems in the phase difference and TOF methods and to determine wave velocities as well as the dispersion characteristics in multi-phase material. The used sweep-signal was a linear time-frequency func-
Different to the time-scale transform methods, the time-frequency transform methods are less analytical from the mathematical point of view and therefore more limited, but faster in use. The Wigner-Ville distribution is often used representative of Cohens class, which are the general class of the quadratic time-frequency formulations. The basis for the high resolution in time and frequency is the current autocorrelation of the time history itself, see eq. 3. W(t, )
t f f 2
*
t e jft d 2
(3)
746
the diffuse waves is the Faynman path summation (Lomax 1999) which describes stochastically arbitrary wave paths through a granular media and the genesis of a wave train as superimposed different wave fields. The analysis of this part is realised by using the normalized windowed cross correlation (CC) in eq. 4 R
( t,T )
(t s )
t T uunp (t )uper (t ts )dt

tT
tT
t T
u2 unp (t )dt
tT
t T
u2 per (t )dt
12
(4)
Figure 1. Time-frequency dependence of transmitted and received signal by using a sweep excitation and a time-scale or time-frequency transform method.
The entity of both the time-scale and time-frequency transform methods is the imaging of the dependency between time and frequency. Therefore, by comparison of the time-frequency dependence of a transmitted and a received signal, the group velocity can be determined, see figure 1. The absolute value of the group velocity is given by the dark lines in the surface plot of 3D timefrequency spectra and mathematically linked with the phase velocity of the wave train. The difference in the detected velocity between the first arrival or time-of-flight methods velocities is given due to the dispersion, attenuation and scattering characteristics in granular multi-phase materials. 3 3.1 ASSIGNMENT OF VELOCITY GRADIENTS Diffuse Waves & Coda wave interferometry
in which uunp represents the wave field in the unperturbed und uper in the perturbed system, T the window length and ts the time markers. The detected time delay in the shifted and cross correlated windows shows a more or less linear gradient with the total time t. In ratio to the time lag t the gradient of the velocity v/v is defined, see figure 2. The normalised CC gives additionally a trust value in the statistical sense. 4 4.1 EXPERIMENTAL STUDIES Experimental Setup
As explained before, the complete wave train consists of an initial coherent part followed by a noisy tail, the coda. The coda itself is compounded by different scattering effects, where the waves are interfered from different wave paths along the grain contacts (Snieder 2002, Snieder 2006). But due to this many different superimposed influences (accumulation of small changes), this part of the wave front is sensitive to any packing details and called diffuse waves. The base of the coda wave interferometry (CWI) is the determination of the velocity gradients between two stages of the system, before and after a small perturbation, as e.g. density change, change of suction or water content. This characteristics promises to find a sensitivity sensor for further monitoring studies. In combination of absolute velocity assessment methods with velocity gradient methods, the absolute velocity can be given by adding the gradient to an initial value in a sensitive manner. The theoretical background of
The control of the laboratory equipment takes place by LabView 8.2. To generate and record the seismic signals during laboratory tests, an extended function generator was implemented in Labview. The content of signal shapes covered harmonic, impulse and sweep signals whereas the change of polarization is possible for each signal. To record the transmitted and received signals at the same time was the most difficult point. As data logger a NI-USB 6251 with a maximal sample rate of 1 MHz was used. The bender elements were impinged by a voltage range of 10V. Because the input signal was
Figure 2. Definition of velocity gradient by the ratio v/v between the total time t and the time lag t of the shifted windows in the coda wave interferometry.
747
in the lower range of mV, the data logger was prefixed by a charge amplifier. The relevant frequency ranges for measurements are given in dependence on the sample size and grain size. The laboratory specimens were prepared in a triaxial cell with a height of 0.2 m and a diameter of 0.1 m whereas the distance between the elements is s 0.184 m. The analysed frequency range was between 3 kHz to 30 kHz. The tests were done on seven different sands as shown in figure 3 and glass beads with four different diameters (0.6, 1.0, 3.5 and 8 mm). During the laboratory tests, the confining pressure was increased in steps of 5 kPa from 100 kPa up to 400 kPa to detect small changes in porosity of the structure. The tests were done under dry and saturated conditions, whereas the study of partially saturated conditions with same material is not finished yet, due to the time effort to achieve the hydraulic equilibrium in the media. 4.2 Experimental analyses
A comparison of all material is given at the end. In the following paragraphs the results of the time-offlight (TOF), phase and group velocity as well as the velocity gradients are presented and discussed. For improvement of TOF picking, the transmitter signal was used twice in normal deviation and inverse polarised deviation, see figure 4. The signals shown in figure 4 can separated into the mentioned above two partsthe initial coherent part and the noisy, superimposed partthe coda. The second part of the given tests was subject of CWI to determine the velocity gradients. In opposite to the impulse excitation, the sweep signal is used to determine the phase and group velocity, see figure 5. Figure 6 shows the surface plot of figure 5 for the transformed transmitter and receiver signal in time-frequency range. From the time delay between both transformed signals, the group velocity was calculated. However, the determination of the phase or group velocity needs an excellent signal quality because
Three different materials are analysed in detail here, Sand 5 and 6 and glass beads of 3.5 mm diameter.
Figure 3.
Grain size distribution of test sands.
Figure 5. Used sweep signals to determine the phase and group velocity in the same sample as well as the time delay between the transmitter and receiver signal.
Figure 4. Typical impulse signals to determine TOF, here for dry sand100 kPa and 9 kHz.
Figure 6. Surface plot of the transformed transmitter and receiver signal.
748
these methods are more sensitive to noise and disturbances. The advantage of these complex and sensitive methods is directly linked to the wave length. Furthermore, the presented group velocity analysis gives a completion to the TOF and phase difference method. This analysis gives an extension of information and an option to overcome or to reduce the existing problems in velocity determination. Figure 7 shows the comparison between the phase, group and TOFvelocity in a dry and saturated sample. To validate the measured group velocity, the relationship between group and phase velocity was used to determine numerically the group velocity from phase velocity measurements. It is obvious that the numerical and measured group velocities are very close. The frequency trend is similar for both casesdry and saturated material. But the TOF values are higher than the phase or group velocity. With presence of a fluid, the phase velocity and TOF velocity is decreasing. In general, the velocity is decreasing from 10 up to 15 percent after sample saturation. In comparison to analysis of the initial coherent part, the analysis of the second part gives much more detailed information about small changes in the structure. As shown in figure 8, the evolution of the velocity gradient over the pressure steps is equivalent to the change of porosity. It is visible that the test sand no. 6 has a different behaviour in the evolution of the velocity gradient after the saturation. The reason was found after more detailed investigation of the void ratio, see figure 9. In comparison between the gradient of void ratio evolution, as derivation of the void ratio in figure 9, and the gradient of velocity, in figure 8, it is visible that there is a direct link between both gradients; see figure 10 and figure 11. The void ratio was determined by measurement of cell water change after each pressure step. This procedure gives the more or less rough void ratio curves in figure 9.
Figure 8. Evolution of the velocity gradient under increasing pressure steps.
Figure 9. pressure.
Evolution of the void ratio change with
By using this not-well-smoothed curve for derivation of the void ratio gradients, the spikes in figure 9 appear. To compare both gradients, a trend function for the void ratio gradient is to be used. Figures 10 and 11 show impressively the sensitivity of velocity gradients for small perturbations. By summation of all velocity gradients to an accumulated function a link to the accumulated volume change is given. Figure 11 shows a comparison between both accumulated functions. Whereby the velocity gradient is multiplied by a constant factor b (60.0 / 40.0) for comparison between the accumulation shapes. 4.3 Conclusions The objective of the presented paper was to analyse tools to assess granular material under field influences (pressure, fluid). To determine initial states in these media or to monitor small changes in a sensitive mannerdifferent information can be extracted from a wave train propagating through a porous media. The methods were validated at dry and saturated material and currently the partially
Figure 7. velocity.
Comparison between TOF, phase and group
749
Figure 10. Comparison between the velocity and the void ratio gradient of dry and saturated sand.
Figure 11. Comparison between the shape of the accumulated volume change and accumulated velocity gradient of the sands.
saturated material is under study. It was shown that the methods can give additional information about the dispersion characteristics as well as the accumulated changes due to small changes in porosity, interparticle pressures and others like water content. ACKNOWLEDGMENT The authors acknowledge the support of the R&D-Programme GEOTECHNOLOGIEN funded by the German Ministry of Education and Research (BMBF) and German Research Foundation (DFG), Grant 03G0636B. REFERENCES
Alramahi, B., Alshibli, K.A., Fratta, D., Trautwein, S. 2008. A Suction-control apparatus for the measurement of P- and S- Wave velocity in soils. Geotechnical Testing Journal. 31(1):1224. Arulnathan, R. & Boulanger, R. 1998. Analysis of bender elements. Geotechnical Testing Journal, 21(2): 120131. Bath, M. 1974, Spectral analysis in geophysics. Developments in geophysics 7, Elsevier Scientific Publishing.
Blewett, J., Blewett, I.J. & Woodward. 1999. Measurement of shear wave velocity using phase-sensitive detection techniques. Canadian Geotechnical Journal. 36:934939. Dziewonski, A., Bloch, S. & Landisman, M. 1969. A technique for the analysis of transient seismic signals. Bulletin of Seismological Society of America. 59(1): 427444. Farge, M. 1992, Wavelet transforms and their application to turbulence. Annual Review of Fluid Mechanics. 24, 395457. Greening, P. & Nash, D. 2004. Frequency domain determination of G0 using bender elements. Geotechnical Testing Journal 27(3): 288294. Gucunski, N. & Shokouhi, P. 2004. Detection and characterization of cavities under the airfield pavements by wavelet analysis of surface waves, in FAA Worldwide Airport Technology Transfer Conference, New Jersey, USA. Hammond, J. & White, P. 1996. The analysis of non-stationary signals using time-frequency methods. Journal of Sound and Vibration 190(3): 419447. Ismail, M. & Rammah, K. 2005. Shear-plate transducers as a possible alternative to bender elements for measuring gmax. Geotechnique. 55(5): 403407. Jovicic, V., Coop, M. & Simic, M. 1996. Objective criteria for determining gmax from bender element tests. Geotechnique. 46(2): 357362. Lee, J.-S. & Santamarina, J., C. 2004. Bender elements: perfocmance and signal interpretation. Journal of Geotechnical and Geoenvironmental Engineering. 131(9):10631070 Lomax, A. 1999. Path-summation waveforms. Geophysical Journal International. 138:702716 Mallat, S. 1998. A wavelet tour of signal processing, Academic Press. Narasimhan, S. & Nayak, M. B. 2003. Improved WignerVille Distribution performance by signal decomposition and modified group delay. Signal Processing. 83: 25232538. Nazarian, S., Yuan, D. & Baker, M. R. 1994. Automation of spectral analysis of surface wave method, in Ebelhar, Drevinch & Kutter, (eds), Dynamic Geotechnical Testing II, 88100. Niethammer, M. & Jacobs, L. J. 2001. Time-frequency representations of lamb waves. Journal of Acoustical Society of America 109(5): 18411847. Santamarina, J. & Fratta, D. 2005. Discrete Signals and Inverse Problems, John Wiley and Sons. Snieder, R. 2002. Coda Wave Interferometry and equilibrium of energy in elastic media. Physical Review. E 66(046615):18 Snieder, R. 2006. The theory of Coda Wave Interferometry. Pure and Applied Geophysics. 163:455473 Somfai, E., Roux, J.N., Snoeijer, J.H., Van Hecke, M. & Van Saarloos, W. 2005. Elastic wave propagation in confined granular systems. Physical review. E 72: 15393755. Torrence, C. & Compo, G. 1998, A practical guide to wavelet analysis. Bulletin of the American Meterological Society 79(1): 6178. Youn, J.-U., Choo, Y.-W. & Kim, D.-S. 2008. Measurment of small-strain shear modulus gmax of dry and saturated sands by bender element, resonant column and torsional shear tests. Canadian Geotechnical Journal. 45:14261438. Yuan, D. & Nazarian, S. 1993. Automated surface wave method: Inversion technique. Journal of Geotechnical Engineering. 119(7): 11121126.
750
A novel non-contacting laser displacement transducer technique for volume change measurements of unsaturated soils
X. Zhang, M. Mavroulidou & M.J. Gunn Z. Cabarkapa J. Sutton
Faculty of Engineering, Science and the Built Environment, London South Bank University, UK Geotechnical Consulting Group Ltd., UK GDS Instruments, Hampshire, UK
ABSTRACT: A novel non-contact total volume change measurement system was designed using an ultra high accuracy laser displacement sensor to measure the water level changes resulting from the volume change of the soil specimen. The suggested system consists of an open inner cell located inside a Bishop and Wesley cell, an external measurement cell connected to the open inner cell (a necessary arrangement due to the limited space in the Bishop and Wesley cell), and a Linearised Charge-Coupled Device (Li-CCD) laser displacement sensor fixed above the measurement cell. A number of calibrations were performed to account for the effects of various factors including water compressibility, temperature, cell pressure and system creep. Comparative measurements of volume changes of saturated samples during triaxial testing using the suggested technique and conventional techniques were then performed. These showed overall very good agreement between the methodologies and gave confidence in the suggested technique. This preliminary study therefore proved the feasibility of the technique and showed promise for the development of a fairly simple tool enabling accurate volume measurements during triaxial testing. Further studies and refinement of the tool are necessary for the adoption of the technique with confidence and its potential commercial exploitation. 1 INTRODUCTION sensors (Cabarkapa & Cuccovilo, 2006), digital image processing (Gachet et al, 2003) and recently, laser scanning techniques (Romero, 1999) and c) indirect measurements of the specimen volume deduced by measuring changes in the volume of the confining cell liquid. For this originally a cathetometer was used (Cui & Delage, 1996) and more recently, a differential pressure transducer (DPT) (Ng et al, 2002). As the measurement error increases with increasing confinement liquid volume, an inner cell or double-walled cell (Wheeler, 1988; Yin, 2003) is often used to minimise the liquid volume and negate the effects of cell creep under pressure. Whilst improved volume measurements have been achieved using liquid cell measurement in combination with a DPT, a number of features of such systems could be further improved: firstly, the pipelines attached to the DPT are complex and need four valves to control the water flow. This can result in corner points that are hard to de-air and this can affect the accuracy of the measurement. Moreover, the use of the system needs extreme care, as it is easy to miss the appropriate switching on and off of all valves. If the bilateral pipeline generates a big pressure difference, the transducer could
Volume change measurement when testing unsaturated soils is much more complex than that of saturated soils, due to the presence of the air phase. Consequently, the measurement techniques used for saturated soils are not directly applicable to unsaturated soil testing. Various methods have been suggested to measure volume changes in an unsaturated soil. These have varying degrees of accuracy and complexity and a number of limitations. Moreover, it can be difficult to achieve similar degrees of accuracy in the measurements using these techniques compared to saturated soil testing. Hence, there remains an outstanding requirement to develop a new volume change technique for unsaturated soils allowing for simpler and more accurate measurements. In general, techniques used for the measurement of unsaturated soil volume changes can be split into three broad categories, namely: a) direct measurements of air-volume and water-volume used to deduce volumetric changes of the soil specimen (Adams et al, 1996), b) direct measurements of the specimen volume using local Linear Variable Differential Transformers (LVDT) and Hall Effect
751
be damaged. Secondly, the water level change in the neck-shaped cell is about 1020 mm but the measurement range for DPTs that are usually employed is around 100 mm. This discrepancy between the measurement range of the transducer and the water volume change range will affect the measurement resolution and accuracy. The accuracy of the system is therefore 0.1% (readable to 0.1 mm), which is lower to that of an external LVDT transducer (this is readable to 0.01 mm). The water capacity in the reference tube and open inner cell is different and this can result in higher apparent pressure differences, due to the compression and thermal expansion of the water. Although it is impossible to eliminate completely errors due to pressure and temperature changes, these should be kept as small as possible. The following sections describe the development of a new volume change measurement technique for unsaturated soil testing, whose design was conceived as a further improvement to the indirect measurements of the cell liquid technique using a DPT, as it attempts to address some of the shortcomings of the DPT technique outlined above. 2 TESTING SYSTEM OUTLINE
Figure 1. Schematics of the volume change measurement system.
The key feature of the suggested volume measurement system is a high accuracy Linearised ChargeCoupled Device (Li-CCD) laser displacement sensor (model: Keyence LK-G 85) of an accuracy of 0.02%, a resolution of 0.0001 mm and a measuring range of 15 mm. The whole volume change measurement system consists of an open inner cell which is made of stainless steel placed inside a Bishop and Wesley cell and an external measurement cell (a necessary arrangement due to the limited space in the Bishop and Wesley cell) supported on a stand bar which is clamped onto the lower chamber beam to ensure it moves together with the inner open cell. A 1/8 inch diameter Legris tube is used to ensure hydraulic continuity between the water in the measurement cell and the water in the inner open cell. The laser sensor is fixed at the top of the measurement cell (see Figures 1 and 2). The laser sensor transmits a laser beam passing through a window (i.e. a 20 mm thick perspex disk, shown in Figure 2). The beam is reflected back to the sensor by a float, which lies on the surface of the water. This is used to monitor the water level change in the measurement cell which is then converted into specimen volume change, as explained in the next paragraph. The unsaturated soil triaxial system is completed by two GDS Instruments hydraulic pressure/volume controllers supplying axial stress and back water pressure respectively, one dual channel pneumatic controller used to supply pore air pressure and cell pressure, a data acquisition system
Figure 2. Detail of the volume change measurement system components.
and a series of transducers. The axis translation technique is adopted to control the suction. An example of how volume change measurements would be obtained assuming a typical compression triaxial test is as follows: during shearing the long top cap moves down following the change in height of the sample. (Displacement in the axial direction can be measured by an external displacement transducer). At the same time the sample typically deforms in the radial direction in a non-uniform way. This deformation in the radial direction causes the water in the measurement cell to rise (as it is hydraulically connected to the inner cell) and the resulting change in the level of water is captured by the laser sensor (see Figure 3). As a numerical example of this, assume a 20% strain in the axial direction and a 5% total volumetric
752
strain (a representative volumetric strain value for sample failure in a triaxial cell). If the triaxial soil sample were of a 38 mm diameter and 76 mm height, the original volume of the sample would be 86.19 cm3. A total volume strain of 5% would therefore be equal to a total volume change of the sample of 0.05*86.19 cm3 4.31 cm3. Upon 20% axial strain the change in the height of the sample would be 0.2*H 60.8 mm and this, without taking into account deformation in the radial direction, would correspond to a change in the volume of the cylindrical sample of V 0.2*A*H 17.24 cm3 (where A area of sample and H height of sample). This means that if the sample had dilated with a total volumetric strain of 5%, the change in water volume in the measurement cell would be 17.24 4.31 21.55 cm3; conversely, if the sample were contracting the water volume change in the measurement cell at a 5% total volumetric strain would be 17.24 4.31 12.93 cm3. Note that in both cases (dilation or contraction) the water volume changes in the measurement cell are positive corresponding to a rise in the height of the water level in the measurement cell (see Figure 4). Having explained the principle of volume change measurements, the reason for using a long and wide top cap of a diameter equal to that of the specimen instead of a small steel ram now becomes clear: if a
small steel ram were used as in a typical stress path cell, the volume changes corresponding to axial displacements/sample height changes would be too small for a satisfactory measurement resolution. 3 3.1 SYSTEM CALIBRATIONS The effect of the compressibility and thermal expansibility of water
Water volume change measurements can be affected by the compressibility of water under fluctuating temperature. In thermodynamics and fluid mechanics, compressibility is a measure of the relative volume change of a fluid or solid as a response to a pressure (or mean stress) change:

1 V V P
(1)
where V volume; and p pressure. The coefficient of thermal expansion is defined as:
1 V V P
(2)
Figure 3. Volume change in measurement cell during shearing.
The compressibility of water is 4.6 1010/Pa (Fine & Millero, 1979) and the volumetric coefficient of thermal expansion of water is 0.21 103/C at 20C. As an example, based on the above values, for a cell volume of 1000 cm3, and an initial temperature of 20C (at constant pressure) the change in volume for a 1C change in temperature (based on Equation 2) would be 210 mm3 and that for a pressure increment of 500 kPa would be 230 mm3 (based on Equation 1). The above compressibility value is the theoretical value if the water is completely de-aired. However, in casual laboratory testing, dissolved air in the water will cause higher compressibility and hence volume change upon pressure/temperature increments. Therefore, in order to obtain correct volume change values, a preliminary calibration test investigating the thermal expansibility of water in the system should be carried out prior to a long term unsaturated triaxial soil testing. 3.2 Laser beam refraction due to covering window
Figure 4. Example of measurement.
specimen volume change
As shown in Figure 2, the measurement cell of the new volume change system is sealed with a perspex window. While sealing the measurement cell and maintaining the required air pressure, the window allows the laser beam to pass through it and reach a reflection surface on the object measured. The window would therefore cause refraction of the laser beam and this effect (varying with the window
753
thickness) had to be determined prior to triaxial testing. The measurement system was assembled as for normal triaxial testing. Four perspex windows of different thickness were tested (3.1 mm, 5.7 mm, 11.3 mm and 20 mm respectively) to measure the water level change. An immediate change was observed due to the thickness of the window affecting the refraction of the laser beam. However after this immediate change giving an offset value (linearly increasing with window thickness) the measured value of the water level change was the same as that without window, independent of the window thickness. Figure 5 shows that a 20 mm-thick window (used in the present design) generates 7.7 mm offset value in terms of the refraction. The measured values can therefore be corrected to eliminate the refraction effect. 3.3 Calibrations for cell pressure, temperature and creep
in Figure 2) was placed on the top of the specimen. The volume change measurement system was then filled with water to its capacity (approximately 750 cm3). A thin layer of silicon oil was dropped on the water surface to reduce water evaporation (Cui & Delage, 1996; Ng et al, 2002). Because of the type of the specimen used, no back pressure was applied. 3.3.1 Calibration coefficient between volume change and the raw output value of the transducer The calibration test was performed at various cell pressure levels (0, 100, 200, 300, 400, 500 and 600 kPa respectively). During the test, the triaxial system was set up as for usual testing, and an advanced GDS pressure/volume 2 MPa/200 cc controller with an empty/fill function was used to adjust the water level. The accuracy and resolution of the volume measurements of the GDS controller is 0.1% and 0.5 mm3 respectively. Before using the GDS controller calibration as given by the manufacturers, a proof test was also performed: the expelled water was weighed using a high accuracy balance (readable to 0.001 g). The reading of the balance was in agreement with the reported value of the volume from the controller (as displayed digitally) multiplied by the water density. A relationship between a given water volume change and the raw output value of the laser transducer was thus established and regression analysis performed to fit the data and obtain the Calibration Factor (CF) for different cell pressures (see Figure 6). It is notable that the CF is different for rising and falling water levels. In general the CF is lower when the water level is falling, than when the water level is rising. This could possibly be due to the fact that the circumference line of the water meniscus in the open cell ( (52 mm 38 mm)) is larger than that of the measurement cell ( 35.5 mm) and this could possibly cause some difference in the water surface tension between the rising and the falling water conditions. Furthermore the viscosity of the thin layer of silicon oil may have contributed to this hysteresis effect. The greatest difference in CF
Measured volume changes can be affected by a number of factors. Other than the effect of water compressibility which has already been investigated, such factors include the response to cell pressure changes, creep of the system, temperature and hysteresis of water flow. Moreover, a calibration factor is needed between the raw output of the transducer and volume change of specimen. To ensure that these effects are taken into account and correct soil volume change measurements made, a number of further calibration tests were carried out. These included both short- and long-term calibration tests. A solid brass specimen with diameter 38 mm and height of 76 mm was used in all types of calibration tests. The bulk moduli of brass and water are 6.1 1010 N/m2 and 2.2 109 N/m2 respectively. This implies that volumetric strains under a 600 kPa pressure are 9.8 106 and 2.7 104 for brass and water respectively, i.e. volume change of brass is much smaller than that of water and can be neglected. The water used was distilled water, which was subsequently de-aired. The specimen was sealed with O-rings in a double latex rubber membrane layer. The specifically designed long top cap (shown
Figure 5. Offset value due to refraction for 20 mmthick window.
Figure 6. Calibration factors (CF) for various cell pressure levels.
754
between the two conditions occurred at 0 kPa. This could be due to small air bubbles dissolved into the water or escaping out of the liquid interface during the calibration progress. It is therefore important to select the correct CF for each cell pressure. Calibration for the apparent volume change with increasing cell pressure While increasing cell pressure, a quasi-instantaneous apparent volume change reading can take place. This can be due to compression of any trapped air, compressibility of water, compressibility of rubber membranes and expansion of connecting tubes and valves. To assess such immediate apparent volume changes in the system, the inner cell was filled with 750 cm3 of water. The target cell pressure was selected and applied through a GDS controller. The selected target pressure was reached within 100 minutes and was then maintained for 30 minutes. The pressure cell was subsequently decreased to 0 kPa at the same rate. As during this short-term tests temperature change was very small, its effect was neglected. The relationship between immediate apparent volume change against cell pressure obtained by this test is plotted in Figure 7. It can be seen that after 200 kPa the curve is approximately linear with the apparent volume change being 0.43 cm3 at 600 kPa. The compression line of air-free water for a temperature of 25C (the average room temperature during the test) can be determined as explained in section 3.1, and its effect removed from the measured apparent volume change. The new corrected apparent volume change line (having subtracted the volume change due to water compressibility) shows that for pressures above 100 kPa, apparent volume changes are very small, meaning that the observed apparent volume change above this pressure level was mainly due to water compressibility. 3.3.3 Long-term calibration for system creep and room temperature effect Because of the time scales needed to complete triaxial tests for unsaturated soils, it is also important 3.3.2
to know the long-term performance of the new volume change measurement system. To this effect, long term calibration testing was carried out. This investigated the effect of the creep of the system and room temperature. In this test the cell pressure was increased from 200 kPa to 300 kPa and then maintained at 300 kPa in a non-controlled environment, for 51 days. The combined effects are shown in Figure 8 whereas Figure 9 represents further measured volume changes due to temperature variation. An immediate volume change is shown after increasing the pressure to 300 kPa. For a 5.2C change in temperature (the maximum temperature change while testing), a change of 0.748 cm3 in volume was observed. This corresponds to a 0.143 cm3/C change which, divided by the approximate water volume in the volume change measurement system (750 cm3) gives a value of 0.19 103/C at 25C (i.e. the average room temperature). This is fairly consistent with the theoretical value of the coefficient of thermal expansion of water at 25C (i.e. 0.2569 103/C) and gives confidence for the performance of the system. In time, further apparent volume change took place. Subtracting the effect of temperature from the total apparent volume change, the creep volume change was found. This depended on the intrinsic characteristic properties of the system. Further tests are scheduled to investigate the
Figure 8. Apparent volume change and elapsed time after an increase in cell pressure from 200 to 300 kPa.
Figure 7. Apparent volume change upon cell pressure variation.
Figure 9. Measured volume changes versus temperature.
755
measurements of volume changes of saturated samples during triaxial testing using the suggested technique and conventional techniques showed overall very good agreement and gave confidence in the technique. Overall this study proved the feasibility of the technique and showed promise for the development of a fairly simple tool enabling accurate volume measurements during triaxial testing. Further studies on factors affecting the performance of the system and refinement of the technique are scheduled in order to gain further data and confidence regarding the performance.
Figure 10. Comparative volume change of saturated sample.
ACKNOWLEDGEMENTS The authors would like to thank Mr. Steve Ackerley (Department of Civil and Environmental Engineering, Imperial College, London) for his constructive comments on the development of the new volume change measurement system. This study constituted part of a larger research programme on the hydro-mechanical properties of lime-treated clays, funded by the UK Engineering and Physical Sciences Research Council (EPSRC) through grant EP/E037305/1. REFERENCES
Adams B.A., Wulfshon D. & Fredlund D. 1996. Air volume change measurement in unsaturated soil testing using a digital pressurevolume controller. Geotechnical Testing Journal 19(1): 1221. Cabarkapa Z. & Cuccovillo T. 2006. Automated triaxial apparatus for testing unsaturated soils. Geotechnical Testing Journal 29(1): 2129. Cui Y.J. & Delage P. 1996. Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique 46 (2): 291311. Fine R.A. & Millero F.J. 1973. Compressibility of water as a function of temperature and pressure. Journal of Chemical Physics 59 (10): 5529. Gachet P., Klubertanz G., Vulliet L. & Laloui L. 2003. Interfacial behaviour of unsaturated soil with smallscale models and use of image processing techniques. Geotechnical Testing Journal 26 (1): 1221. Ng C.W.W., Zhan L.T. & Cui Y.J. 2002. A new simple system for measuring volume changes in unsaturated soils. Canadian Geotechnical Journal 39(3): 757764. Romero E. 1999. Characterization and thermohydromechanical behaviour of unsaturated Boom clay: an experimental study. PhD thesis. Barcelona: Universitat Politecnica de Catalunya. Wheeler S.J. 1988. The undrained shear strength of soils containing air bubbles, Gotechnique 38(3): 399413. Yin J. H. 2003. A Double Cell Triaxial System for Continuous Measurement of Volume Changes of an Unsaturated or Saturated Soil Specimen in Triaxial Testing. Geotechnical Testing Journal 26 (3): 16.
repeatability of the creep and temperature effects under different room temperature conditions in order to confirm the calibration coefficient. 4 COMPARATIVE VOLUME CHANGE MEASUREMENTS DURING TRIAXIAL TESTING
After calibrations the suggested technique has been used for comparative volume change measurements of saturated soil samples tested in a GDS triaxial system using a set of three advanced GDS controllers. Figure 10 shows comparative measurements of volume changes of a saturated fine sand dense sample obtained respectively from the back pressure line using a GDS controller and from the laser transducer using the suggested methodology to determine volume changes. This particular test was performed at a cell pressure of 200 kPa and a back pressure of 100 kPa at a rate of shearing of 2.234 mm/hour. Overall the results show very good agreement and give confidence for the feasibility of the suggested technique. Note that there appear to be some fluctuations in the results of the back pressure line especially at axial strains of 10% and above, as opposed to the values obtained using the suggested methodology with the laser sensor. These are believed to have been caused by some fluctuations of the axial load during shearing. 5 CONCLUSIONS
A new volume change measurement system based on a non-contact laser displacement transducer and the cell liquid measurement technique was introduced and demonstrated. The design and operating principles of the system were presented and a number of careful calibrations made. Comparative
756
Modelling
Fundamentals and constitutive modelling
Mechanical model for partially saturated sandy soils: Application to liquefaction

H.B. Bian
LPMM, FRE CNRS 3236, Universit Paul Verlaine-Metz, Ile du Saulcy, Cedex 01, France
T. Nishimura I. Shahrour
Ashikaga Institute of Technology, Japan LML, UMR CNRS 8107, Universit de Lille I, Cit Scientifique, France
ABSTRACT: A numerical model is presented for the partially saturated sandy soil in seismic area, with the particular interests on the liquefactions. The formulation is based on Coussys unsaturated poromechanical theory with reasonable assumptions for sandy soils. Based on the conception of effective stress, a constitutive equation with two loading surfaces is used for describing the mechanical behaviour of sandy soils under cyclic loading. And then it is applied to study the dynamic response of free filed under the earthquake loading. It is shown that, with the decrease of the initial water saturation, the liquefaction resistance increases. The variation of initial water saturation will also have the influence on the dynamic responds. And its frequency will also have some changes. 1 GENERAL INSTRUCTIONS the saturated case, the hydro-mechanic coupled theory founded by Biot (Biot, 1941) together with the concept of effective stresses proposed by Terzaghi (Terzaghi, 1936) are used to study the liquefaction phenomena in saturated soils (Zienkiewicz et al., 1984). New developments in the area of unsaturated soils (Fredlund, 1993; Coussy, 2004) together with the generalization of the concept of effective stress to unsaturated soils (Bishop, 1959) allow dealing the unsaturated problem as a coupled problem. In the present work, the study will focus on the mechanical model for unsaturated sandy under the earthquake loading by taking the influence of water saturation. It aims to establish an adequate and robust constitutive equation for partially saturated sandy soils in view to investigate the phenomena of liquefaction of sandy soil under the dynamic loading. 2 2.1 NUMERICAL MODEL Basic assumptions
Soil liquefaction constitutes a major cause of damage induced by earthquakes. It is widely recognized that liquefaction during earthquakes is due to the progressive build-up of excess pore-water pressure resulting from cyclic shearing (Sawada et al., 2006). During the past decades, intensive efforts have been carried out to understand the mechanism of liquefaction, and to develop methods for evaluating liquefaction potential in saturated soils. However, in recent years many investigations indicated that that initial water saturation has great influence on the liquefaction resistance of sandy soil. Both laboratory and in situ observations showed that a reduction in the degree of saturation results in a significant increase in the soil resistance to liquefaction (Martin et al. 1978; Yoshimi et al., 1989; Sawada et al., 2006). More important, most soils encountered in geotechnical engineering are unsaturated, even located under the phreatic surface the soils are not as usual supposed, full saturated, but in an initially unsaturated state. So it is of great interest to extend the study of liquefaction to unsaturated state. The partially saturated sandy soils could be described as the supposition of 3 continuous phases: the soils skeleton, the pore water and pore air. It is considered partially saturated for a referenced fluid, generally chosen in liquid form (pore water). For
This part is devoted to present the numerical model for partial saturated free field with the focus on liquefaction problem. The numerical model is based upon several assumptions. Firstly, we supposed that the soil grain is incompressible, the volumetric deformation is due to the variation of the pore volume (the drainage of pore water and the
763
compression of pore air). The mechanical property of pore water is supposed as linear elastic, e.g. the compressibility of pore water is constant. While for the pore air, it is assumed as ideal gas. When liquefaction is considered, the air flux is supposed as zero. For dynamic problem, the interaction between pore air and pore water is neglected, the dissolution of air into water will not be taken into consideration. Isothermal condition will be adopted; the exchange between water vapour and liquid water is neglected. 2.2 The coupling equations
After a small perturbation, under drained or undrained conditions, the assumption of the soil grains incompressibility leads to the following expressions: V trV Va Vw (4)
According to the ideal gas law, the variation of the pore-air volume is equal to: p0lq pa 0 0 1 Va Va plq pa 0 (5)
Based on the theory of Coussy (Coussy 2004), Bian and Shahrour (2009) have developed a numerical model for partially saturated sandy soils, supposing the pore air pressure equals to that of pore water. The increment of the total stress and pore pressure pw is given as the function of the water flux mw and deformation increment , as illustrated below:
The current pore-water and pore-air volumes as follows:

0 Vw Vw V dVa ; Va Va0 Va
(6)
The current water saturation is equal to: Sw Vw Vw Va (7)
m C : M w I w
m pw pa M tr( ) M w w
(1)
Where C is the drained elastic matrix, I is the unit matrix with two order, The Biot modulus M for unsaturated sandy soils is the function of the compressibility of water Kw, the porosity n and the water saturation Sw, as: n n n(1 Sw ) nSw 1 a w M K a K w pw pa 0 Kw (2)
Combination of Equations (3) to (7) leads to the following expression for the current water saturation: Sw
0 n0Sw v
p0 w pa 0 pw pa 0
0 ) 1 n0 (1 Sw
n0 v
(8)
2.3
Field equations
It can be observed that equation (1) are similar to that used for saturated soil (the U-P model proposed by Zienkiewicz), but with a Biot modulus depending on the porosity, pore pressure and water saturation, while for saturated soil, the modulus is constant. It is of interest to indicate that, in equation (2) if the water saturation is unit, that means the soil is saturated, the formulation for saturated soil is recovered. It also means that this formulation can be used for both saturated and unsaturated problems. The water saturation in equation (2) should be solved. Normally for unsaturated soil, the water retention curve is applied. However in case of sandy soil, the idea gas law together with zero air flux assumption could be used to solve the current water saturation. We consider an unsaturated soil specimen, with a volume V and initial measured pore-pressure p0lq. The initial porosity is n0 and the 0 initial water saturation is Sw . So the initial pore-air 0 Va and pore-water volume Vw0 are:
0 0 0 Va0 n0 (1 Sw )V ; Vw n0Sw V
By choosing the displacement and the water pressure as principal variables, the problem of the partially saturated free field is controlled by a set of field equations: the balance equation, the diffusion equation and the conservation of the mass. 1) The balance equation: When adopting the small perturbation hypothesis and neglecting the body force, the balance equation is: div() 0 (9)
is the total stress tensor and is the acceleration vectors of soil skeleton. is the soil density, in case of unsaturated, it depends on respectively soil porosity, the densities of the solid grains and pore-water, while the density of the pore-air is neglected: (1 n)s nSww (10)
(3)
With n the porosity, Sw the water saturation, s the density of soil grains and w the density of water.
764
2) The diffusion law: For instance, the generalized Darcys law is used: k rl w kins w grad ( pw ) w u w (11)
pa is a reference pressure, N is a constitutive parameters; while p and q stand for the mean stress and the deviatoric stress, respectively. The constitutive relation uses two loading surfaces. The first one is the limit surface, which describes the soil behavior under monotonic loading: fm q Mf pRm (16)
rl Where w is the pore-water mass flux, kins and kw are the soil intrinsic permeability and the relative permeability associated with water, respectively. is the viscosity of pore water. While for the pore air, zero mass flux is assumed, so the diffusion equation for pore water is sufficient. However, the presence of the air will significantly affect the diffusion of pore water. As a soil becomes unsaturated, air replaces firstly some of the water in the large pores, and this causes the water to flow through the smaller pores with an increased tortuosity to the flow path. As a result, the coefficient of permeability with respect to the water phase decreases rapidly as the space available for water flow reduces. The relative permeability is normally expressed as the function of water saturation as:
rl kw
The hardening function Rm depends on the plastic deviatoric strains ( dp): Rm

p d p b d
(17)
b is a model parameter. The constitutive relation takes into consideration the variation of the soil resistance in the deviatoric plane using the following expression for Mf: Mf 6 sin 3 sin sin 3
Sw (1 (1 Sw
1 m ) m )2
(12)
Where m is the model parameter, it can be calibrated from the laboratory measured data. 3) The conservation of the mass: The expression of the overall mass conservation must take into account the relative masse flow of each fluid phase: pore-water and pore-air. In the absence of overall creation and null gas flux, the overall mass conservation will only applied to pore water and can be written as follows: w div(w ) m 2.4 The constitutive equations for sandy soil (13)
and denote the internal friction angle and the Lodes angle. Non associated flow rule is used in the model. The gradient of plastic potential is given by:
p ) gm exp( 0 d q Mc Mc p p p g m 1 Mc p q
(18)
With Mc 6 sin cv 3 sin cv sin 3
(19)
The present constitutive equation for sandy soils is characterized by using two loading surfaces, monotonic loading surface and cyclic loading surface respectively, with a non associated plastic flow rule. The elastic part of the model is governed by nonlinear elasticity. The Shear modulus and the bulk modulus are determined according to the following expressions. p K K 0 A( p, q ); pa
N
0 is a constitutive parameter,cv stands for the characteristic angle. The cyclic loading surface is assumed to be conic in the stress domain. Its expression is given by: fc ql - plRc With
l l q l sij sij l sij ij pl ij
(20)
p G G0 pa
pl ij ij (14)
Rc
p dc p b d c
(21)
1 1v 0 A( p, q ) 1 . .N 9 1 2v0
N q . p
1
(15)
Rc controls the isotropic hardening of the cyclic p loading surface, dc denotes the plastic deviatoric deformation associated to the cyclic loading, which should be initialized at each loading inversion.
765
The unit tensor ij denotes the axis of the cyclic loading surface; its evolution is governed by: d ij d Hij
3 3.1
LIQUEFACTION OF PARTIALLY SATURATED FREE FIELD Introduction
l Hij cc (1 AF )Rc sij
(22)
d is the plastic multiplier; cc is a constitutive parameter. A and F are determined as follows: A 1

l if sij sij 0
l A 1 if sij sij 0 q F ( F 1) M f pRm
(23)
The gradient of the cyclic plastic potential is given by: g m d ij 0 ij gc A q 3 sij Mc ij ij p p 2M c p q gm if d ij 0 ij l sij 3 jl gc 3 A l Mc l 2 ij l l ij p 2Mc p j2 p p B exp( 0 d ) With A ; B exp( bc vp ) 3Mc if
The proposed numerical model is implemented in a finite element calculation code. For solving the problem in the time domain, the Newmark method is used. After several validation examples, we now consider a free field. For the reason of simplicity, the free filed is supposed as homogenous of only one material: Nevada sand. The geometry of the considered free field is presented in Figure 1. The depth of the soil layer is equal to 30 meters; the width of the domain considered is arbitrary supposed as B. The boundary conditions are defined as following: 1) at the base of soil layer, the vertical displacement is blocked, and it is impermeable; 2) at the lateral boundaries, the equivalent displacement and pore pressure are imposed; 3) at the top of the soil layer, it is permeable. The physical properties of soil (Nevada sand) are extracted from the VELCAS experiment report (Arulmoli et al., 1992). They are listed in Table 1, in fact, the soil consider is looses sand with relative density as Id 0.4. The loads applied at the base of soil layer are the real earthquake records of 1995 Kobe earthquakes. The displacement history and the acceleration history recorded during the earthquake are presented in Figure 2. 3.2 Initial conditions The problem considered is nonlinear; the initial stresses may have a great influence on the numerical results. The initial conditions include the initial stress distribution and initial pore water pressure distribution.
(24)
vp is the plastic volumetric deformation, bc is a constitutive parameter which controls the plastic deformation under cyclic loading. One of the key points in this constitutive equation is to determine which loading surface should be used during the loading process. The criterion is defined as follows: firstly, the loading process is checked with the monotonic loading surface, in the event of unloading, the current active loading surface (monotonic loading surface) is memorized and a new loading surface (cyclic loading surface) is initialized and activated. During the loading process, at each increment, we check the monotonic loading surface. If this surface is checked as active, the cyclic loading surface is deactivated; the behaviour is then controlled by the monotonic loading surface. The present constitutive equations for sandy soil is based on the effective stress conception; so it can be easily implemented in present numerical model presented in the previous section.
Figure 1. Geometry of the liquefiable free filed (Id 0.4).
766
Table 1. Model Parameters of Nevada sand. Youngs modulus Poison ratio Parameter Friction angle Dilation angle Parameter Parameter Parameter Parameter E0 v0 N cv b cc bc 100 kPa 0.23 0.5 34.8 30.5 0.00094 8 0.024 100
0.25 0.20 0.15 0.10 0.05 0.00 -0.05 -0.10 -0.15 -0.20 -0.25 8 6 4 2 0 -2 -4 -6 -8 0 0
Acceleration(g)
Time(s) 5 10 15 20 25 30 35 40 45
the water table, it is not as usually assumed, fully saturated, but in a state of partially saturated state; the gas (the mixture of dry air and water vapour) in form of small air bubbles locked in the corner of soil grains. These gases are sometimes called residual gas in the pore, and it is difficult to remove it out. The residual gas may come during the deposit history or it may caused by the fluctuation of the phreatic surface caused by natural or manmade behaviour. The amount of the residual gas in the soil pore strongly depends on the soil grain size and the distribution of pores. The distribution of water saturation could be supposed as uniform if the soil is homogenous. The laboratory and in situ observations have already confirmed this hypothesis. For example, the installation of sand compaction pile will introduce great deal of air into soil (Okamura et al., 2006). In our case, we suppose five different saturations for soil under the phreatic surface, Sw 1.0, 0.99, 0.98, 0.95, and 0.90. Then these soils columns are submit to the earthquake loading. 3.3 Influence of water saturation on liquefaction resistance
Displacement(cm)
Time(s) 5 10 15 20 25 30 35 40 45
Figure 2. Ground acceleration and displacement records during Kobe Earthquake in 1995, Japan.
Generally, the soils are divided into two distinguished parts by the phreatic surface. Those below the phreatic surface are considered as saturated, while those above the water table are unsaturated. In fact, in natural state, the interstitial pores of soil are saturated with dry air, water vapour and liquid water. For the reason of simplicity, the distribution of water saturation is supposed as homogenous for both soils located at the two sides of phreatic surface. For those above the phreatic surface is supposed as dry soil, while those below the phreatic surface are as with uniform water saturation. Normally the initial stress in soil layer is due to the gravity action. According to equation(10), we can easily give the initial stress distribution if we know the distribution of water saturation. However, for the reason of comparison, the same density is used for all unsaturated cases considered in this study. While for the initial distribution of pore water pressure, it only relates to the distance from the phreatic surface. We suppose that the phreatic surface located at 2 m below the ground surface. For the soil below
The profile of the distribution of excess pore pressure is illustrated in Figure 3. The variation of the pore-pressure is normalised by the initial effective stresses; while the depth is normalised by the total height of the soil column. The liquefaction phenomena have been observed for the three previous cases (Sw 1.0, 0.99, and 0.98), while for the two other with lower water saturation (Sw 0.95 and 0.90) there is no liquefaction. It can be seen that with the decrease in the water saturation, the liquefaction zone will decrease. For example, for fully saturated case, the liquefaction zone concerns about 40% of the layer. In the case of water saturation 98%, the liquefaction zone is reduced to about 20%. The liquefaction resistance of partially saturated soil is greater than that of saturated soil. For instance, at the depth Z 0.2H, the ratio of excess pore-water pressure to initial effective stresses is about 0.87 for saturated case, while this ratio decreases to 0.82,
1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 dpw/' 1
Z/H
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90
Figure 3. Influence of water saturation on the distribution of excess pore-pressure in function of the depth.
767
1.2E-02 1.0E-02 8.0E-03 6.0E-03 4.0E-03 2.0E-03 0.0E+00
Amplitute
Z=30m
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90 REC_DIS
Amplitute
1.4E-04 1.2E-04 1.0E-04 8.0E-05
4.0E-01 3.5E-01 3.0E-01 2.5E-01 2.0E-01
Amplitute Z=30m
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90 REC_ACC
Amplitute
9.0E-09 8.0E-09 7.0E-09 6.0E-09 5.0E-09 4.0E-09 3.0E-09 2.0E-09
6.0E-05 4.0E-05
1.5E-01 1.0E-01 5.0E-02 0.0E+00 0 1.0E-01 9.0E-02 8.0E-02 7.0E-02 6.0E-02 5.0E-02 4.0E-02 3.0E-02 2.0E-02
Frequency (Hz)
0 0.1 0.2 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
2.0E-05 0.0E+00
Frequency (Hz)
0.2 0.4 0.6 0.8 1
1.0E-09 0.0E+00
7.0E-04 6.0E-04 5.0E-04 4.0E-04 3.0E-04 2.0E-04 1.0E-04 0.0E+00 0
Amplitute
Z=15m
Amplitute
1.4E-04 1.2E-04
Amplitute
Z=15m
Amplitute
9.0E-09 8.0E-09 7.0E-09
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90 REC_DIS
1.0E-04 8.0E-05 6.0E-05 4.0E-05
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90 REC_ACC
6.0E-09 5.0E-09 4.0E-09 3.0E-09 2.0E-09
Frequency (Hz)
0.2 0.4 0.6 0.8 1
2.0E-05 0.0E+00
1.0E-02 0.0E+00 0
Frequency (Hz)
0.2 0.4 0.6 0.8 1
1.0E-09 0.0E+00
Figure 4. Displacement spectrums for the nodes at the ground surface (Z 30 m) and at the centre layer (Z 15 m).
3.0E-02 2.5E-02 2.0E-02 1.5E-02 1.5E-04 1.0E-02 5.0E-03 1.0E-04
Figure 6. Acceleration spectrums for the nodes at the ground surface (Z 30 m) and at the centre layer (Z 15 m).
Amplitute
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90 REC_VEL
Z=30m
Amplitute
3.5E-04 3.0E-04 2.5E-04 2.0E-04
Frequency (Hz)
0.0E+00 0 0.2 0.4 0.6 0.8 1
5.0E-05 0.0E+00
8.0E-03 7.0E-03 6.0E-03 5.0E-03 4.0E-03 3.0E-03 2.0E-03 1.0E-03 0.0E+00 0
Amplitute
Sw=1.00 Sw=0.99 Sw=0.98 Sw=0.95 Sw=0.90 REC_VEL
Z=15m
Amplitute
3.5E-04 3.0E-04 2.5E-04 2.0E-04 1.5E-04 1.0E-04
displacement has great influence. And the same results can be observed on nodal velocity and nodal acceleration for the two given nodes; these results are presented in Figure 5 and 6 respectively. It seems that, with the decrease of water saturation, the period of free vibration of the soils column has the tendency of increase. For example, in Figure 6, the acceleration spectrum of the node at the top of soil column (Z 30 m), the most impotent frequency is about 0.51 Hz in case of initial water saturation equal 0.98, while it equals to 0.54 Hz and 0.59 Hz for the initial water saturation as 0.95 and 0.90 respectively. 4 CONCLUSION
Frequency (Hz)
0.2 0.4 0.6 0.8 1
5.0E-05 0.0E+00
Figure 5. Velocity spectrums for the nodes at the ground surface (Z 30 m) and at the centre layer (Z 15 m).
0.76, 0.64 and 0.51 for the initial water saturation 99%, 98%, 95% and 90% respectively. 3.4 Influence of water saturation on dynamic responds
In the present work, a numerical model is presented for the partially saturated sandy soil in seismic area. It is focus on the influence of water saturation on the liquefaction. The numerical example shows that the water saturation has great influence on the liquefaction resistance. With the decrease of water saturation, the liquefaction has the increase tendency. It also shows that, with the decrease of water saturation, the period of free vibration of the soils column has the tendency of increase. REFERENCES
Arulmoli K., Muraleetharan K.K., Hossain M.M., Fruth L.S., 1992, VELACS Verification of liquefaction analyses by centrifuge studies laboratory testing program soil data report.
Two nodes displacements, at the ground surface (Z 30 m) and at the centre of soils layer (Z 15 m) respectively, are analyzed by using the fast Fourier transform (FFT) method. The results are presented in Figure 4. From the figure, it can be seen that with decrease of water saturation, the nodal
768
Bian H., Shahrour I., 2009, Numerical model for unsaturated sandy soils under cyclic loading: Application to liquefaction. Soil Dynamic and Earthquake Engneering, Vol. 29, pp. 237244. Biot M.A., 1941, General theory of three dimensional consolidation, Journal of application physic, Vol. 12, pp. 155164. Bishop A.W., 1959, The principle of effective stress, Teknisk Ukeblad 39, 859863. Coussy O., 2004, Poro mechanics, John-Wiley. Fredlund D.G., 1993, Rahardjo H., Soil mechanics for unsaturated soils, John-Wiley. Martin G.R., Finn W.D.L., Seed H.B., 1978, Effects of system compliance on liquefaction testes, Journal of the geotechnical engineering division, Vol. 104, No. GT4. Okamura M., Ishihara M., Tamura K., 2006, Degree of saturation and liquefaction resistances of sand improved with sand compaction pile, Journal of
geotechnical and Geoenvironmental engineering, ASCE, Vol. 132, No. 2, pp. 258264. Sawada S., Tsukamoto Y., Ishihara K., 2006, Residual deformation characteristics of partially saturated sandy soils subjected to seismic excitation, Soil dynamics and earthquake engineering, Vol. 26, pp. 175182. Terzaghi K., 1936, The shearing resistance of saturated soils and the angle between the planes of shear, Proceedings of the 1st International Conference on Soil Mechanics and Foundation Engineering, Vol. 1, 5456. Yoshimi Y., Tanaka K., Tokimatsu K., 1989, Liquefaction resistance of a partially saturated sand, Soils and foundations, vol. 29, No. 3, pp. 157162. Zienkiewicz O.C., Shiomi T., 1984, Dynamic behaviour of saturated porous media; the generalized Biot formulation and its numerical solution, International journal for numerical and analytical methods in geomechanics, Vol. 8, pp. 7196.
769
Saturation induced instability on bonded geomaterials

G. Buscarnera & R. Nova
Politecnico di Milano, Department of Structural Engineering, Italy
ABSTRACT: Certain types of unsaturated bonded geomaterials, as for example residual soils, loess or volcanic deposits, are often characterized by the presence of two types of bonding effects, one due to cementation among grains and the other due to surface tensions in water menisci. In these soils unstable processes can arise from coupled hydro-mechanical degradation phenomena, which can even produce the vanishing of either of the two contributions. In this work, both mechanical bonding and partial saturation are accounted for within the same strain-hardening constitutive model. The constitutive model is then used in the light of a theoretical approach aimed at interpreting the onset of hydro-mechanical instabilities in unsaturated soils. Oedometric tests are numerically simulated, showing the possibility of unstable volumetric collapses taking place as a result of saturation processes. The influence of different degrees of initial mechanical bonding on the occurrence of such phenomena is finally investigated, pointing out the existence of a critical range of bonding percentages which exacerbate the potential for mechanical instability upon saturation. 1 INTRODUCTION be the driving factor for unstable phenomena. Some authors have introduced mechanical bonding and partial saturation within the same conceptual framework (Garitte et al. 2006, Koliji et al. 2008). Very few works, though, considered the link between the vanishing of these two bonding contributions and the concept of mechanical instability (Buscarnera & Nova 2009c). The main goal of this paper is to address the problem of soil instability for those natural materials characterized by a non negligible degree of mechanical bonding. For this purpose the modelling approach proposed by Jommi & di Prisco (1994) for unsaturated soils is extended to bonded geomaterials by following the procedure suggested by Gens & Nova (1993). The resulting constitutive law is then used in the light of a recently developed theoretical framework aimed at dealing with soil instability in the unsaturated regime (Buscarnera & Nova 2009b). Particular attention is given to instability in oedometric testing of unsaturated bonded soils. Model predictions are shown, focusing on the role played by hydraulic test control conditions and the initial degree of mechanical bonding. It is shown that under certain circumstances wetting induced collapse can be interpreted as an unstable process. In particular, a range of critical bonding percentages can be envisaged, which is associated to a material fabric prone to develop a catastrophic collapse upon suction reduction processes.
The mechanical properties of natural soils influencing the analysis of engineering problems are the outcome of a complex interaction with the surrounding environment. The resulting behaviour of such soils is often difficult to understand within the classical soil mechanics framework and more sophisticated approaches are necessary in order to describe the experimental evidence. The behaviour of a variety of natural soils (e.g., loess deposits, residual soils, geomaterials of volcanic origin etc.) is often influenced by the combined presence of mechanical bonding and partial saturation. Even though these two contributions derive from rather different microscopic causes, their most significant effects on the macroscopic mechanical response of soils are quite similar. Both the mechanical bonding due to grain cementation and the hydraulic bonding related to water menisci, in fact, increase the overall material stiffness and confer a non zero tensile strength to an otherwise unbonded soil. Partial saturation effects, however, can disappear if the material undergoes a wetting path (Alonso et al. 1990), while the bonding between grains can be progressively removed as a consequence of a mechanical damage at the micro-scale (Leroueil & Vaughan 1990). The action of these two degradation processes can be simultaneous in geomaterials characterized by both bonding contributions, producing a much more intense weakening of the material that can
771
A CONSTITUTIVE MODEL FOR UNSATURATED BONDED SOILS
In order to investigate the occurrence of instabilities in unsaturated bonded soils it is first necessary to define an appropriate constitutive model for these materials. The first approach for modelling the constitutive behaviour of unsaturated soils was based on the adoption of two independent stress variables, namely net stress and suction, and on the introduction of the loading collapse yield curve concept (Alonso et al. 1990). Following these fundamental ideas, great advances have been achieved in unsaturated soil constitutive modelling. Several works have in fact been oriented towards a more complete hydro-mechanical description of unsaturated soils and the use of different stress-strain variables (see the review by Gens et al. 2006). In this work the constitutive approach suggested by Jommi & di Prisco (1994) has been adopted in the light of the energy framework put forward by Houlsby (1997). The following measure of stress, hereafter referred to as average skeleton stress, has been used for modelling purposes:
ij ij Sr uw ij (1 Sr ) ua ij
Starting from these premises, a coupled hydromechanical constitutive law for unsaturated soils has been developed. The model is conceptually similar to other models already available in the literature, but it is formulated in a way that is convenient for the analysis of the possible occurrence of instabilities, which constitutes the major goal of the paper. The constitutive law is based on a coupled formulation which includes modified suction and degree of saturation as further stress and strain variables, as follows:
. . D D *ij w ij D ext E D w D ww Sr s
(4)
(1)
where ij is the total stress, uw the pore water pressure, ua the air pore pressure, s ua uw the suction, Sr the degree of saturation and ij the Kronecker delta. In this way the skeleton stress ij represents the intensive measure work conjugate to the strain rate. In the particular case of partially saturated soils the energy input is not only due to the mechanical work done by the external forces, however. Since the soil is unsaturated, the possible change in water content can alter the volume occupied by fluids without causing any strain of the soil skeleton. This water volume variation is associated to an energy exchange between water and air at the water menisci. According to Houlsby (1997) the total specific energy is then given by:
ij ns S dW ij r
The components of the extended coupled constitutive matrix will depend on the mathematical expression adopted for the constitutive functions. At this reference, it must be noted that hydraulic hysteresis and strain dependency of the water retention curve have been neglected for the sake of simplicity and that a non associated flow rule has been used to model irreversible strains. Only some details of the mathematical features of the hardening laws will be discussed in the following, being these laws a fundamental component to describe hydro-mechanical degradation processes. A detailed description of the constitutive functions chosen to model the elastic behaviour, the yield surface and the plastic potential is available in Buscarnera & Nova (2009a). The adoption of the stress measure given by Equation (1) provides several advantages. As pointed out by Jommi (2000), in order to capture some key features of unsaturated soil response, modifications have to be introduced into the classical modelling framework for saturated soils, however. For this purpose, the hardening law controlling the size of the yield domain of the unbonded soil is assumed to be governed by two separate contributions, a mechanical one and a hydraulic one. The analytical expression of the hardening law used in this work is expressed as follows:
p p V S s s ps s rsw psS p r
(2)
(5)
where n the porosity of the soil. Equation (2) confirms the experimental evidence that a single effective stress theory is not able to describe the behaviour of an unsaturated soil. Three phase porous media are indeed affected by a further energy term, which can be expressed by the product of a smeared suction: s* n s (3)
having the opposite of the rate of the degree of as conjugate strain measure. saturation S r
where s, s and rsw are hardening constitutive parameters. The first term is the usual hardening relationship for granular materials, that is conveniently expressed in terms of both volumetric and deviatoric plastic strains. The second term in Equation (5) represents instead the mathematical form of a hydraulic bonding effect. In the framework herein suggested, in fact, the degree of saturation is seen as a global measure of the distribution of water menisci, and therefore of the capillary forces associated to surface tensions.
772
As a consequence, the constitutive model assumes a homothetic expansion/contraction of the elastic domain as a result of drying/wetting processes. Figure 1 schematically shows the evolution of the yield surface in the skeleton stress space during a soaking process. The structure of the model presented so far can describe the behaviour of unsaturated soils in which the only bonding contribution is provided by the apparent cohesion related to partial saturation. Thus, in order to account for an actual mechanical bonding among grains, the approach suggested by Gens & Nova (1993) has been followed. A second internal variable pm is then introduced in the formulation, its effects being conceptually related to the mechanical effects of bonds. The new internal variable produces a further expansion of the elastic domain which is completely independent from that linked to partial saturation. The overall size of the elastic domain is then governed by the following variable: pc ps pm pt ps (1 k) pm (6)
Figure 2. surface.
Effect of a mechanical bonding on the yield
where pt represents the isotropic tensile strength of the bonded material (here assumed equal to pt k pm as shown in Figure 2). The hardening law for the internal variable pm is defined as follows:
p p V S m m pm m p
has been implicitly assumed a negligible influence of suction on their evolution during soaking. As a consequence, it is assumed that suction plays a major role on the behaviour of the soil matrix only. This is of course an assumption that may not be in agreement with some experimental findings (Leroueil & Barbosa 2000). Nevertheless, given the lack of systematic observations on this subject, it can be considered as a reasonable first approximation to address the problem. The introduction of non-mechanical degradation processes affecting the degree of bonding, however, can be straightforward, as pointed out by Nova et al. (2003). This aspect will not be considered in the following, assuming mechanical degradation as the only degradation process affecting the degree of mechanical bonding.
(7) 3 A THEORETICAL APPROACH FOR UNSATURATED SOIL INSTABILITY
where m and m are hardening constitutive parameters. Equation (7) governs the evolution of the degree of bonding with volumetric and deviatoric plastic strains. According to this law, yielding will induce a decrease in the amount of bonding, and therefore a tendency to a contraction of the yield surface (mechanical degradation phenomenon). By adopting such a hardening law for the behaviour of the bonds among grains, it
q CSL
Sr<1 wetting path Sr=1
p"
Figure 1. Contraction of the yield surface due to wetting processes depicted in the skeleton stress.
Several approaches can be used to deal with the mechanical instability of soils (e.g., energy approaches, static approaches, bifurcation analyses etc.). In this work the theoretical framework of reference is that proposed by Nova and co-workers (Nova 1994, Imposimato & Nova 1998), often referred to as Theory of Test Controllability. According to this approach soil instability is tackled from the mathematical viewpoint as a bifurcation problem, focusing on the role of test control parameters on the onset of an instability. It can be proven that several types of loss of controllability, and therefore several instability modes, are possible within the classical failure domain (Imposimato & Nova 1998). These instability modes depend on test control conditions, and can be identified by studying the vanishing of suitable mathematical operators which are always associated to the particular loading programme followed during the test. The theory of test controllability can be extended to the case of unsaturated soils by making reference to the hydro-mechanical constitutive stiffness
773
matrix of Equation (4). This idea has been firstly suggested by Vaunat et al. (2002) and subsequently used by Buscarnera & Nova (2009b). In this paper, only the most important features of this extension are presented. A more detailed description of the reference theoretical framework is available in Buscarnera & Nova (2009b). In order to define the test control conditions characterising a laboratory test on an unsaturated material, more complex loading programmes than those used for saturated soils have to be considered. Therefore, either the smeared suction or the saturation index will vary, together with appropriate combinations of stresses and strains. The definition of a loading programme implies the introduction of a control vector , grouping the hydro-mechanical variables actually governed during a test, and a response vector , which is energetically associated to the former set of control variables. A formal representation of both control variables and response variables can be expressed as follows: E (8)
det( Cext) det(1) 0 or equivalently when: det(Dext ) det(2) 0
(11)
(12)
The two mathematical quantities given in Equations (11)(12) can then be defined as instability indices, since their vanishing implies a loss of controllability for the specific control considered. Using such a mathematical approach, the inception of a potential instability can be predicted, provided that the loading programme is suitably defined. Equations (11)(12) show that the suggested instability criterion only depends on the two matrices and (i.e., on the specific loading programme) and either on the extended constitutive stiffness matrix Dext or the extended constitutive compliance matrix Cext (i.e., on the current hydromechanical state). 4 MODEL PREDICTIONS OF INSTABILITY IN UNSATURATED BONDED SOILS
The rate of the control vector is always known as a function of stress and strain variables, being the loading programme fully defined. By contrast, the response vector is defined only from a formal viewpoint. This vector groups the work conjugate variables associated to , and it has to be properly defined on the basis of energy considerations. The two sets of variables can be linked by means of a control matrix X as follows: X (9) The mathematical expression of X is derived from (8) and (4): X Dext Dext

Cext

Cext

1
1
(10)
in which Cext is the extended constitutive compliance matrix. The above expression assumes that both matrices ( Dext ) and ( Cext) are not singular. For a general laboratory test, the rate of control variables can be explicitly defined in terms of generalised stresses and strains. It is then possible to obtain the loss of controllability condition for a given mixed stress-strain hydro-mechanical loading programme. The control of variables is indeed lost when matrix X calculated for the considered loading programme becomes singular, i.e., when:
In a previous study Buscarnera & Nova (2009c) showed that stability can be lost during a soaking test depending on the way in which saturation is governed. In particular, model simulations showed that if the saturation stage is applied in the form of water injection (i.e., under water volume control rather that under suction control), an abrupt loss of controllability can take place. These hydro-mechanical instabilities are identified by the vanishing of any of the two indices of instability (11)(12). When suction is not directly controlled, indeed, the amount of hydraulic degradation is a result of the entire stress-strain response, possibly leading to a dramatic volumetric collapse. This sort of predicted phenomena can be classified as a fully coupled instability, in which both mechanical and hydraulic variables suffer an abrupt evolution. Following this previous work, the influence of an initial mechanical bonding contribution on the stability of unsaturated soil response is studied. In particular, the role played by the initial degree of mechanical bonding in suction controlled oedometric tests is addressed. For this purpose, a certain initial stress state is considered, represented by the stress point A in Figure 3. Yielding upon suction constant loading occurs at a stress level which depends on the initial size of the elastic domain. As already pointed out in the previous sections, the size of the elastic domain will be a function of both the initial suction and the amount of bonds between grains. The initial relative proportion of these two contributions is generally difficult to be known a priori in practical cases.
774
Yielding point
q
0.72
Initial stress state

0.7
pm0 /pc0 =50% (No yielding) Loss of controllability at yielding for test #3 pm0 /pc0 =30%
0.66 (tendency to develop softening collapse)
A
Hydraulic contribution pm0 (Mechanical bonding)
0.68
void ratio
pm0 /pc0 =0%
0.64
test #1: pm0 /pc0 =0% test #2: pm0 /pc0 =15% test #3: pm0 /pc0 =30% test #4: pm0 /pc0 =50% pm0 /pc0 =15%
ps0 (unbonded saturated material)
ps0 (unsaturated material)
p"
0.62
Figure 3. Schematic description of yielding upon oedometric loading at constant suction.
0.6 0.75
0.8
0.85
0.9
0.95
1.05
Sr
At this reference, in fact, Buscarnera & Nova (2009c) showed that similar qualitative responses can be predicted in terms of post-yielding compressibility, irrespectively of the ratio between mechanical and hydraulic bonding. This feature of unsaturated bonded geomaterial response was predicted by using different proportions of mechanical and hydraulic bonding contributions (i.e., by using different values for the ratio pm0/pc0) and by keeping at the same time a constant size for the initial elastic domain (i.e., obtaining yielding at the same level of stress in all tests). The authors pointed out that the mechanical response in terms of oedometric compressibility was qualitatively identical for the lowest values of initial mechanical bonding, which makes it difficult to quantify the amount of mechanical and hydraulic bonding simply from the compression response. The predicted response upon saturation can be significantly affected by the relative proportion of the two contributions, however. In order to study the response upon saturation, then, a series of four soaking tests has been performed, by imposing the same initial hydro-mechanical state to all tests. By contrast, different initial conditions have been imposed for the internal variables representing mechanical and hydraulic bonding contributions. Suction controlled wetting stages have been simulated under oedometric conditions and material stability has been checked on the basis of the theoretical approach previously presented. Figure 4a shows the evolution of the void ratio during the four saturation tests, while Figure 4b shows the evolution of det 1. It is readily apparent a transition from swelling to collapse for all the simulations which experienced yielding upon saturation. At low bonding percentages (test #1 and test #2), a sudden decrease in the instability index value is observed at yielding. The value of det 1 never vanishes, however, showing a fully controllable response of the material under suction control conditions. Similarly, no loss of controllability is obtained in the highest
detX1
Yielding upon saturation
test #4 (No yielding)
test #1 test #3
test #2
Softening due to bond degradation: sudden loss of controllability
Sr
Figure 4. Evolution of the void ratio (a) and of the instability index (b) during saturation for different bonding percentages.
level of mechanical bonding (test #4). In this case saturation does not produce any yielding during the suction reduction stage. The material therefore undergoes an elastic path and no instability is possible under constant external stresses. By contrast, an unstable response can be envisaged at a critical percentage of mechanical bonding (test #3). If, in fact, the material is characterized by an intermediate proportion of mechanical and hydraulic bonding, a sudden saturation induced instability can occur under constant vertical stress. For such critical conditions, indeed, mechanical and hydraulic bonding effects are collaborating to sustain the external state of stress. Saturation produces a weakening of the material that requires bonds to carry larger stresses. If the effect is such that yielding is produced, a relevant plastic softening of the bond structure can be developed, which results in the inability of the material to withstand the applied state of stress. This type of unstable response is predicted when two independent conditions are fulfilled. Firstly, mechanical bonding has to be large enough to play a dominant role on the hardening modulus, having then a softening
775
c
Softening requirement Yielding requirement
Range of critical bondig percentages pm0<pm_yielding
p >p
m0
m_softening
pm0
Figure 5. Critical range of bonding percentages able to satisfy the two necessary conditions for instability (i.e., softening at yielding and occurrence of yielding upon saturation).
behaviour at yielding. It can be proved that this occurs when the variable 1 m pm/s ps 0 (Nova 2003), i.e., when at yielding pm0 s ps0 /m. Secondly, mechanical bonding has to be small enough so that yielding can occur upon saturation, i.e., pm0 pmy, in which pmy is the minimum value at which yielding is achieved (Fig. 5). The example points out an interesting feature of the mechanical response predicted by the model. Even though neither relevant variations of volumetric compressibility nor instabilities were predicted upon oedometric loading, the resulting behaviour in terms of mechanical stability upon soaking completely changes. A critical bonding percentage can in fact be identified for a fixed initial state of the material. The range of critical mechanical bonding depends on the initial state of stress, the initial degree of saturation and on the constitutive parameters governing the hydro-mechanical hardening rules. These critical conditions are likely to be affected by a sudden instability upon saturation, in which the decrease of suction activates an abrupt mechanical softening. This type of prediction can be obtained irrespectively of hydraulic control conditions. In this case, in fact, the instability mode can be considered as an uncoupled mode, i.e., as a failure mode in which only mechanical variables increase dramatically. Therefore, the degree of saturation represents in this case only an external triggering factor of a purely mechanical instability which takes place because of a brittle response at yielding. 5 CONCLUSIONS
the onset of instability in unsaturated bonded geomaterials. The model is featured by the use of a modified version of the Bishops stress for unsaturated soils (average skeleton stress) and it is formulated in a coupled hydro-mechanical form. Generalized hardening laws are used in order to describe both the hydraulic and the mechanical bonding contributions. The model has been employed in the light of a theoretical approach able to capture soil instability phenomena in the unsaturated regime. The approach is based on the definition of two instability indices, the vanishing of which is related to the inception of a loss of controllability. These mathematical indices depend on both the constitutive matrices and the specific hydro-mechanical loading programme. When the unsaturated material is characterized by a non zero initial mechanical bonding, important conclusions arise from model predictions. No instabilities are in fact obtained under suction controlled saturation both at very low and very high levels of bonding between grains. On the contrary, a critical intermediate proportion between mechanical and hydraulic bonding is found at which instabilities are possible. This critical range can be identified by means of the mathematical expressions chosen for the hydro-mechanical hardening laws. The predicted instability takes place irrespectively of the saturation mode (i.e., even when suction is directly governed). The saturation path in this case represents a simple trigger of an intense material softening, which results in the inability of sustaining the applied vertical stress, and which leads to a potentially catastrophic collapse even under oedometric conditions. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Buscarnera, G. & Nova, R. 2009a. An elastoplastic strainhardening model for soil behaviour allowing for hydraulic bonding-debonding effects. Int. J. Numer. Anal. Meth. Geomech. 33: 10551086. Buscarnera, G. & Nova, R. 2009b. Modelling instabilities in triaxial testing on unsaturated soil specimens. Published online on the Int. J. Numer. Anal. Meth. Geomech. D.O.I. 10.1002/nag.832. Buscarnera, G. & Nova, R. 2009c. Modelling the onset of instability in oedometric tests on unsaturated bonded soils. Proc. 1st COMGEO, S. Pietruszczak, G.N. Pande, C. Tamagnini & R. Wan eds: 226223. Garitte, B., Vaunat, J. & Gens, A. 2006. A constitutive model that incorporates the effect of suction in cemented geological materials. Proceedings of the Fourth International Conference on Unsaturated Soils, Miller G.A. et al (eds.); Carefree, Arizona: 19441955.
A constitutive law for unsaturated bonded soils has been formulated with the main goal of modelling
776
Gens, A. & Nova, R. 1993. Conceptual bases for a constitutive model for bonded soils and weak rocks. Int.Symp. Hard soils-soft rocks, Athens. Anagnostopoulos et al. eds, Balkema, Rotterdam: 485494. Gens, A., Sanchez, M. & Sheng, D. 2006. On constitutive modelling of unsaturated soils. Acta Geotechnica 1: 137147. Houlsby, G.T. 1997. The work input to an unsaturated granular material. Gotechnique 47(1): 193196. Imposimato, S. & Nova, R. 1998. An investigation on the uniqueness of the incremental response of elastoplastic models for virgin sand. Mech. of Cohes.-Frict. Mat. 3: 6587. Jommi, C. 2000. Remarks on the constitutive modelling of unsaturated soils. Experimental Evidence and Theoretical Approaches in Unsaturated Soils; Proceedings of the International Workshop on Unsaturated soils, Trento: 139153. Jommi, C. & di Prisco, C. 1994. Un semplice approccio teorico per la modellazione del comportamento meccanico dei terreni granulari parzialmente saturi. Atti Convegno sul Tema: Il Ruolo dei Fluidi nei Problemi di Ingegneria Geotecnica, Mondov: 167188. Koliji, A., Vulliet, L. & Laloui, L. 2008. Advanced constitutive model for unsaturated structured soil with double porosity. Proceedings of the 12th IACMAG conference. 16 October, Goa, India.
Leroueil, S. & Barbosa, A. 2000. Combined effect of fabric, bonding and partial saturation on yielding of soils. Proc. Asian Conf. On Unsaturated Soils: 527532. Leroueil, S. & Vaughan, P.R. 1990. The general and congruent effects of structure in natural soils and weak rocks. Gotechnique 40(3): 467488. Nova, R. 1994. Controllability of the incremental response of soil specimens subjected to arbitrary loading programmes. J. Mech. Behav. Mater. 5(2): 193201. Nova, R. 2003. The failure concept in soil mechanics revisited. Bifurcation and Instabilities in Geomechanics, J.F. Labuz & A. Drescher eds., Balkema, Lisse: 316. Nova, R., Castellanza, R. & Tamagnini, C. 2003. A constitutive model for bonded geomaterials subject to mechanical and/or chemical degradation. Int. J. Numer. Anal. Meth. Geomech. 27: 705732. Vaunat, J., Gens, A. & Pontes Filho, I.D.S. 2002. Application of Localization Concepts to Discontinuos Water Content Patterns in Unsaturated Media. Eighth International Symposium on Numerical Models in GeomechanicsNUMOG VIII, Roma. Lisse: Swets & Zeitlinger. p. 179184.
777
On the definition of constitutive models for compacted soft rocks

R. Cardoso
Instituto Superior Tcnico, Lisbon, Portugal
E.E. Alonso
Universitat Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: This paper presents some results of an experimental study performed to characterize the hydro-mechanical behaviour of compacted marls from Abadia, Portugal. The main purpose of the study was to understand the effects of the degradation of the fragments on the global behaviour of the aggregate. The knowledge earned allowed the identification of the main factors to be considered in the definition of constitutive models for compacted soft rocks. From a general point of view, traditional unsaturated constitutive models are not adequate to reproduce the large amplitude of the deformations observed in experimental tests on samples of compacted marls. Models for hard rockfill materials are not adequate either because they are not able to reproduce the behaviour of the material after the strong degradation of the fragments. The comparison between the model requirements and those of existing models for clayey materials and rockfill is performed. Suggestions are made on possible modifications of some known existing models from the literature. 1 INTRODUCTION strong degradation of the fragments. Suggestions are made on possible modifications of an existing model for double structure compacted materials. BExM is the model selected to reproduce the tests. This model was slightly modified following the suggestions presented. 2 EXPERIMENTAL EVIDENCE
It is often unavoidable to build embankments made of soft rocks despite of the evolutive nature of these materials. In fact, large fragments remain after compaction and their degradation when subjected to suction cycles irreversibly affects overall behaviour. The performance of these embankments cannot be suitably reproduced with existing models. An experimental study on the hydro-mechanical behaviour of compacted marls from Abadia, Portugal, was performed. The main purpose of the study was to understand the effects of the degradation of the fragments on the global behaviour of the mixture. It was observed that the behaviour is similar to the behaviour of rockfill while the fragments are intact (stiff fragments) but, when the mechanical properties of the fragments degrade due to stress or suction cycles, the material suffers strong structural changes and the granular structure progressively becomes similar to a compacted clayey material. Some experimental results showing this behaviour are presented in the paper. From a general point of view, traditional unsaturated constitutive models are not adequate to reproduce the large amplitude of the deformations observed in the compacted marls. On the other hand, models for hard rockfill materials are not adequate because they are not able to reproduce the behaviour of the material after the
The hydro-mechanical behaviour of compacted marls from Abadia formation in Portugal (upper Jurassic in age) was investigated. The marls were used in the construction of some embankments from A10 Motorway in Portugal (Arruda dos Vinhos). The behaviour of the AT1 embankment is reported in Cardoso (2009). Some characteristics of the material can be found in Cardoso and Alonso (2009a,b). Uniform grading size samples were prepared with fragments having dimensions varying between 9.5 mm and 4.75 mm (retained by ASTM sieve #4 and passing sieve #3/8). The samples were prepared with an initial void ratio of 1.078 0.005 and a water content of 14.5% 1%. The adoption of a relatively high void ratio was motivated by the need to ensure high air permeability to speed up the process of suction application using the vapour equilibrium technique. The drying effect was analysed by comparing the results of an oedometer loading test under a
779
constant suction s 2 MPa (test U) with two oedometer tests at higher applied suctions (test UD1 at s1 38 MPa and UD2 at s2 230 MPa). Two additional tests were performed to quantify the effect of a fully wetting-drying cycle before loading because major structural changes are observed after full saturation. They are tests UWD1 and UWD2, in which the samples, once compacted to the same initial state, were first fully saturated and then dried respectively to s1 38 MPa (sample UWD1) and s2 230 MPa (sample UWD2). The main results found in the oedometer tests performed are plotted in Figure 1. They show that drying has some effect in the structure of the compacted material because the compressibility of samples UD decreases with increasing suction. This result is typical of unsaturated compacted materials. The effect in the mechanical behaviour of wetting before drying (tests UWD) is studied by means of the comparison between tests UD and UWD for the suctions 38 MPa and 230 MPa. In the figure, wetting at v 50 kPa in tests UWD led to a strong collapse and therefore to a reduced void ratio. Subsequent drying did not modify the void ratio. The denser UWD specimens are, as expected, stiffer than samples UD. This is shown if Cc values are compared (Cc 0.422 and 0.379 for specimens UD1 and UD2, and 0.107 and 0.070 for specimens UWD1 and UWD2). Small differences were found in the elastic compressibility independently from the past suction history. Under 600 kPa, a larger collapse was observed for samples UD than for samples UWD, a result expected due to the different void ratios before full saturation. The final void ratios after full saturation are very close in all samples. The compressibility curves for all samples loaded under fully saturated conditions are very similar. The similarities found in final void ratios and in compressibility indicate that the fully saturated material has similar structure independently from suction history before full saturation.
1.10 1.00 0.90
Similar results were found in specimens prepared with initial void ratio achieved by changing the grading sizes of the fragments or the compaction energy. Samples U (e 1.078) and F (e 0.915) were prepared with the same compaction effort but adopting uniform grading size and a Fuller grading size distribution with five diameters, respectively. Sample H was prepared with the same uniform grading size adopted for sample U but with reduced void ratio (e 0.678) achieved by applying higher compaction effort. The compaction water content was identical in all of them. The photographs of the different samples after compaction are presented in Figure 2. The samples were fully saturated in oedometric cells under 9 kPa vertical stress (Fig. 3). Only the denser sample F exhibited swelling deformations. Therefore the collapse of samples U and F, having larger void ratios, can be interpreted to result from the swelling of the fragments invading the large pores between them. In Figure 3, swelling deformations measured in another sample prepared with uniform grading size distribution and with the same energy applied in the preparation of sample U but with the voids filled with fine fraction confirms this degradation mechanism. The specimens were loaded in oedometric cells after being fully saturated under low vertical stress. All specimens after full saturation show similar compressibility. This result indicates similar structures, which correspond to the more homogeneous material originated by the swelling of the fragments. Since the fragments do not recover their initial shape upon further drying, these volume
End of drying (shrinkage)
void ratio, e
0.80 0.70 0.60 0.50 0.40 0.30 1
Collapse due to full saturation End of drying (shrinkage) Full saturation
U UD1 UWD1 UD2 UWD2 US

10
Full saturation
Loading in fully saturated conditions

100 1000 10000
vertical stress (kPa)
Figure 1. Oedometric tests on uniform grading size samples U.
Figure 2. Specimens prepared with different void ratios: a) U (e 1.078); b) F (e 0.915); c) H (e 0.678).
780
1.5 1.0
vertical deformation (%)
0.5 0.0 -0.5 -1.0 -1.5 -2.0 -2.5 -3.0 0.00
Uniform grading size with the voids filled with fines
e=0.637 e=0.681
H: Uniform grading size heavy compacted F: Fuller grading size with 5D
e=0.971
U: Uniform grading size

e=1.078
0.01
0.10
1.00
10.00
100.00
time (h)
Figure 3. Imbibition tests under 9 kPa vertical stress for samples compacted with different grading sizes.
changes will be interpreted as plastic deformations and the designation fragments degradation is used in the remaining of the paper. The values of the intrinsic permeability were measured for all saturated samples under 100 kPa and are very similar (2.6 1013 m2 for specimens U and F and 2.2 1013 m2 for specimen H). This result can be explained by similar dimensions of the large voids, a result expected since the void ratios after full saturation are identical in all samples. These structural changes are confirmed in the analysis of the photographs of the samples with uniform grading sizes (named U) at the compaction water content under 1000 kPa, shown in Figure 4 (specimens U and US). After full saturation the fragments can hardly be distinguished; therefore the material is approaching a clayey matrix. Other cases are studied in Cardoso (2009). Irrespective of the degradation of the fragments when fully saturated, the compacted material, when dry, behaves as a rockfill. For the unsaturated samples a marked time-dependent behaviour was observed under each loading increment in oedometer tests performed under constant suction. Increasing slopes of the curves in a plot deformation dv vs. ln(time) were measured. The secondary compression index t (t dv/d(ln t)) was measured in each test loaded in unsaturated conditions (Cardoso & Alonso 2009b). The values measured for the vertical stress v 1000 kPa are in Figure 5. The value found for the saturated test is included in the figure. The peak reached before full saturation shown in Figure 4 marks the transition between rockfill behaviour (above a given suction) and clayey soil behaviour (below that suction). The threshold suction identified is the minimum suction studied (2 MPa) but further study is required to investigate this value. Figure 5 also includes the values of t measured in the unsaturated oedometer tests performed in compacted rockfill presented by Oldecop & Alonso
Figure 4. Macrostructure of the compacted sample under different suction cycles for the vertical stress of 1000 kPa: a) dry to s 230 MPa (e 0.978); b) fully saturated (e 0.505).
Clayey soil behaviour 0.30 0.25 0.20 Samples U and UD Pancrudo Rockfill Rockfill behaviour
v=1000kPa
t
0.15 0.10 0.05 0.00 0 50 100 150 200 250 300
total suction (MPa)
Figure 5. Values of t for samples UD and full saturated US: transition between rockfill and clayey soil behaviour.
(2007). The comparison shows that the peak is found at full saturation when the breakage is more intense. The comparison confirms that compacted marls experience a qualitative change in behaviour when wetted. Particle breakage dominates at high suctions. Regular soil deformation is found at lower suctions. The characterization tests performed by Cardoso (2009) on compacted samples of marl included the quantification of structural changes caused by different loading paths. The results of mercury intrusion porosimetry (MIP) tests (micropores) were considered jointly with data from the measurement of the macropores dimensions in
781
order to cover the large range of dimensions of the voids of the compacted material. The results confirm that compacted marl is a double structure material because a predominant size in the low diameters attributed to the microvoids and several peaks in the large diameter range attributed to the macropores. 3 CONSTITUTIVE MODEL
Fragments (fragments volume changes = microstructural volume changes)
(Microstructure=marl) Swelling em Collapse e pM Effect of degradation epb
Figure 6. Effect on global behaviour of fragments degradation due to full saturation (Cardoso 2009).
The two aspects of behaviour described for compacted marls are covered by constitutive models for rockfill and for regular compacted soils. Therefore it can be concluded that the mechanical behaviour of this material is some sort of combination of the mechanical behaviour of the two materials. They can be assumed to be the boundary cases, in terms of behaviour. The model must be able to simulate the deformations due to the degradation of the fragments under loading paths involving stress and suction changes because this is what makes the material different from the two limiting materials. From a general point of view, traditional unsaturated constitutive models are not adequate to reproduce the large amplitude of the deformations observed in the compacted marls caused by their degradation on wetting. On the other hand, models for hard rockfill materials are not adequate because they are not able to reproduce the behaviour of the material after the strong degradation of the fragments. These two reasons, added to the finding of the double structure detected in the material, lead to the conclusion that a model for regular compacted soils is probably adequate. For this reason, BExM proposed by Gens & Alonso (1992) for expansive clays is the model used as a starting point. It is assumed that the marl fragment is the microstructure and the set of fragments resulting from the compaction process is the macrostructure. The main assumption is that the compacted marls behave like an assembly of fragments to which is added the effect of their degradation, as illustrated in Figure 6. In this case, global irreversible volumetric deformations ep (Eq. 1) are the sum of the irreversible volumetric deformations of the fragments epb (debonding or loss of structure) and of the irreversible volumetric deformations of the macrostructural rearrangement epM. The new term epb can be seen as the contribution of the plastic deformations of the microstructure epm to the plastic deformations of the macrostructure. These deformations are not easily quantified: ep epM epb (1)
As shown in Figure 6, the consideration of the two independent causes of deformations allows understanding better the physical phenomena involved in the deformation process of the compacted material. An important simplification is to treat plastic deformations ep as only macrostructural plastic deformations epM,b instead of the existence of plastic deformations at the level of the microstructure (fragments degradation). The macrostructural deformations due to structural instability and fragments degradation (the two terms of Equation 1) are then converted into a single term. f e p M ,m e p e p M e p m em e e m e p m e e m e p m (2)
The adoption of a unique interaction function (Eq. 2) to represent the two effects of the microstructure volume changes on global behaviour is desirable because it is difficult to obtain experimental data to calibrate the irreversible effect of the plastic volumetric deformations of the fragments given by epm (also in Fig. 6). These deformations plus their effects on global arrangement are merged into the single term epM,m. The new definition of the interaction function therefore includes also information on the effect on global behaviour (on macrostructural arrangement) of the degradation of the fragments. The sign of the interaction function has physical meaning. Some examples are illustrated in Figure 7, where case 1 is the initial configuration, cases 2 and 3 illustrate wetting paths and cases 4, 5 and 6 represent drying paths. In case 2, the interaction function is positive if the fragments do no recover their initial volume and in case 3 the interaction function is negative. If the microstructural deformations are irreversible, resulting from degradation, the interaction function is negative. In case 4, the interaction function is negative. If global rearrangement becomes unstable this is case 5 and the interaction function is positive. In case 6, the interaction function is positive if the fragments do not recover their initial volume.
782
2.20 2.10 2.00 1.90 1.80 1.70 1.60 1.50 1.40 0 1 10 UD2-1 Model fitting
s S2=230MPa S1=39MPa S3=12MPa
UD2 UD2-1 UD2-3 p
100
1000
p (kPa)
Figure 7. functions.
Physical
meaning
of
the
interaction
a)
s S2=230MPa S1=39MPa S3=12MPa UWD2 UWD2-1 UWD2-3 p
The presence of the fragments was considered in the definition of the interaction functions. Because the fragments degrade (which explains the strong structural changes suffered by the compacted material) two interaction functions were necessary for each suction path. Two sets of functions were defined before and after being full saturated to account with the effects on the macrostructural arrangement of fragments degradation.
b)
2.20 2.10 2.00 1.90 1.80 1.70 1.60 1.50 1.40 1.30 1.20 0 1 10 100 UWD2-1 Model fitting
1000
p (kPa)
REPRODUCTION OF THE TESTS
As mentioned before, BExM was adapted to reproduce the behaviour of compacted marls. Oedometric tests on compacted specimens were conceived to find BBM parameters for the macrostructure. The tests were performed in uniform grading size samples type U previously described. Two sets of tests were performed, named UD and UWD (stress paths in Figures 8 and 9). The tests from set UD were performed on samples dry to s2 230 MPa following loading paths defined to find BBM parameters. Two of the samples were partially wetted under 600 kPa vertical stress to suctions s1 39 MPa and s3 12 MPa (specimens (UD21 and UD23, respectively). The tests also provided the experimental compressibility curves to be reproduced with the model. Maximum vertical stress reached in the oedometric loading process under constant suction was 1000 kPa. Then the samples were fully saturated after being unloaded to 600 kPa, and then reloaded to 1000 kPa and unloaded. The tests from set UWD are similar to tests UD but the samples were fully saturated before being dried to s2 230 MPa. Attention was taken in the calibration of parameters r and necessary for the definition of the elastoplastic compressibility index (s) for suction s and of the reference net mean stress pc to make them equal for the two sets of tests (Alonso et al., 1990). The adoption of a single set of parameters for the material before and after
Figure 8. Stress paths and reproduction of the tests using the properties of the remoulded material for the microstructure (Cardoso, 2009).
being fully saturated is justified because it is the same material. This is equivalent to consider that the material becomes denser after the degradation of the fragments. The values found are r 0.23, 0.02 MPa-1 and pc 250 kPa. The other BBM parameters adopted were measured in the set UD ((0) 0.200, M 0.006 and s 0.008). This set is chosen because it is when the material has the most intact structure since it is believed that the initial fragments arrangement is not much affected until the fragments begin to degrade due to full saturation. In a first attempt to reproduce the behaviour of the compacted material, the elastic model adopted for the microstructure was calibrated to reproduce the behaviour of the destructured material (Fig. 8) or of the rock (Fig. 9), the two boundary cases of the fragments in terms of structure. The differences between the two cases are in the values for the elastic compressibility index. For the destructured material it was used m 0.030 and for the rock it was used m 0.007. The values were measured in tests described in Cardoso (2009). In both cases, the effective stresses for the microstructure were defined using Equation 3, where p is the isotropic effective stress, s is suction, Sr
783
2.20 2.10 2.00 1.90 1.80 1.70 1.60 1.50 1.40 0 1 10 100 UD2-1 Model fitting
s S2=230MPa S1=39MPa S3=12MPa
UD2 UD2-1 UD2-3 p
Table 1. 2009).
Calibration of the transfer functions (Cardoso, 2nd Wetting
1st Drying 2nd Drying 1st Wetting
Elastic properties of the destructured material f0 0.5 0.9 1.8 4.0 f1 0.0 0.0 0.5 2.0 n 1.0 1.0 1.0 1.0
1000
p (kPa)
a)
s S2=230MPa S1=39MPa S3=12MPa UWD2 UWD2-1 UWD2-3 p
Elastic properties of the rock f0 0.0 3.0 5.0 f1 0.0 0.0 0.5 n 1.0 1.0 1.0
3.0 0.0 1.0
2.20 2.10 2.00 1.90 1.80 1.70 1.60 1.50 1.40 1.30 1.20 0 1 10 UWD2-1 Model fitting
100
1000
b)
p (kPa)
Figure 9. Stress paths and reproduction of the tests using the properties of the rock for the microstructure (Cardoso, 2009).
is the saturation degree and p is isotropic total stress. Parameter quantifies the immobile water within the soil that does not affect volume changes (Alonso et al. 2010). p p sSr (3)
In both cases the saturation degree was calculated with the water retention curve for the compacted material (first cycle) and 1 were used. This value of was used in the absence of data for better calibration this parameter for the compacted material. The constants f0, f1 and n of the interaction functions, given by Equation (4), were calibrated numerically. Two sets of functions were adopted, one to be used before full saturation (1st wetting and 1st drying) and the other to be used after full saturation (2nd wetting and 2nd drying). The values are in Table 1 for the two alternatives considered for the microstructure. p f p f0 f1 1 p0
n
(4)
For wetting paths, the constants of the transfer function have the same sign for the two models. Their signs followed the discussed in Figure 7. Larger absolute values are calculated when the rock elastic compressibility index is considered. This means that the microstructural deformations of the microstructure need larger correction for the rock, which may be explained by the smaller elastic compressibility for this material. The comparison between the numerical and experimental curves for the tests is given in Figure 8 (using the properties of the destructured material) and Figure 9 (using the properties of the rock) for the samples where the increment of vertical stress was applied under s1 39 MPa (specimens UD21 and UWD21). It can be observed that the main features of the behaviour of the compacted material are reproduced: shrinkage occurs on drying, collapse occurs in partial wetting and after full saturation. In both the shapes of the numerical and experimental curves are similar under the application of vertical stress in unsaturated conditions. A better curve fitting in test UD21 is observed for the model considering the parameters of the rock because the fragments were not completely degraded. Nevertheless, it can be seen the figures that the large collapse deformations observed in test UWD2-1 after the first full saturation (under 50 kPa) cannot be reproduced with any of the models. The error is compensated in the second full saturation (under 600 kPa) because the void ratio reached at the end of the simulation is similar to the value measured in the test. This correction is only possible because the second set of transfer functions is introduced. However this option reveals not to be able to reproduce the large collapse deformations caused by the degradation of the fragments therefore another alternative must be investigated.
784
CONCLUSIONS
This work presents the reproduction of the tests performed on compacted marls where large collapse deformations were observed due to the degradation of the fragments of rock. The double structure model BExM was used. The definition of the interaction functions was altered to enclose all information concerning the effects of the degradation of the fragments on the macrostructural arrangement. Two alternatives were studied for the calibration of the elastic constitutive model adopted for the microstructure. They correspond to the two boundary cases of the fragments in terms of structure. The results found indicate that the transfer functions by themselves are not able to reproduce the strong collapse deformations occurring in the first full saturation under small vertical stress. This indicates that the degradation of the fragments must be included explicitly in the model. As the fragments correspond to the microstructural level if a double structure framework is adopted for the compacted material, the results indicate that the key factor is the model adopted for the microstructure. This will be object of future work. ACKNOWLEDGEMENTS The authors would like to thank to the Portuguese Foundation for Science and Technology, FCT,
for the financial supports that allowed this study (SFRH/BD/25846/2005, POCTI/CM/59320/2004). Acknowledgement is also due to Professor Emanuel Maranha das Neves for his useful comments during the development of the research presented. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. (1990). A constitutive model for partially saturated soils. Gotechnique, 40(3), pp. 405430. Alonso, E.E., Pereira, J.-M., Vaunat, J. and Olivella, S. (2010). A microstructurally-based effective stress for unsaturated soils. Gotechnique, 60. (in print). Cardoso, R. (2009). Hydro-mechanical behaviour of compacted marls. PhD Thesis. High Technical Institute, Lisbon, Portugal. Cardoso, R. & Alonso, E.E. (2009a) Degradation of compacted marls: a microstructural investigation. Soils and Foundations, Vol. 49(3). Cardoso, R. & Alonso, E.E. (2009b). Structural changes of compacted marls subjected to suction cycles. Proc. Asian Pacific Conference on Unsaturated Soils, Australia. Gens A. & Alonso E.E. (1992). A framework for the behaviour of unsaturated expansive clays, Can Geotech. J., 29: 10131032. Oldecop, L.A. & Alonso, E.E. (2007). Theoretical investigation of the time-dependent behaviour of rockfill. Gotechnique, 57(3), pp. 289301. Van Genuchten, M.T. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, pp. 892898.
785
Effect of anisotropy on the prediction of unsaturated soil response under triaxial and oedometric conditions
F. DOnza, D. Gallipoli & S. Wheeler
Department of Civil Engineering, University of Glasgow, Glasgow, UK
ABSTRACT: The present work aims to investigate the influence of anisotropy on the prediction of unsaturated soil behaviour. The first part of the paper describes the improvement of an existing anisotropic elasto-plastic model for unsaturated soils. In the second part, the predictions of such novel model are compared with those of an isotropic elasto-plastic model and against experimental data for a number of triaxial and oedometric stress paths. In particular, elasto-plastic models for unsaturated soils can lead to incorrect predictions of the soil response during oedometric stress paths, especially if model parameter values have been selected on the basis of isotropic and triaxial tests. In this paper, particular attention has therefore been devoted to Ko conditions and to the investigation of the sensitivity of soil response to anisotropy during oedometric loading. Despite their experimental simplicity, oedometric tests are seldom used in the calibration of unsaturated constitutive models because of the uncertainties in the stress path resulting from the imposed radial constraints. Understanding the influence of anisotropy on the soil response during Ko loading is therefore an important step towards a wider use of oedometer tests during calibration of unsaturated elasto-plastic models. 1 INTRODUCTION (1990) and a variation of this model including an associated flow rule. All models have been calibrated by using tests performed by Barrera (2002) on an isotropically compacted silt. Experimental results from isotropic compression, shearing and Ko loading have then been compared against predictions from the anisotropic model with associate flow rule and from two versions of BBM, with associated and nonassociated flow rule respectively. The stress paths computed by the different models during a multistage oedometric test (including loading/unloading at constant suction and wetting/drying at constant vertical stress) have also been compared to assess the sensitivity of the predictions to anisotropy and the particular choice of flow rule. 2 2.1 ADOPTED MODELS The Barcelona Basic Model (BBM)
Some of the most common limitations of current elasto-plastic models for unsaturated soils include inaccuracy of the plastic flow rule, erroneous shape of yield locus and incorrect dependency of flow rule or yield locus on suction. These deficiencies are particularly important during Ko loading, when the development of radial stress is governed by the predicted stress-strain response and, in particular, by mechanical anisotropy caused by both initial fabric and subsequent loading history. During Ko loading, the particular choice of elastic law (e.g. the assumption of either a constant shear stiffness or a constant Poisson ratio) can also result in the prediction of different stress paths inside the yield locus and, hence, significantly different predictions of yield stresses. Some of the above issues are investigated in the present work through the simulation of a number of suction-controlled triaxial and oedometer tests by using different elasto-plastic unsaturated soil models that assume isotropy and anisotropy of mechanical behaviour. The anisotropic elasto-plastic model used in this work is based on the proposal by Stropeit et al. (2008) but includes improvements regarding the predicted shape of the yield locus at different suctions. Two isotropic elasto-plastic models have been used, namely the Barcelona Basic Model (BBM) by Alonso et al.
The BBM by Alonso et al. (1990) is a well-known constitutive model for unsaturated soils developed within the framework of classical elasto-plasticity. It is expressed in terms of net stresses, and suction, s and reverts to the Modified Cam Clay model at the saturated limit (i.e. when s 0). The model incorporates the influence of suction on yielding under general stress states and on shear strength. It also links volume changes and shearing
787
within a single elasto-plastic framework similar to Modified Cam Clay for saturated soils. One of the most important contributions of the BBM is the interpretation of plastic volumetric deformation during both loading and wetting/drying as a single yielding process explained by means of a unified modelling approach. The original BBM includes a non-associated flow rule to improve prediction of stress paths during Ko loading. In this work, two versions of BBM, with the original non-associated flow rule and an associated flow rule respectively, will be used for comparison. 2.2 The modified Anisotropic Barcelona Basic Model (ABBM1)
The ABBM by Stropeit et al. (2008) extends traditional elasto-plastic critical state models for unsaturated soils, such as BBM, to take into account the effect of anisotropy during plastic loading. Similarly to BBM, the ABBM is developed in the triaxial stress space in terms of net stresses, ua (where is the total stress and ua is the pore air pressure) and suction, s ua uw (where uw is the pore water pressure). The ABBM only models anisotropy of plastic strains while the elastic behaviour is isotropic and follows a formulation identical to BBM. The ABBM incorporates anisotropy by introducing an additional rotational component of hardening to describe development of soil fabric during plastic deformation. An initial version of the model limited to saturated soils (S-CLAY1) was proposed by Wheeler (1997) and subsequently modified by Ntnen et al. (1999). Wheeler et al. (2003) also gave a detailed description of S-CLAY1 together with its extension to general stress states. This work introduces some refinements to ABBM and the modified model presented here has therefore been named ABBM1. In ABBM1, the constant suction cross sections of the yield surface are expressed in the q : p plane as:
p q p 2 M 2 2 M f ( s ) pm ( s ) p M
ellipse with two vertical tangent points A and B (see Figure 1) falling on a line of gradient passing through the origin. Point B has abscissa coordinate equal to pm(s) while point A has abscissa coordiM f ( s ) (therefore depending on nate equal to M the current inclination of the yield locus). This is different from ABBM where point A has instead 3 f ( s ). This constant abscissa coordinate equal to 2 modification with respect to ABBM is necessary to ensure that, for equal to zero (i.e. corresponding to isotropic behaviour), the yield locus expression of ABBM1 reverts to that of BBM. Moreover, the yield locus expression adopted in ABBM1 (see Equation 1) implies that the intersection between the top of the constant suction ellipse and the critical state line has always a horizontal tangent regardless of the inclination . This is again different from ABBM where the intersection between the yield ellipse and the critical state line had a horizontal tangent only for the inclination corresponding to critical state. The size of the constant suction ellipse is governed by the hardening parameter pm(s), which depends on suction according to the LC yield curve expression of BBM: pm s pc p 0 ( s ) mc p
( 0 )
(2)
where pm(0) gives the size of the yield curve at s 0 and pc is a reference pressure (i.e. a soil constant). In Equation 2, the variation of the slope of the normal compression lines with suction, (s) follows the same expression as in BBM. Unlike the original BBM, an associated flow rule is assumed in ABBM1: d sp 2 d vp M 2 2 (3)
(1)
csl (s= s1) csl (s=0) 1 1
where M is the saturated critical stress ratio while and pm(s) govern the rotational and volumetric component of hardening respectively (i.e. they govern the inclination and size of the constant suction cross section of the yield surface), f(s) defines the increase of apparent cohesion with suction, q is the deviator stress and p is the mean net stress. Unlike pm(s), the parameter is assumed to be independent of the suction level. In ABBM1, the constant suction cross section of the yield surface has the form of a sheared
1 f ( s)
A
M f (s) M-
p 0(0) p m (0) p0(s) p m ( s ) p'
Figure 1. surface.
Constant suction cross sections of yield
788
where sp and vp are the plastic shear and volumetric strains respectively and is the gradient of the line from the vertical tangent point A to the current stress point, which in ABBM1 is equal to: M f (s) M M f (s) p M q
Table 1. Test SAT-1
Stress paths of tests in reference data set. Type iso Net stress Stage (kPa) e l/u Suction (kPa)
(4)
TISO-1 iso
Both volumetric and kinematic hardening laws are incorporated in the model. Volumetric hardening regulates the change in size of the yield surface depending on plastic volumetric strain: dpm 0 v d vp pm 0 0 (5)
e l w/d l/u w l/u
IS-OC-03 IS-NC-06 IS-NC-12 IS-OC-06
tx tx tx tx
where v is the specific volume. Kinematic hardening governs the change in inclination of the yield surface depending on both shear and volumetric plastic strains: 3 d 4 d vp b d sp 3 4 (6)
where parameters b and control the relative effectiveness of volumetric and deviatoric plastic strains in rotating the yield ellipse and the rate at which the inclination of the ellipse changes. Parameters b and , together with the starting value of the inclination , are the only three additional parameters in ABBM1 with respect to BBM. 3 EXPERIMENTAL DATA SET
e l s e l s e l s e l s e l w
IWS-OC-01 tx
IWS-NC-02 tx
The models described in the previous sections have been calibrated using tests performed by Barrera (2002) on samples of silt with sand and clay. All samples have been prepared by static isotropic compaction under a confining pressure of 600 kPa at water content of 11 0.2%. The suction, measured after compaction, was equal to 800 kPa. The experimental dataset, summarized in Table 1, consists of 1 isotropic saturated test (SAT-1) and 8 suction controlled tests, including 1 isotropic test (TISO-1), 6 triaxial tests (IS-OC-03; IS-NC-06; IS-NC-12; IS-OC-06; IWS-OC-01 and IWS-NC-02) and 1 oedometer test (EDO-1). During the isotropic stages of the triaxial tests and during the oedometric test, stress paths have been applied by step increments. In triaxial tests, samples have been consolidated prior to shearing under quasi-isotropic conditions at a nominal deviator stress of 10 kPa. Shearing has been performed at an axial strain rate of 1.0 m/min with at least one
EDO-1
s e l w s oed e B-C l C-D-E w/d E-F-G l/u G-H H-I-J w l/u
saturation at 10 kPa effective stress 10 1300 50 25 800 800 25 600 600 800 10 150 150 600 1400 600 600 150 20 20 600 2000 50 25 800 800 25 300 800 300 693 25 800 800 25 600 600 1183 800 25 800 800 25 1200 1200 2138 800 25 800 25 1600 800 600 600 1179 800 25 800 800 25 600 600 800 10 800 600 1212 800 25 800 800 25 600 600 800 20 20 600 997 25 800 800 25 600 600 800 50 300 300 600 1600 600 600 300 50 50 600 2400 50
iso isotropic, tx triaxial, oed oedometric, e equalization, l load, u unload, w wetting, d drying, s shearing.
unloading/reloading stage before reaching critical state (not shown in Table 1). A total of 6 suction levels have been explored (0, 20, 50, 150, 300 and 800 kPa) investigating both elastic and elasto-plastic behaviour. Mean net stresses up to 2000 kPa have been explored during the isotropic
789
tests and the quasi-isotropic consolidation stages of triaxial tests while shearing to critical state has been investigated at suction levels of 20 and 800 kPa. During the oedometric test, the sample has been loaded up to 2400 kPa of vertical net stress. 4 MODELS CALIBRATION
The equation of BBM describing the variation of isotropic yield stress with suction is recast in a linear form in the semi-logarithmic y : ln p0 plane: y
(0) * ln p0 ln p0 r ( 0 ) ( 0 )
(10)
The BBM parameters have been determined on the basis of the isotropic and triaxial compression tests described in the previous section. A sequential calibration procedure has been employed whereby parameter values have been selected one at a time in a specific order without making assumptions on the values of remaining parameters. The procedure is described in detail by Gallipoli et al. (submitted) and is very briefly summarized in the following. The five parameters governing isotropic plastic behaviour, namely , (0), r, pc, N(0) are first determined. The parameter is selected by fitting the relative spacing of the constant suction normal compression lines in the v* : ln p plane, where v* is the mapped specific volume defined as: v* v s ln s pat pat (7)
where suction s is replaced by the auxiliary variable y defined as: pc y s* ln p0 (11)
and pat is atmospheric pressure. The parameter is the single parameter that controls relative spacing between normal compression lines and its value is therefore selected without having to estimate the values of the other parameters (0), r, pc and N(0) at the same time. The equation of BBM describing the variation of the slopes of constant suction normal compression lines is recast in a linear form by replacing suction s with mapped suction s* defined as: s* 1 exp s (8)
* The initial value of p0 is then determined as the intercept, at y 0, of the best fit line of Equation 11 to the measured isotropic yield stresses plotted against the auxiliary variable y. The elastic compressibilities for changes of mean net stress and suction, i.e. and s, are determined from isotropic unloading/reloading and drying/wetting paths respectively, while the shear modulus, G is determined from unloading/reloading paths during constant suction shear stages. All BBM parameter values are summarized in Table 2. Given that the tested samples had been isotropically compacted, it is realistic to assume an initial null inclination of the yield locus in ABBM1 and, hence, an initial value of equal to 0. The ABBM1 reverts to BBM for equal to 0, thus the same values of , (0), r, pc, N(0) determined for the BBM can also be adopted for the ABBM1. In addition, since the initial value of is equal to 0 in ABBM1, the initial value of pm(0) coincides with * the initial value of p0 in BBM.
Table 2. Selected parameter values for BBM and ABBM1. BBM s G (MPa) N(0) (at s 0; p pc) (0) r (kPa 1) pc (kPa) k M b Initial value of p*(0)/pm(0) (kPa) Initial value of 0.0097 0.0045 167.3 0.779 0.0810 1.15 0.00164 9027853 0.421 1.18 250.6 ABBM1 0.0097 0.0045 167.3 0.779 0.0810 1.15 0.00164 9027853 0.421 1.18 0.0075 1 250.6 0
The parameters (0) and r are then selected by a linear fit of the experimental slopes of the constant suction normal compression lines. The value of pc is determined by fitting the absolute spacing between constant suction normal compression lines while the value of N(0) is fixed by matching the intersection point between normal compression lines. The two parameters, k and M, defining apparent cohesion and the slope of constant suction critical state lines in the q : p : s space, are simultaneously determined by fitting the critical state surface: q Mp M s to critical state points from the triaxial tests. (9)
790
The same function describing the increase in apparent cohesion with suction is assumed in ABBM1 as in BBM, i.e. f(s) ks. Therefore, the same critical state surface in the q : p : s space, with same values of k and M, is used in both models. The value of in ABBM1 has been determined using the empirical method suggested by Wheeler et al. (2003) as:
15 0
(12)
and, in the absence of direct evidence on the relative effectiveness of volumetric and deviatoric strains during kinematic hardening, a value of unity has been selected for parameter b. ABBM1 also assumes isotropic elastic behaviour with a formulation that is identical to BBM for increments of both volumetric and deviatoric strains. As a consequence, the values of , s and G for ABBM1 are the same as those specified for BBM. The selected parameter values for ABBM1 are summarized in Table 2. 5 PREDICTION OF SOIL BEHAVIOUR
deviator stress of 10 kPa. Nevertheless, the predictions from ABBM1 and BBM are undistinguishable confirming that this small deviator load is incapable of producing significant kinematic hardening in ABBM1 and, hence, significant rotation of the yield surface from its initial isotropic inclination. During the subsequent shearing, the two BBM versions, with associated and non-associate flow rules, predict the same evolution of volumetric strains with stresses due to their identical volumetric hardening rules. These predicted volumetric strains are significantly larger than experimental values as shown in Figure 4. The BBM with associated flow rule predicts larger shear strains, and hence larger axial strains, than the original BBM with non-associated flow rule. It is interesting to note that the larger plastic shear strains predicted by the associated flow rule in BBM, together with the above overestimation of plastic volumetric
1.70 1.65 1.60 1.55 1.50 1.45 1.40 10 100 1000 10000 simulations by all models experimental data
Figure 2 compares the predictions from ABBM1 and BBM with the experimental data in the v : lnp plane for test TISO-1 (see Table 1). As expected, during isotropic loading all models show identical predictions due to the initial null inclination of the yield locus (i.e. the initial value of is equal to 0). Predictions also appear to be in good agreement with experimental data. A similar comparison is shown in Figure 3 for test IS-NC-12 (see Table 1). In this case, the stress path during the consolidation stage is quasi-isotropic due to the application of a nominal
1.70 1.65
p' (kPa)
Figure 3. Simulations and experimental data of specific volume versus mean net stress for test IS-NC-12.
0.08
0.06
v
1.60 1.55 1.50 1.45 1.40 10 simulations by all models experimental data 100 1000 10000
0.04 BBM non associated BBM associated ABBM1 experimental data 0.00 0.04 0.08 0.12 0.16 0.20 0.24
0.02
0.00
s
Figure 4. Simulations and experimental data of volumetric strain versus shear strain for the shear stage in test IS-NC-12.
p' (kPa)
Figure 2. Simulations and experimental data of specific volume versus mean net stress for test TISO 1.
791
strains, result in an unexpected good match to the experimental axial deformations and, hence, in an accurate prediction of the stress-strain path in the q : a plane (see Figure 5). The combination of kinematic and volumetric hardening in ABBM1 produces more accurate predictions of plastic volumetric strains during shearing compared to BBM. The distance between the predicted constant suction normal compression lines and critical state lines in the v : lnp plane is smaller in ABBM1 than in BBM. Hence, the volumetric strain undergone by the soil during triaxial shearing from a normally consolidated condition is also smaller in ABBM1 than in BBM. This is demonstrated by the curve predicted by ABBM1 in the v : s plane, which is closer to the experimental data than the corresponding two curves predicted by BBM (see Figure 4). On the other hand, the kinematic hardening law and associated flow rule of ABBM1 are incapable of predicting plastic shear strains correctly and, hence, axial strains are significantly underestimated by this model (see Figure 5). Figure 6 compares the predictions by ABBM1 and the two versions of BBM with the experimental data in the v : ln v plane for test EDO-1 (see Table 1). The model predictions are all very similar and the two versions of BBM produce almost undistinguishable simulations lying slightly below the curve predicted by ABBM1. Inspection of Figure 6 seems to suggest that the choice of flow rule in BBM and the incorporation of material anisotropy in ABBM1 have no significant effect on the prediction of specific volume during oedometric tests. A different conclusion is obtained if predictions are analyzed in terms of computed stress paths. Figure 7 compares the stress paths predicted by all models in the q : p plane for tests EDO-1. Unlike isotropic or triaxial compression, the stress
3000 2400 1800 1200 600 0 0.00 0.04 0.08 0.12 BBM non associated BBM associated ABBM1 experimental data 0.16 0.20
1.70 1.65 1.60 1.55 1.50 1.45 1.40 10 100 1000 10000 BBM non associated BBM associated ABBM1 experimental data
v'(kPa)
Figure 6. Simulations and experimental data of specific volume versus vertical net stress for test EDO-1.
ax
Figure 5. Simulations and experimental data of deviator stress versus axial strain for shear stage in test IS-NC-12.
path during Ko loading depends on the choice of the constitutive model because the radial stress is computed by the constitutive model following the imposition of zero radial deformation. Note that, in Figure 7, there are not experimental data for comparison because radial stresses were not measured during experiments. The different stages during the oedometric test are denoted by letters in Figure 7 following the description given in Table 1. Inspection of Figure 7 indicates that the stress path predicted by the original non-associated version of BBM is almost coincident with the stress path predicted by ABBM1. On the other hand, the associated version of BBM produces a stress path that increasingly diverges from the other two as test progresses. During wetting/drying stages (C-D-E and G-H in Table 1), the stress paths predicted by all models have the same slope q/p 3/2. This can be understood by considering that, during wetting/drying stages, no changes of vertical stress take place but only radial stress varies. The same is not true for the loading/unloading stages (B-C, E-F-G and H-I-J in Table 1), where the slope of the stress path depends on the current stress state and the particular choice of flow rule. During the elastic part of an unloading/loading stage, the slope of the stress path depends on the current mean net stress because the Poisson ratio depends on the value of mean net stress due to the assumption of a constant shear modulus. In addition, during the plastic part of an unloading/loading stage, the slope of the stress path is influenced by the particular choice of flow rule. Note that the three predicted end points of the last wetting stage G-H (coinciding with the start points of the subsequent loading stage H-I) are far apart from each other but aligned on the same line of slope q/p 3/2. Interestingly, during the two plastic loading stages that yield on the wet side of the ellipse at
q (kPa)
792
1200
F
800 400 q (kPa) 0
C E B D G H J
0 200 400 600 800 1000 p' (kPa) 1200 1400 BBM non associated BBM associated ABBM1
-400 -800 -1200
1600
1800
2000
Figure 7.
Evolution of stress path in terms of deviator stress versus mean net stress during test EDO-1.
suctions of 300 kPa and 50 kPa respectively (i.e. stages E-F and H-I in Table 1), the stress paths predicted by both ABBM1 and the original nonassociated BBM tend towards the same approximately straight line. The non-associated flow rule in the original BBM is defined to ensure that the stress path attained by the soil during prolonged Ko loading is consistent with Jakys expression for the coefficient of pressure at rest. The fact that the stress path predicted by ABBM1 tends towards the same approximately straight line as the original non-associated BBM therefore suggests that the associated flow rule of ABBM1 is indeed be capable of capturing the correct stress path evolution during oedometric virgin loading. Finally, inspection of Figure 7 indicates that the stress path predicted by ABBM1 suddenly departs from the stress path predicted by the original nonassociated BBM towards the end of the last unloading stage I-J. This break is caused by the inclination of the yield ellipse in ABBM1, which produces earlier yielding in extension compared to the stress path predicted by the original non-associated BBM. 6 CONCLUSIONS
The paper presents an anisotropic elasto-plastic model for unsaturated soils, named ABBM1, which improves previous proposals by assuming a yield locus equation that reduces to that of BBM when soil fabric is isotropic. Moreover, regardless of the inclination of the constant suction yield ellipse in ABBM1, a horizontal tangent always occurs at the intersection between the critical state line and the ellipse. Simulations of multi-stage suction-controlled isotropic, triaxial and Ko compression tests have been performed by using two isotropic and one anisotropic elasto-plastic constitutive models. The
two isotropic models are the original BBM and a modified version of BBM with associated plastic flow rule. The anisotropic model is the ABBM1 proposed in this work. All models have been calibrated by using isotropic and triaxial compression tests performed by Barrera (2002) on isotropically compacted samples of silt with sand and clay. Predictions by the three models have been compared against experimental results from an isotropic compression test, a triaxial shearing test and an oedometric test from the same set of data. Results from the simulation of triaxial shearing indicate that ABBM1 is capable of predicting more accurate evolution of volumetric strains than BBM and, in particular, ABBM1 produces a smaller overestimate of the volumetric strain. This is because, in ABBM1, the distance between the predicted constant suction normal compression lines and critical state lines in the v : ln p plane is smaller than in BBM. On the other hand, the associated flow rule of ABBM1 tends to underestimate plastic shear strains, similarly to BBM. The simulation of the oedometer test show that predictions from all models are very similar in the v : ln p plane. Hence the choice of flow rule in BBM and the incorporation of material anisotropy in ABBM1 seem to have little effect on the prediction of specific volume. On the other hand, the chosen constitutive model influences noticeably the prediction of the stress path in the q : p plane. The stress path predicted by the original BBM is almost coincident with that predicted by ABBM1 while the stress path predicted by the version of BBM with associated flow rule increasingly diverges from the other two as the test progresses. Finally, the stress paths predicted by ABBM1 and the original BBM tend towards the same approximately straight line during plastic loading on the wet side of the constant suction
793
yield ellipse. Given that the original BBM includes a non-associated flow rule to ensure consistency of predictions with Jakys coefficient of pressure at rest, it appears that the assumption of an associated flow rule in ABBM1 is equally capable of capturing the correct evolution of the stress path during oedometric virgin loading. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40(3): 405430. Barrera Bucio, M. 2002. Estudio experimental del comportamiento hidro-mecnico de suelos colapsables. Ph.D. Thesis. Gallipoli, D., DOnza, F. & Wheeler, S.J.I. in press. A sequential method for selecting parameter values in the Barcelona Basic Model. Canadian Geotechnical Journal. Ntnen, A., Wheeler, S., Karstunen, M., & Lojander, M. 1999. Experimental investigation of an anisotropic hardening model for soft clays. In Proceedings of the 2nd International Symposium on Pre-failure Defor-
mation Characteristics of Geomaterials, Torino, Italy, 2830 September, Edited by M. Jamiolkowski, R. Lancellotta, and D. Lo Presti. A.A. Balkema, Rotterdam, 541548. Stropeit, K., Wheeler, S.J. & Cui, Y.-J. 2008. An anisotropic elasto-plastic model for unsaturated soils. Unsaturated Soils: Advances in Geo-Engineering, Proc. 1st European Conf. on Unsaturated Soils, Durham, UK, edited by Toll, D.G., Augarde, C.E., Gallipoli, D. & Wheeler, S.J., Leiden: CRC Press/ A.A. Balkema, Rotterdam, 625632. Wheeler, S.J. 1997. A rotational hardening elasto-plastic model for clays. In Proceedings of the 14th International Conference on Soil Mechanics and Foundation Engineering, Hamburg, A.A. Balkema, Rotterdam. 1: 431434. Wheeler, S.J., Ntnen, A., Karstunen, M. & Lojander, M. 2003. An anisotropic elastoplastic model for soft clays. Canadian Geotechnical Journal, 40(2): 403418.
794
Modelling criteria for a microbiologically stabilised compacted soil in the framework of elastoplasticity
G. Della Vecchia
Politecnico di Torino, Torino, Italy
L. Morales & E. Garzn C. Jommi
Universidad de Almera, Almera, Spain Politecnico di Milano, Milano, Italy
E. Romero
ABSTRACT: A novel stabilisation technique, based on a microbiological treatment, is being studied at the Universidad de Almera on different Iberian soils used in compacted linear earth constructions. The microbiological action results in the production of carbonate by bio-mineralisation, which should improve the mechanical properties of the original soil. To evaluate the advantages of the stabilisation procedure at the material level, laboratory tests were run on different soils, adding different amounts of microorganisms and at different as-compacted densities. With the aim of providing useful constitutive models for the stabilised soils in view of the field applications, an elastic-plastic framework is proposed, accounting for the unsaturated conditions characterising the earth construction, and relying on a suitable description of the effects of the microbiological treatment. Typical features of this approach are discussed by comparison with relevant experimental data coming from the ongoing experimental investigation. 1 INTRODUCTION the pore size distribution is reflected consistently on the water retention properties and on the compressibility of the compacted samples. In view of the field applications, the data were exploited to try to provide a unitary framework to develop useful, albeit simple, constitutive models for the stabilised soil, starting from the constitutive properties of its natural untreated counterpart. To this aim, a simple Modified Cam-Clay, extended to model the mechanical behaviour of the soil in unsaturated states (Jommi 2000), was interfaced to a recent advanced constitutive model for the retention domain (Della Vecchia 2009), accounting for soil microstructure evolution. The resulting coupled hydro-mechanical formulation was slightly modified to include the effects of calcite precipitation. In the following, the modelling criteria will be presented and discussed, and the resulting constitutive equations will be summarised. The modelling approach is evaluated by comparison against selected experimental data, describing the main features of the volumetric coupled hydro-mechanical behaviour of the natural and the stabilised soil.
A novel stabilisation technique, based on a microbiological treatment, is being studied for different Iberian soils used in compacted linear earth constructions. The microbiological action results in the production of calcite by bio-mineralisation, which should improve the mechanical properties of the original soil (see, e.g., Jimnez-Lpez et al. 2004). Besides the microbiological agent, the design parameters for the earth-work include soil type, carbonation elapsed time and compaction energy. To evaluate the advantages of the stabilisation procedure at the material level, laboratory tests have been run on different soils, treated with different amounts of microorganisms for different treatment time intervals, and compacted afterwards at different dry densities. The experimental data collected up to now suggest that precipitation of calcium carbonate from bacteria takes place in the pores which are slightly bigger than the characteristic dimension of the bacteria themselves. Comparison of hydro-mechanical data on the natural and the treated soils suggests that the change in
795
pore size density function (---)
MICROBIOLOGICAL TREATMENT AND EFFECTS ON THE HYDROMECHANICAL BEHAVIOUR
0.6
MIP B-5 MIP BT-5 MIP BT-5_R
3 m 30 m
Soil improvement via the proposed microbiological treatment relies on the capacity of the microorganisms to precipitate calcium carbonate (see, e.g., Boquet et al. 1973). In particular, ureolytic bacteria are able to influence the precipitation of calcium carbonate by the production of a urease enzyme. This enzyme catalyses the hydrolysis of urea 2 (CO(NH2)2) to CO3 and ammonium, resulting in an increase of the pH and carbonate concentration in the bacterial environment (Stocks-Fischer et al. 1999):
2 CO(NH2)2 2H2O 2 NH4 CO3 .
0.4
0.2
0 0.001
0.01 0.1 1 10 100 equivalent entrance diameter, D (m)
1000
(1)
Figure 1. Pore Size Distributions of natural and treated samples, from MIP data.
percentage of CaCO3 (%)
50 40 10 8 6 4
wL PI % CaCO3
In the presence of calcium ions, the solution can become supersaturated with respect to calcium carbonate leading to its precipitation:
wL, PI (%)
2 CO3 Ca2 CaCO3.
(2)
30 20 10 0 0
Precipitation of calcium carbonate crystals occurs by heterogeneous nucleation on the bacterial cell wall, once supersaturation is achieved. In the experimental programme performed at the Universidad de Almera, a microorganism of the Bacillaceae family was added to natural soil samples of a silty-clayey sand from the southeast of the Iberian Peninsula (B-5). Single (BT-5) and double doses of bacteria (BT-5_2D) were added to the original soil. Samples of natural and treated soils were then prepared by Proctor compaction with a constant amount of energy. The procedure resulted in samples having different initial void ratios. Few samples were further statically compacted again, to achieve a higher initial dry density (BT-5_R). All samples were prepared at similar water content, w0 0.15 0.02, following a standard practice in earth construction. A wide laboratory investigation allowed for studying the microstructure characteristics of the natural and the treated soil samples, together with their most relevant hydraulic and mechanical properties. As for the coupled hydromechanical volumetric behaviour, the main outcomes of the experimental investigation may be summarised starting from the pore size density function of the natural and the treated soil, which are compared in Fig. 1. The comparison between the data for natural soil and those for the treated soil shows that the appreciable effect of bacteria is to reduce the porosity in the limited range between 3 m 30 m, while the pores of larger dimensions are unaffected. The characteristic dimension of the last ones is a function of the compaction energy, hence of the
2 0 1 number of doses 2
Figure 2. Calcite content, liquid limit and plasticity index of the natural soil and of the soil after the microbiological treatment.
mechanical loading history. Consequently, at the scale of a laboratory sample, the influence of the treatment was appreciated on the characteristics mainly depending on the void size distribution, such as the plastic compressibility, and the soil retention properties, which will be discussed later with reference to the data presented in Sections 3 and 4. Besides, calcite precipitation modifies the plasticity characteristics of the soil, as the data in Fig. 2 show. Increasing the microorganism dose increases the percentage of CaCO3 in the soil. A tendency for both the liquid limit and the plasticity index to decrease, albeit non linearly, may be clearly appreciated from the data. 3 CONSTITUTIVE FRAMEWORK
Elasto-plasticity with generalised hardening was chosen to provide a convenient framework for the development of a constitutive model for unsaturated microbiologically treated soils. A simple Modified Cam-Clay was chosen for the mechanical behaviour of the soil. Among the different
796
possibilities to extend the model to unsaturated soil conditions, the guidelines suggested by Jommi (2000) were followed. A formulation recently proposed by Della Vecchia (2009) was adopted as the starting point for modelling the water retention behaviour. 3.1 Mechanical behaviour
The average soil skeleton stress, ij uaij Sr ij (ua uw) ij, together with suction, s ua uw, were adopted as stress variables. In the previous expressions, ij is the total stress tensor, uw and ua are the pressures of water and air, respectively, and ij is the Kronecker delta. The average soil skeleton stress is substituted for effective stress in the formulation of the mechanical constitutive equations and a reversible dependence of the preconsolidation pressure on the degree of saturation is introduced. With reference to axisymmetric stress variables, the yield surface and the associated plastic potential read: f g q 2 M 2 p p0 p 0 (3)
hydro-mechanical paths. At low water content, the influence of initial and current porosity is negligible. The relationship between the suction and the amount of water stored inside the aggregates depends mainly on the mineralogical composition of the clay and on the water chemical composition. In the inter-aggregate region a storage mechanism dependent on void ratio and void structure dominates. A conceptual separation of the amount of water in the intra-aggregate void space from that in the inter-aggregate void space can be achieved by an additive decomposition of water ratio ew w Gs Sr e, into intra- and inter-aggregate water ratios, ewm and ewM, ew Vw Vwm VwM ewm ewM , Vs Vs Vs (6)
where p0 is the isotropic hardening variable, M is the critical state stress ratio, p and q the constitutive triaxial stress variables. The preconsolidation pressure p0 evolution is ruled by both the degree of saturation,
* p0 1 b1 p0 b2 1 Sr 1 exp ,
where Vwm and VwM are the volume of water in the micro and macro pores, respectively, and Vs is the volume occupied by the solid particles. Micro- and macrovoids retention mechanisms may be described separately, yet guaranteeing continuity of the whole retention model and of its first derivatives, as proposed by Della Vecchia (2009). For water ratios lower than e* wm, that is the water ratio corresponding to completely dry macro-voids and saturated micro-voids, s b ln max * sm ew 1 , s b ln s ln max s* * sm m
* bewm
(4)
and the irreversible volumetric strain

* 0 p * 1 e p0 p vol ,
(7)

(5)
where and are the slope of the normal consolidation line and of the unloading-reloading lines p in the e log p plane, respectively, and vol is the volumetric plastic strain. The parameters b1 and b2 rule the shape of the projection of the yield surface on a p Sr plane. A simple linear elastic law can be adopted for the deviatoric elastic strains, with a constant shear modulus, G. 3.2 Retention domain
where s* m is the suction corresponding to e* wm, smax the maximum suction attainable, and b a parameter describing the average slope of the curve for high values of suction. The inter-aggregate retention mechanism depends on void ratio (Romero and Vaunat 2000, Kawai et al. 2000). For ew ranging between e and em, a suitable expression for the water ratio is s m ln 1 s 1 m ew em (e em ) 1 , (8) 1 s n ln 2 where m and n are independent parameters of the model, em is the current value of intra-aggregate void ratio and sm is the suction corresponding to em. Continuity between the two analytical expressions (7) and (8) for s sm, ew ewm, together with
The pore size distribution, together with its evolution along hydro-mechanical paths, rules the retention behaviour of porous materials. For compacted clayey soils, two distinct water retention regions may be recognised (see, e.g., Romero et al. 1999), respectively intra-aggregate and inter-aggregate. The two retention regions evolve in different ways along generalised
797
their first derivatives, provides a unique dependence of on the set of independent parameters, and naturally leads to a correct dependence of the air entry value on void ratio. Experimental data on compacted moderately active soils suggest that the intra-aggregate water ratio is not constant of a given soil. Instead, it evolves with water content, as a result of the swelling-shrinking tendency of the aggregates. The evolution of em may be described by a bilinear reversible envelope, as a function of two independent parameters, e* m and , depending on the consistency limits of the soil (Della Vecchia 2009),
* em em * for ew em ,
0.5
e*m, ebio, e*m+ ebio
e*m
0.4 0.3 0.2 0.1 0 3 4
ebio
e*m+ebio
percentage of CaCO3 (%)
10
Figure 3. Variation of microscopic pore space with calcite content at the end of the microbiological treatment.
0.016 0.012 0.008 0.004 0
(9a)
(-) oed (-)
* em .
em
* em +
* ew em
for ew
(9b)
The difference (ew em *) represents the maximum void space in which the aggregates may expand for a given void ratio, while em * represents the minimum intra-aggregate pore space, which is an increasing function of the soil plasticity indexes. The value of em * was calculated for the natural soil B-5 as the total void space having an equivalent diameter smaller than the limiting value which appears to be affected by mechanical loading. 3.3 Modelling the effect of calcite precipitation
B-5 BT-5 BT-5_2D BT-5_R
0.04
0.08
e bio (-)
0.12
0.16
0.2
Figure 4. Plastic compressibility from oedometer tests: experimental data plotted as a function of calculated eBIO.
Based on the results of the microstructure investigation, microbiological precipitation of calcite seems to produce two effects at the microscopic level. Firstly, it changes the pore size distribution due to microbiological calcite filling the pores having diameter in the range 3 m 30 m. Secondly, the liquid limit and the plasticity index of the soil decrease as the amount of microorganism increases. The voids in which calcite precipitates do not seem to be affected by changes in the total void ratio, as the comparison between the relevant pore size distributions suggests (Fig. 1). Under this hypothesis, precipitation of calcite produces a decrease of the inter-aggregate pore space, with respect to the untreated sample, of an amount eBIO, so that eM e em e BIO e em e BIO .
Besides, as the plasticity indexes of the soil decrease slightly at increasing calcite content, the * is expected to decrease consistently. value of em The PSD data confirm this idea, as the elaboration presented in Fig. 3 shows. The sum of that part of intra-aggregate pore space which directly depends on the soil plasticity indexes, em *, and of that part of the void space which is subtracted to the inter-aggregate space, eBIO, is assumed to define the current effective microscopic void ratio (Eq. 12a). Given the new definition, the previous water retention model (Eq. 7 and Eq. 8) may be adopted, with Eq. 12b, c defining the effective microscopic void ratio:
* * em em e BIO , * em em * for ew em
(12a) (12b) for ew

* em .
(10)
em
* em
* ew em
(12c)
The microbiologically affected pore volume may be estimated as: e BIO

30 m 3m
PSDnat PSDtreated dD.
(11)
It is interesting to note that oedometer experimental data highlight that the irreversible compressibility, measured by ( ), decreases with the amount of microorganisms introduced to stabilise the soil. The decrease seems to be well approximated by a linear function of eBIO, if samples
798
compacted at the same energy level are compared (Fig. 4). Increasing the compaction energy further reduces noticeably the soil compressibility. 4 EVALUATION OF THE CONSTITUTIVE APPROACH
The constitutive formulation was adopted to simulate the response of natural and treated samples in oedometer compression tests. To fully characterise the compacted samples, the water retention domain had to be calibrated first. To this aim, WP4 experimental data were exploited to calibrate the parameters for the original soil (B-5) and for the soil treated and compacted with higher energy (BT-5_R). The comparison between experimental data and the calibrated model is reported in Fig. 5. The calibrated parameters for the retention model are summarised in Table 1. It is worth noting
that, in the calibration, * m was assigned on the basis of the theoretical elaboration summarised in Fig. 3. All the other parameters were kept the same, but for the sample compacted at higher energy, which shows a different inter-aggregate pore structure. The wetting branches of the water retention domain predicted by the model for the original and the treated soil are compared in Fig. 6, for two different void ratios. The difference between the curves is only due to the different intra-aggregate pore space. 4.1 Oedometer test data After compaction, specimen of the different samples were loaded up to vertical stresses typical of the field application, flooded at constant applied stress, and further subjected to a loadingunloading cycle. The experimental tests showed that the amount of collapse induced by wetting was mainly a function of the initial void ratio, irrespective of the microbiological treatment. Nonetheless, the treatment reduced the plastic compliance, ( ), after collapse (Fig. 4). The initial void ratios and degrees of saturation of the different samples, together with the vertical net stress at wetting are reported in Table 2. The oedometer tests were numerically simulated with the constitutive approach suggested. The comparisons between numerical simulations and experimental data for the different samples are reported in Figs. 7, 8 and 9. They show that very good match can be obtained for both the original and the treated soil adopting the same constitutive framework, simply by a suitable calibration of few constitutive parameters.
1000
1000 B-5 exp B-5 sim BT-5_R exp BT-5_R sim
100
suction, s (MPa)
10
0.1 equivalent diameter 3 m - 30 m 0.01
0.001 0 0.2
degree of saturation, Sr (-)
0.4
0.6
0.8
1
100
suction, s (MPa)
Figure 5. Retention domain of the original soil B-5 and of the treated soil after heavy compaction BT-5_R. Table 1. samples. Retention parameters for the different soil Branch B-5 BT-5 BT-5_2D BT-5_R Drying Wetting Drying Wetting Drying Wetting Drying Wetting b 30 50 30 50 30 50 30 50 n 1.5 1.5 1.5 1.5 m 0.33 0.33 0.33 0.33 * m 0.230 0.295 0.290 0.295 s* m (MPa) 10 2 10 2 10 2 5 0.1
wetting branches B-5 BT-5 BT-5_2D
10
0.1 equivalent diameter 3 m - 30 m 0.01
e=0.5
0.001 0 0.2 0.4 0.6
e=0.7
water ratio, ew (-)
0.8
Figure 6. Wetting branches of the water retention domains of the original and the treated soil: numerical predictions for two different void ratios.
799
Table 2. Oedometer tests: initial conditions. vc (kPa) B-5 BT-5 BT-5_2D BT-5_R 50 100 50 50 100 125 300 e0 0.59 0.70 0.55 0.64 0.62 0.49 0.46 Sr0 0.58 0.52 0.81 0.69 0.72 0.59 0.57 p0 (kPa) 90 80 80 90 100 220 350
0.8 BT-5_R, vc=125 kPa 0.7 BT-5_R, vc=250 kPa numerical simulation
void ratio, e (-)
0.6
0.5
0.8
B-5, vc =50 kPa B-5, =100 kPa numerical simulation

c v
0.4
0.7
0.3 10
void ratio, e (-)
100
1000
10000
0.6
0.5
Figure 9. Comparison between experimental data (symbols) and numerical simulations of the oedometer tests on the treated samples compacted with high energy (BT-5_R). Table 3. Oedometer tests: variable constitutive parameters. b2 3.5 4.8 4.8 1.5
0.4
0.3 1
10
100
1000
10000
Figure 7. Comparison between experimental data (symbols) and numerical simulations of the oedometer tests on the original compacted soil (B-5).
0.8 BT-5 BT-5_2D, vc =50 kPa 0.7 BT-5_2D, vc =100 kPa numerical simulation
B-5 BT-5 BT-5_2D BT-5_R
0.090 0.085 0.080 0.060
0.6
0.5
0.4
Most of the mechanical parameters were kept constant for the different samples, namely the critical state stress ratio, M 1.46, the inclination of the unloading-reloading line, 0.0075, the shear modulus, G 15 MPa, and one of the parameters ruling saturation hardening, b1 0.3. Only the slope of the normal consolidation line, , and the parameter b2, ruling the sensitivity of the preconsolidation pressure to the degree of saturation, were changed in the different simulations. The values calibrated for the different samples are reported in Table. 3, while the initial pre-consolidation pressure, p0 , adopted in each simulation can be found in Table. 2. 5 FINAL REMARKS
void ratio, e (-)
0.3 1
10
100
1000
10000
Figure 8. Comparison between experimental data (symbols) and numerical simulations of the oedometer tests on the compacted samples microbiologically treated (BT-5 and BT-5_2D).
An elastic-plastic framework, originally conceived for compacted natural soils, was adapted to model the coupled hydro-mechanical behaviour of soils compacted after microbiological treatment. The approach relies on the experimental observation that calcite precipitation simply seems to modify
800
the inter-aggregate pore structure, but does not have any cementing effect. Although more experimental data are needed to confirm definitely this interpretation, a coupled hydro-mechanical model with an embedded microstructure measure, such as the one suggested here, proves to be a valuable simple choice to account for changes in the inter-aggregate microstructure due to the microbiological treatment, and, secondly, due to the compaction energy level. ACKNOWLEDGEMENTS This work was done in the framework of the BIOLIN project financed by ACCIONA Infraestructuras. The results presented herein are those of the authors only and they do not, necessarily, reflect the views of the company. Laura Morales acknowledges the exchange programme Estancias en Otros Centros financed by the Universidad de Almera for her stay at the Politecnico di Milano. REFERENCES
Boquet, E., Boronat, A. & Ramos-Cormenzana, A. 1973. Production of calcite (calcium carbonate) crystals by soil bacteria is a general phenomenon. Nature 246: 527529.
Della Vecchia, G. 2009. Coupled hydro-mechanical behaviour of compacted clayey soils. Phd Thesis, Politecnico di Milano. Jimnez-Lpez, C., Jroundi, F., Pascolini, C., RodriguezNavarro, C., Piar-Larrubia, G., Rodriguez-Gallego, M. & Gonzlez-Muoz, M.T. 2008. Consolidation of quarry calcarenite by calcium carbonate precipitation induced by bacteria activated among the microbiota inhabiting the stone. International Biodeterioration & Biodegradation 62: 352363. Jommi, C. 2000. Remarks on the constitutive modelling of unsaturated soils. Experimental Evidence and Theoretical Approaches in Unsaturated Soils. A. Tarantino and C. Mancuso (eds.): 139153. Rotterdam: Balkema. Kawai, K., Kato, S. & Karube, D. 2000. The model of water retention curve considering effects of void ratio. Unsaturated Soils for Asia. H. Rahardjo and E. Leong (eds): 329334. Rotterdam: Balkema. Romero, E., Gens, A. & Lloret, A. 1999. Water permeability, water retention and microstructure of unsaturated Boom clay. Engineering Geology, 54: 117127. Romero, E. & Jommi, C. 2008. An insight into the role of hydraulic history on the volume changes of anisotropic clayey soils, Water Res. Res,. 44, W12412: 116. doi:10.1029/2007 WR006558. Romero, E. & Vaunat, J. 2000. Retention curves of deformable clays. Experimental Evidence and Theoretical Approaches in Unsaturated Soils. A. Tarantino and C. Mancuso (eds.): 91106. Rotterdam: Balkema. Stocks-Fischer, S., Galinat, J.K. & Bang, S.S. 1999. Microbiological precipitation of CaCO3. Soil Biology and Biochemistry 31 (11): 15631571.
801
Some remarks on the hydro-mechanical constitutive modelling of natural and compacted Boom clay
G. Della Vecchia & C. Jommi A. Lima & E. Romero
Politecnico di Milano, Milano, Italy Universitat Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: Natural and compacted Boom clay have been extensively studied for their potential use as environmental barriers. In such applications, a series of coupled chemo-thermo-hydro-mechanical processes characterise the response of the soil, and advanced constitutive models are mandatory for a reliable description of its behaviour. The consequences of hydro-mechanical coupling are explored herein, from a constitutive point of view, starting from the analysis of few microstructure features of the natural and the compacted material, respectively. Numerical simulations of selected data are presented, coming from an extensive experimental programme performed at the Universitat Politcnica de Catalunya. A fully coupled hydro-mechanical constitutive model, conceived in the framework of elastoplasticity with generalised hardening, is adopted to model the behaviour of the natural and the compacted Boom clay. It is shown that the same constitutive framework can be adopted for the two materials provided the differences in the retention mechanisms and in the evolution of anisotropic fabric are correctly taken into account. 1 INTRODUCTION 2 2.1 TESTED MATERIAL Natural Boom clay
Natural and compacted Boom clay are the subject of many researches dealing with the coupled phenomena that may possibly affect the performance of this clay as natural or engineered environmental barrier (e.g. Romero et al. 1999, Delage et al. 2007). Such multi-phase systems are governed by a series coupled processes, involving mechanical, hydraulic, thermal and chemical interactions. Advanced constitutive models are mandatory for a comprehensive description of the behaviour of these systems. Here, the attention is restricted to the hydromechanical response of the natural and the compacted materials under generalised stress paths. In particular, the paper explores the consequences of hydro-mechanical coupling from the constitutive point of view, by presenting numerical simulations of selected tests on both natural and compacted Boom clay, coming from an extensive experimental programme performed at the Universitat Politcnica de Catalunya. Experimental data, integrated by fabric investigation by means of mercury intrusion porosimetry, are interpreted in the framework of classical elastoplasticity with generalised hardening. The attention is posed on the link between the microstructure features of the two materials and their consequences on phenomenological modelling.
An experimental study has been carried out on intact Boom clay specimens, collected at a depth of 230 m in the Underground Research Facility (URL) of Mol, Belgium. The in situ condition of the Tertiary clay at the depth of sampling is characterised by a vertical stress vsitu 4.5 MPa and a situ water pressure uw 2.25 MPa. From oedometer tests on the undisturbed material (Lima 2010) it is possible to estimate the maximum vertical effective stress max 5.2 MPa at the end of the deposition v process, before elastic unloading, diagenesis and creep (Delage et al. 2007), resulting in an overconsolidation ratio OCRsitu 2.3. Assuming a value of 20 for the friction angle of the natural material (Lima 2010), the earth pressure coefficient at rest K0 can be estimated:
dep NC K0 K0 1 sin 0.658 situ NC K0 K0 OCR situ 0.9978 1
(1a) (1b)
The preconsolidation pressure of the undisturbed soil and the in situ effective mean stress can be estimated to be pdep 4 MPa, and, psitu 2.25 MPa, respectively.
803
When the soil samples are retrieved from the ground, stress variation due to unloading can be situ quantified as p p situ psitu uw . Assuming that constant effective pressure, p 0, could be maintained throughout soil sampling, the corresponding variation in pore water pressure and the pore water pressure in the laboratory can be estilab mated to be uw p p 4.5 MPa and uw situ uw uw 2.25 MPa. Therefore, a value of suclab lab tion of about slab ua uw 2.25 MPa is expected for the samples in the laboratory. Natural soil specimens were trimmed in the laboratory from intact block samples. The initial state variables of two of the samples tested is summarised in Table 1. The estimated value of suction falls in the range between the two values measured in the laboratory. Specific gravity of Boom clay is Gs 2.7, its liquid and plastic limits are in the range wL 0.56 0.76 and wP 0.28 0.30, respectively (further information on natural Boom clay can be found in Lima (2010)). A description of the initial fabric of the natural material can be obtained by Mercury Intrusion Porosimetry (MIP) and Environmental Scanning Electron Microscopy (ESEM). A clear unimodal pore size distribution characterises the initial state of the material (Fig. 1). The ESEM images shown by Lima et al. (2008) confirm the absence of any distinction between different levels of porosity. In Figure 1, the pore size distributions of two further samples are shown. One refers to the end of a drying path from the undisturbed conditions,
Table 1. Sample 1 2
1
performed by controlling inflow of air at a given temperature and relative humidity. The second one was unloaded from the natural state by consolidation at pc 0.4 MPa in saturated conditions. Both hydro-mechanical paths influence the position and the value of the dominant pore size, but do not affect the topology of the void space, which remains clearly unimodal. The comparison allows excluding relevant variations in the fabric of the natural material in the stress range of interest for engineering applications. 2.2 Compacted Boom clay
The experimental tests on the compacted material were run on samples prepared by static compaction from Boom clay powder, at dry unit weight d 0 13.7 kN/m3 and water content w0 0.15, on the dry side of the optimum Proctor. The reference initial state of the compacted samples is summarised in Table 2, where e0, Sr0 and s0 are the initial value of void ratio, degree of saturation and matric suction respectively. The pore size density function of the ascompacted material is shown in Figure 2. It is characterised by two dominant pore modes, representative of the intra-aggregate porosity and of the inter-aggregate porosity (Romero et al. 1999). In the same figure, the pore size distributions of two other samples are shown. One refers to a sample saturated in oedometer at null applied vertical stress, and the second one describes a sample loaded to 1.2 MPa in oedometer, after saturation at null stress.
Table 2. w0 0.15
1.2
Initial state of the natural Boom clay samples. w0 0.28 0.24 e0 0.73 0.61 Sr0 1 1 s0 (MPa) 2. 3. Compacted Boom clay: as-compacted state. e0 0.97
as compacted saturated in oedometer saturated and loaded
Sr0 0.44
s0 (MPa) 1.9
pore size density funcion, e/logd (-)
0.8
1d drying
pore size density funcion, e/logd (-)
undisturbed p' =400 kPa c
0.8
0.6
0.4
0.4
0.2
0
10 0 10 1 10 2 10 3 10 4 10 5 10 6
0 10
0
10
10
10
10
10
10
pore equivalent diameter (nm)
pore equivalent diameter (nm)
Figure 1. Pore size distribution of natural Boom clay: undisturbed sample, sample after unloading in saturated conditions, sample after one dimensional drying.
Figure 2. Pore size distribution of compacted Boom clay: initial state of the as-compacted sample, after saturation, and after loading.
804
It is evident that different hydro-mechanical paths lead to different interactions between the two levels of structure in this case, contrarily to the natural material. A prevailing unimodal pore size distribution is observed after saturation, due to the increased size of the aggregates, which invade the macropores, being the size of the aggregates proportional to the water content of the sample. The peak of the resulting distribution is collocated between the two peaks characterising the ascompacted material. If the sample is loaded, a reduction of its macro porosity is observed, until a unimodal distribution results, with a dominant size in the intra-aggregate range. As evidenced by Romero et al. (1999), and more recently by Della Vecchia (2009), irreversible changes of the pore size distribution refer to variation of the pores between the aggregates. The behaviour of micro-pores seems to be reversible and uniquely dependent on the current water ratio, ew w Gs Sr e. 3 MODELLING THE HYDRO-MECHANICAL BEHAVIOUR
Table 3. Retention curve parameters for natural Boom clay (subscript d for drying and w for wetting). d (1/MPa) 0.15 nd 2.8 md 0.19 w (1/MPa) 0.50 nw 2 mw 0.22
100
suction, s (MPa)
10
1
model vapour equilibrium - wetting WP4 dewpointmeter - drying WP4 dewpointmeter - wetting
0.1 0 0.2
degree of saturation, S r (-)
0.4
0.6
0.8
Figure 3. Water retention domain for natural Boom clay.
3.1 Retention curve The pore size distribution, and its evolution along hydro-mechanical paths, rules the retention behaviour of a material. For natural Boom clay, mechanical and hydraulic paths do not seem to have significant effects on pore structure, which remains unimodal after all paths investigated, neither on the dominant pore size, which changes only slightly. The evidence suggests that the water retention behaviour of the natural material can be modelled by simply linking the degree of saturation to the suction, and neglecting its dependence on the evolution of other state variables. The classical Van Genuchten relationship can be used to the aim, 1 Sr , n 1 s
m
show distinct evolutions along hydro-mechanical paths, consistently with the porosimeter data presented in the previous section. At low water contents the influence of initial and current porosity is negligible, and the relationship between suction and the amount of water in the aggregates mainly depends on the mineralogical composition of the clay and on the water chemical composition. In the interaggregate region a storage mechanism, dependent on void ratio and void structure, dominates. To model this behaviour, a key point is the conceptual separation of the amount of water in the intraaggregate void space (micro pores) from that in the inter-aggregate void space (macro pores). A suitable choice is the definition of intra- and inter-aggregate water ratios, ewm and ewM, providing an additive decomposition of the total amount of water: ew Vw Vwm VwM ewm ewM Vs Vs Vs (3)
(2)
where two different sets of the material parameters , n, m are calibrated on the drying and the wetting branches (Tab. 3). Experimental water retention data for natural Boom clay, obtained by different suction control and measurement techniques, are presented in Figure 3, together with the calibrated model. For compacted clays, experimental evidences (e.g. Delage et al. 1996) suggest the presence of two distinct water retention regions, intra-aggregate and inter-aggregate, the latter corresponding to water contents sufficiently high to fill partially the voids between the aggregates. The two retention regions
In the previous relationship, Vwm and VwM are the volume of water in the micro and macro pores, respectively, and Vs is the volume occupied by the solid particles. Micro and macro voids retention mechanisms may be described separately, yet guaranteeing continuity of the whole retention model and of its first derivatives, as proposed by Della Vecchia (2009). An analytical expression, independent from the void ratio, may be assumed for water ratios lower than e* wm, that is the water ratio corresponding to completely dry macro voids and saturated micro voids,
805
where s* m is the suction corresponding to e* wm, smax is the maximum suction attainable, and b is a parameter describing the average slope of the curve for high values of suction. The dependence of the macroscopic part of the curve on void ratio may be achieved by adopting the linear scaling criterion proposed by Romero & Vaunat (2000), and adding a suitable continuous dependence of the air entry value on the void ratio (Kawai et al. 2000). For ew ranging from e to em, the water ratio reads
m ln 1 s sm 1 , (5) ew em (e em ) 1 ln 2 1 s n
suction, s (MPa)
smax b * * ln sm bewm ew 1 , smax s ln * b ln * sm sm

100
(4)
10
1 e=0.98 0.1 e=0.63
0.01 0 0.2
water ratio, ew (-)
0.4
0.6
0.8
Figure 4. Water retention domain for compacted Boom clay at two different void ratios (e 0.98, e 0.63).
3.2 Stress-strain constitutive model Elasto-plasticity with generalised hardening provides a sufficiently complete and general framework for the description of the coupled hydro-mechanical stress-strain response of both compacted and natural Boom clay. The average soil skeleton stress, ij ij uaij Sr (ua uw) ij, together with suction, s ua uw, can be adopted as stress variables. In the previous expressions, ij is the total stress tensor, uw and ua are the pressures of water and air, respectively and ij is the Kronecker delta. The anisotropic formulation proposed by Romero & Jommi (2008) is adopted here. With reference to axisymmetric stress variables, the yield surface and the associated plastic potential read:
2 f g ( q M p )2 ( M 2 M ) p( p0 p ) 0
where m and n are independent parameters of the model, em is the current value of intra-aggregate void ratio and sm is the suction corresponding to em. Imposing the two analytical expressions (4) and (5) to be continuous, together with their first derivatives, in s sm, ew ewm, gives a unique dependence of on the set of independent parameters, and naturally leads to a correct dependence of the air entry value on the void ratio. Finally, experimental data of compacted Boom clay suggest that the evolution of em may be described by a bilinear reversible envelope, as a function of two independent parameters, e* m and . The latter represents the swelling tendency of the aggregates (Della Vecchia 2009): em em
* em * em +
(7) where p0 and M are the isotropic and the rotational hardening variables, respectively, M is the critical state stress ratio, p and q are the constitutive triaxial stress variables. The preconsolidation pressure p0 evolves with the degree of saturation,
* p0 p0 1 b1 b2 1 Sr 1 exp ,
for ew
* (ew em )
* em ,
(6a)
* em .
for ew
(6b)
For compacted Boom Clay, 0.4 and e* m 0.4. Table 4 summarises the other parameters of the model, which were calibrated on the e 0.98 curve. The model prediction of the evolution of the water retention domain of compacted Boom clay with void ratio is shown in Figure 4, together with some experimental data (Romero et al. 1999).
Table 4. Retention curve parameters for compacted Boom clay. Branch Drying Wetting b 5 7.5 n 0.8 0.8 m 0.25 0.25 s* m (MPa) 5 2
(8)
and with irreversible volumetric strains,

* 0 p * (1 e ) p0 p vol ,
(9)
where and are the slopes of the normal consolidation line and of the unloading-reloading lines, p respectively, and vol is the volumetric plastic strain. Rotational hardening is governed by the difference
806
between the current stress ratio q / p and the current inclination of the yield surface M:
void ratio, e (-)
0.8
p c vol ( M ). M
(10)
0.7
The dimensionless parameter c governs the rate of evolution of M, and controls the target value of M for a given stress ratio. At the stress levels and at the void ratio of interest for compacted clays, for which the model was originally conceived, plastic strains induce significant variations of the anisotropic fabric of the material, and dominate the overall volumetric response. Therefore, elastic anisotropy was neglected. 3.2.1 Stress-strain response of natural Boom clay Due to its geological history, natural Boom clay has an initial inherent anisotropic structure, which can not be erased, unless very high stress levels, of the order of tens of MPa, are reached. A large part of the hydro-mechanical paths of interest are located inside the elastic domain. Elastic strains assume in this case a relevant role and elastic anisotropy should not be neglected. To account for this observation, a simple hyper-elastic anisotropic law was introduced in the model: p e vol J p (1 e ) e q J 3G d
0.6
0.5
experimental data numerical simulation
0.01
vertical stress, v (MPa)
0.1
10
Figure 5. Oedometer test on natural Boom clay: contact with water at 0.02 MPa.
0.8
experimental data numerical simulation

0.7
void ratio, e (-)
0.6
0.5
0.1
(11)
vertical stress, v (MPa)
10
Figure 6. Oedometer test on natural Boom clay: contact with water at 3. MPa.
Inherent anisotropy can be taken into account in the previous plastic framework by assuming constant not null values for M and J, and impeding rotational hardening, by assuming a value of c 0. Comparison between experimental data and numerical simulation of two oedometer test performed on natural Boom clay is shown in Figures 5 and 6. In the first test, the sample was posed in contact with synthetic Boom clay water at low vertical stress (0.02 MPa) and then subjected to a loadingunloading cycle up to a maximum vertical stress of 6 MPa. The evolution of void ratio with the vertical stress is shown in Figure 5. The model is able to simulate initial swelling of the sample, together with the subsequent accumulation of compressive plastic strain. The almost complete recovery of volumetric strain observed experimentally is well caught by the theoretical simulation. The numerical result is due to predicted negative volumetric plastic strain accumulated during the unloading stage, when the stress path reaches the inclined yield surface for negative values of the deviatoric stress. A second sample was compressed under undrained conditions up to a vertical stress of 3 MPa, and later posed in contact with synthetic Boom
clay water. The sample was then compressed up to 8 MPa, and finally unloaded (Fig. 6). The numerical simulation presented in Figure 6 shows the ability of the proposed framework to reproduce both the drained and the undrained behaviour of the material, together with the stress-state dependence of the volumetric response. Also in this case, the negative volumetric plastic strain developed during unloading causes large swelling, bringing the sample to a void ratio greater than the initial one, albeit smaller than that recorded in the experimental test. 3.2.2 Stress-strain response of the compacted clay Differently from the natural material, compacted Boom clay, starting from the induced initial anisotropy, shows a tendency to adapt more rapidly to the current stress direction at the stress level of interest. Romero & Jommi (2008) demonstrated that to model properly the experimental evidence, an evolution law for the rotational hardening, M, must be introduced (Eq. 10). Examples of the resulting behaviour of the compacted material in one-dimensional compression tests are reported in Figures 7 and 8.
807
Table 5. Parameters of the mechanical model for natural Boom clay. G J (MPa) c (MPa) M 0 3.3 b1 b2
0.175 0.075 57
0.78 1.48 0.10 2.00
Table 6. Parameters of the mechanical model for compacted Boom clay. G (MPa) c J (MPa) M b1 b2
0.125 0.014 40
1
120 0
0.87 1.48 0.54 5.25
0.085 MPa and 0.6 MPa, respectively, and finally subjected to a wettingdryingwetting cycle. The model correctly predicts the onset of volumetric plastic strain during first wetting, and the accumulation of further irreversible compressive volumetric strain during the subsequent dryingwetting cycle. The shape of the loading-collapse curve has a key influence in predicting such a response. The effects of the stress state on the amount of volumetric plastic strain during the first wetting and the subsequent drying-wetting cycle has been discussed in detail in Romero & Jommi (2008), as well as the influence of the evolution of anisotropy.
4
void ratio, e (-)
0.9
CONCLUSIONS
0.8
Experimental data Numerical simulation

0.7 0.01 0.1 1
suction, s (MPa)
Figure 7. Compacted Boom clay: wetting-dryingwetting cycle in oedometer at v 0.085 MPa.

1
Experimental data Numerical simulation
void ratio, e (-)
0.9
0.8
0.7 0.01
suction, s (MPa)
0.1
Figure 8. Compacted Boom clay: wetting-dryingwetting cycle in oedometer at v 0.600 MPa.
The comparison between numerical simulation and experimental data (Romero & Jommi 2008) along a wetting-drying-wetting cycle is reported in terms of the evolution of void ratio with suction. The two samples were one-dimensionally compacted at water content w 15%, and dry density d 1.37 Mg/m3, then unloaded, and mounted in an oedometer cell allowing control of suction. The samples were then reloaded at constant water content, up to final vertical net stress values of
The same constitutive framework seems to be applicable to natural Boom clay and to the material compacted starting from the clay powder, provided few changes which reflect the different micro-structure evolution observed along coupled hydro-mechanical paths. Pore topology and evolution of anisotropy mostly distinguish natural from compacted Boom clay. Natural Boom clay has a clear unimodal pore size density function, which is not affected much by drying-wetting cycles, neither by mechanical loading. On the contrary, compacted Boom clay is characterised by a clear bimodal pore size distribution. The macropores suffer significant irreversible evolution promoted by plastic volumetric strain. Micropores evolve, almost reversibly, in dryingwetting cycles (Della Vecchia 2009). The two distinct evolution mechanisms observed in compacted Boom clay do not permit normalisation of its water retention domain by means of the degree of saturation alone, which, on the contrary, seems sufficient for a reliable description of the water retention domain of the natural clay. Irreversible strains due to stress changes rule a clear evolution of the initial anisotropy of the compacted soil, at the stress level of interest. The anisotropic structure of natural Boom clay, due to its geological history, does not evolve easily, so that it can be considered as inherent, unless very high stresses are applied.
ACKNOWLEDGMENTS Travel expenses of A. Lima for the development of the present work have been funded by the Spanish Ministry of Science and Innovation, in the frame of the Acciones Integradas Hispano-Italiana (Ref. HI2000038).
808
REFERENCES
Delage, P., Audiguier, M., Cui, Y.J. & Howat, M. 1996. Microstructure of a compacted silt. Canadian Geotechnical Journal 33: 150158. Delage, P., Le, T.T., Tang, A.M., Cui, Y.J. & Li, X.L. 2007. Suction effects in deep Boom Clay block samples. Gotechnique 57(1): 239244. Della Vecchia, G. 2009. Coupled hydro-mechanical behaviour of compacted clayey soils. Phd Thesis, Politecnico di Milano. Kawai, K., Kato, S. & Karube, D. 2000. The model of water retention curve considering effects of void ratio. In H. Rahardjo & E. Leong (eds), Unsaturated Soils for Asia: 329334. Rotterdam: Balkema. Lima, A. 2010. Thermo-hydro-mechanical behaviour of natural Boom clay: an experimental study. Phd Thesis, Universitat Politcnica de Catalunya (in preparation).
Lima, A., Romero, E. & Pineda, J.A. 2008. Low-strain shear modulus dependence on water content of a natural stiff clay. In Associao Brasileira de Mecnica dos Solos e Engenharia Geotcnica (ed.), XIV Congresso Brasileiro de Mecnica dos Solos e Engenharia Geotcnica, Bzios, Rio de Janeiro, Brazil: 17631768. Romero, E., Gens, A. & Lloret, A. 1999. Water permeability, water retention and microstructure of unsaturated Boom clay. Engineering Geology 54: 117127. Romero, E. & Jommi, C. 2008. An insight into the role of hydraulic history on the volume changes of anisotropic clayey soils. Water Res. Res. 44, doi:10.1029/2007WR006558. Romero, E. & Vaunat, J. 2000. Retention curves of deformable clays. In A. Tarantino & C. Mancuso (eds): Experimental Evidence and Theoretical Approaches in Unsaturated Soils: 91106. Rotterdam:Balkema.
809
Soil water characteristic curves and adsorption potential

S. Frydman & R. Baker
TechnionIsrael Institute of Technology, Haifa, Israel
ABSTRACT: In previous studies the authors suggested that in the low potential (high suction) range, clay water matrix potential is due, essentially, to adsorption and not capillary effects. In many soils, the adsorption potential appears to be dominated by van der Waal forces, which are related to the specific surface area (Sa) of the clay particles. A theoretical expression for clay soil-water characteristic curves (SWCC) was consequently developed for cases in which the osmotic potential is small relative to the matrix potential, and van der Waal forces are dominant. Use of this expression requires knowledge of Sa, which is not commonly available in standard geotechnical data. The present paper presents a relation between Sa and liquid limit, based on data from the literature, and uses this relation to compare theoretical SWCCs with measured curves presented by other researchers. 1 INTRODUCTION
Baker and Frydman (2009) and Frydman and Baker (2009) discussed the nature of water potential, , or suction, in unsaturated clays, and presented a simplified model for the SWCC in the range of potentials lower (i.e. suction higher) than the cavitation pressure of the water in the soil, uw-cav. The model is based on the notions of cavitation, adsorption, and a double porosity. The main simplifying assumptions of the model are: a) The choice of a particular geometric structure for the double porosity model (see Fig. 1). The micro-pores are assumed to be saturated pods with parallel arrangement of clay crystals, while the macro-pores have angular geometry and may contain both water and air. b) All the water in the pods is assumed to be adsorbed, and the adsorption mechanism is dominated by van der Waal forces. c) The water in a macro-pore is located mainly in the corners of the pore, and its potential, cp, is the result of capillarity. d) The proposed model does not explicitly consider the osmotic potential. A basic element in the review of Baker and Frydman (2009) was the consideration of the possibility of cavitation of the water in the macropores under the tensile stresses generated by the capillary potential. The cavitation process limits the magnitude of the tensile stresses in the macro pore water. On the basis of these assumptions, one may consider how the SWCC develops during a drying process, from an initial condition in which the potential, , is above the cavitation pressure, with water in both the pods and the macropores. This system is now subjected to drying by exposing it to an external water potential lower than uw-cav. At the
Figure 1. The double porosity model.
first stage of the drying process both the adsorption potential of the water in the pods, and the capillary potential of the water in the macro pores, cp, will be reduced. This process will continue until cp uw-cav. At this stage, cavitation of the capillary water will occur. The adsorption potential of the water in the pods will continue to decrease until it reaches equilibrium with the external solution, and as a result of the potential gradient between the capillary and pods water, the capillary water in the macro pores will migrate into the pods. At equilibrium, all the
811
capillary water will be adsorbed inside the pods, and the macro pores will be effectively empty. Now consider the effect of an external mechanical stress applied to the system. The pods are saturated, and therefore will move as rigid bodies, and the effect of the external load will merely reduce the volume of the empty macro-pores, without affecting the matrix potential, m, which, at this stage, is controlled by the adsorption potential in the pods. As a result, m f() (where is the water content) is independent of the dry density d or the degree of saturation S. Fig. 2 (Tarantino and De Col, 2008) shows that for a significant range of water contents of engineering interest, f(d). This form of the SWCC function indicates that adsorption potential is the dominant mechanism controlling m. It is suggested that this situation is a consequence of the limitation imposed on the capillary potential by the cavitation process. The idea that f(d) is associated with conditions in which the macro pores are effectively empty has been proposed previously (e.g. Romero, 1999; Villar, 2007), and Baker and Frydman (2009) contributed to this discussion by relating the start of the migration process to the cavitation phenomenon which occurs when the capillary potential reaches the limiting value of cp uw-cav. In view of the above discussion, the use of axis translation techniques (e.g. pressure plate device), which prevent onset of cavitation, appears to be unsuitable for the study of the mechanical behaviour of unsaturated soils. Consequently, in the present paper, whose aim is to compare measured and theoretical SWCCs, only curves obtained using techniques other than the pressure plate are considered. 2 THEORETICAL SWCC
adsorption potential, ad, of the water in the pods. ad consists of different physical components which have been discussed in the literature (e.g. Churaev and Derjaguin,1985; Derjaguin et al., 1987; Iwata et al., 1995; Tuller et al., 1999). Or and Tuller (1999) suggested that the adsorption potential, ad, is dominated by the van der Waal component, vdw. The justification for this assumption is tested in the present paper by the analysis of a series of experimental SWCC functions. Adsorptive potential, based on van der Waals interaction, was studied by Iwamatsu and Horii (1996). Tuller et al. (1999) utilized this work in order to establish the following approximate theoretical form for this potential:
m
A 6 h3
(1)
where A (Joules) is the Hamaker constant (negative), and h is the thickness of the adsorbed layers. For a soil water system, A is of the order of 1020 to 1019 J., and Or and Tuller (1999) recommended adopting a value of A 6 1020 J. Mitchell and Soga (2005) quoted values of 3.1 1020 J for kaolinite, 2.5 1020 J for illite and 2.2 1020 J for montmorillonite. Medout-Marere (2000) found values of 6.8 1020 J, 8.6 1020 J and 7.8 1020 J for kaolinite, illite and montmorillonite, respectively. Or and Tullers value of 6 1020 [J] is adopted in the present paper. In eq. (1) when h is expressed in [m] and A in [J], then the resulting potential, m, is given in [Pa]. Assuming a complete coverage of all the clay surface area by adsorbed water, the gravimetric water content can be evaluated by the expression:
In the present section, a theoretical SWCC is presented, corresponding to the range of potentials and water contents relevant to the condition in which all the water is adsorbed in the pods, and the macro-pores are essentially empty. As a result of this situation, the matrix potential, m, is equal to the
ad 100 Sa h w
(2)
where [%] is the gravimetric water content, Sa [m2/kg] is the specific surface area of the minerals, which is different for different clay minerals, and w 103 [kg/m3] is the mass density of water. Since water adsorbed on the external surface of the pods is considered as pod water, Sa is the total specific surface area, and not just the surface area internal to the pods. Eliminating h between eqs. (1) and (2), and solving the resulting equation for m results in: 10 5 Sa m
3
(3)
Figure 2. Compaction data for kaolin showing equipotential lines (after Tarantino and De Col, 2008).
where A has been taken as 6 1020 J, w 103 kg/m3 and m is given in Pa. Eq. (3) is a theoretical expression for a SWCC in the relevant potential range, assuming only van der Waal contribution to the adsorption potential, and full water coverage of the pod particles.
812
This expression is unique, in that it is a function of parameters {A, Sa, w} which have physical significance in terms of basic colloid physics, do not depend on empirical curve fitting parameters, and can be established directly independently of suction measurements. A procedure for establishing A is described by Medout-Marere, (2000), while several methods are used to measure Sa. From comparisons of some of these methods (e.g. Santamarina et al., 2002; Yukselen and Kaya, 2006; Arnepalli et al., 2008) it was found that while some methods (e.g the Brunauer, Emmett and Teller (BET) gas adsorption method) measure only external surface area, two measure total surface areathe EGME method based on adsorption of ethylene glycol monoethyl ether and the MB method based on absorption of methylene blue. Consequently, if the total Sa of the particles of a particular soil is known, a theoretical SWCC can be established for the low (sub-cavitation) potential range, based on eq. (3), if van der Waal forces are dominant in adsorption potential. Tuller and Or (2005) used an expression similar to eq. (3) to estimate Sa from the dry end of measured SWCC data. 3 SPECIFIC SURFACE AREA AND LIQUID LIMIT
Table 1. Source
Data relating Sa (Methylene Blue) and wL. wL Sa 71.7 101 127 156 183 212 188 171 144 164 196 47 152 792 365 64 176 350 278 79 179 10 49 6 12 12 61 122 47 18 243 242 647 37 149 330 120 181 104 239 140 121 172 164 157 120 172 168 221 110 128 67 268 191 174 156 wL 155 180 205 230 255 75 70 69 73 120 56 129 335 280 288 295 277 349 415 72 61 114 70 465 112 331 396 54 227 411 175 200 49 61 78 77 58 55 52 146 67 59 60 59 83 72 81 60 83 38 39 78 Sa 238 267 307 333 359 225 215 235 213 87 122 388 709 615 617 623 610 730 851 185 73 244 158 851 705 778 949 65 646 1058 400 685 143 158 205 210 225 84 86 365 190 223 211 136 149 67 203 115 209 115 85 164
Since Sa is not a commonly available soil parameter, an attempt has been made to consider possible correlations between it and other soil parameters. The best of these correlations was obtained with liquid limit, wL. Cases found in the literature in which wL and total Sa were reported using the MB or EGME procedure, are presented in Tables 1 and 2. Fig. 3 shows the relationship between wL and Sa. Referring to Table 1, it should be noted that Cokca (2002), Cokca and Birand (1993) and Meseina (2006) used the MB method to obtain methylene blue values (i.e the amount of methylene blue absorbed per gram of soil), and these values have been translated into Sa using an expression presented by Santamarina et al. (2002),
Figure 3. Sa (by MB and EGME procedures) versus wL.
42 48 71 93 112 135 Meisina (2006) 68 65 66 73 71 Sivapullaiah et al. (2008) 38 134 374 82 55 57 52 82 46 59 27 Yukselen & Kaya (2008) 58 30 52 42 58 62 45 25 Arnepalli et al. (2008) 97 61 145 Inparajah & Wong (2008) 60 110 150 Coksa & Birand (1993) 39 52 43 58 53 50 60 60 51 44 60 57 57 50 42 34 69 57 62 64
Cokca (2002)
813
Table 2. Source
Data relating Sa (EGME) and wL. wL Sa 281 128 332 15 26 341 534 534 637 767 25 30 49 23 38 158 224 106 30 37 26 135 149 70 35 76 190 540 484 120 135 310 327 69 180 560 373 289 303 304 458 139 222 369 90 wL 54 227 411 37 35 92 40 53 51 45 100 57 61 64 434 97 560 72 61 114 70 465 112 331 396 27 120 56 129 335 280 288 295 277 349 76 76 72 67 66 74 61 37 840 Sa 320 353 510 56 122 135 145 161 171 186 204 245 255 381 593 675 704 95 62 299 104 578 393 442 582 12 95 80 410 550 425 411 423 321 510 333 460 350 414 355 210 298 36 655
PREDICTED AND MEASURED SWCC
Arnepalli et al. (2008)
97 61 145 Cerato & Lutenegger 42 (2002) 70 202 400 142 519 130 25 40 55 Lutenegger & Cerato 60 (2001) 65 33 69 Yuklselen & Kaya 58 (2008) 30 52 42 58 62 45 25 Sivapullaiah et al. 38 (2008) 134 374 82 55 57 52 82 46 59 415 Andrejkovicova et al. 73 (2008) 70 64 72 80 Liu (2007) 53 64 81 32
A number of SWCCs presented in the literature, in which suction (negative potential) was controlled or measured by methods other than axis translation, are considered in this section. Fig. 4, showing theoretical and measured curves of negative potential versus w/Sa, for a number of soils in which Sa was measured by EGME, was presented by Frydman and Baker (2009). This figure indicated the overall compatibility between the theoretical and the measured SWCCs, with the possible exception of kaolin rich soils (labelled Ridley, and Tarantino in Fig. 4). Cokca (2002) tested mixtures of kaolinite (K) and bentonite (B) in proportions 100%K 0%B to 50%K 50%B, in increments of 5%. In addition to measuring wL and Sa (by MB proceduresee Table 1), he also measured total suction in compacted samples, wetted to different moisture contents, using thermocouple psychrometers. If the osmotic potential in these artificial soils is assumed to be insignificant, his data may be used, together with eq. (3), to compare the theoretical SWCC with measured curves in which the values of Sa are taken as those measured, or as those calculated from his specific correlation against wL, or from eq. (4). Fig. 5 shows the relationship between Sa and wL as found from Cokcas data (Table 1), as well as the relationship, eq. (4), based on the data from all the MB and EGME tests. It is seen that Cokcas data falls below eq.(4); this is a result of the fact that his data points are amongst the lowest ones in Fig. 3. On the basis of Kokcas measured Sa and wL pairs, the following relation is obtained: Sa 1.3 wL 32.2 (5)
Figure 6 shows the theoretical SWCC, together with measured curves (Potential versus w/Sa) based on Cokcas dataset I based on actual Sa measurements, set II based on Sa estimated from eq. (5), and set III based on Sa estimated from eq. (4). It is seen that all the points based on actual
and Chiappone et al. (2004). On the basis of Fig. 3, Sa may be estimated from wL using the best fit relation (r2 0.68): Sa 233.15 Ln(wL) 778.51 (4)
Figure 4. Negative potential versus w/Sa for soils with known values of Sa (after Frydman & Baker, 2009).
814
Figure 5. SSA (MB) versus wL from data of Cokca (2002).
Figure 7. Measured (with Sa estimated) and theoretical SWCCs from data of Melgarejo Corredor (2004).
Figure 6. Theoretical and measured SWCC from data of Cokca (2002).
Sa measurements, and Sa obtained from calibration of Sa to wL from Cokcas data (i.e. eq. 5) lie around the theoretical line. Use of eq. (4) for estimation of Sa results in points lying about a half an order below the theoretical line, but quite parallel to it. Melgarejo Corredor (2004) developed drying curves for 6 clay soils (Speswhite kaolin, Bogota campus clay, London Emankment underground clay, Brazil residual granodiorite clay, Weald clay from Gatwick, and Colombia granodiorite). Matrix suction was measured by filter paper and by the Imperial College tensiometer. Liquid limits varied between 45 and 98 corresponding to estimated Sa values (from eq. 4) between 116 and 227 m2/g. Measured and theoretical SWCCs are shown in Fig. 7, showing reasonable correspondence except in the very high negative potential range. Fig. 8 shows data from some additional reported cases in which Sa has been estimated from wL using eq. (4). The data used was taken from Krahn and Fredlund (1972), Parreira and Goncalves (2000), Stenke et al. (2006), and Villar (2007). The degree of agreement between the theoretical SWCC and points based on measured SWCC together with
Figure 8. Measured (with Sa estimated) and theoretical SWCCs based on data from various sources.
estimated Sa values appears reasonable overall, but does show some significant deviation. 5 CONCLUSIONS
Use of the theoretical expression for a SWCC in the sub-cavitation potential range (eq. 3), which is based on the assumption that the matrix potential is essentially adsorption potential and controlled by van der Waal forces, requires a knowledge of the specific surface area of the soil particles, Sa. As Sa is not commonly known, a relation between Sa and liquid limit, wL, has been developed on the basis of published data, and this has been used to compare measured SWCCs with the theoretical curve. There appears to be overall, reasonable agreement, although some of the data shows significant deviation. This could be due to various reasons, the most basic being that it may be that van der Waal forces to not always dominate the adsorption potential; Frydman and Baker (2009) found that this could be the case, for example,
815
in kaolin rich soils. Alternatively, the considerable dispersion in the SawL relation (Fig. 3) may lead to erroneous estimates of Sa. The results of Cokca (2002) (Fig. 6) illustrate this; when the Sa value actually measured by Cokca was used in the comparison, excellent compatibility resulted, while use of the value obtained from the Sa-wL relation (eq. 4) led to considerably inferior compatibility. It is also possible that some of the negative potential values used for developing the measured SWCCs were unreliable. Measurement of water potential, whether by filter paper, tensiometer or psychrometer, is extremely sensitive and can easily lead to erroneous values. The overall compatibility does, however, suggest that the basic mechanism responsible for suction in the sub-cavitation potential range is adsorption potential, with van der Waal forces usually playing a major role. REFERENCES
Andrejkovicova, S., Janotka, I. & Komadel, P. 2008. Evaluation of geotechnical properties of bentonite from Lieskovec deposit, Slovakia. Applied Clay Science. 38: 297303. Arnepalli, D.N., Shanthakumar, S., Hanumantha, B. & Singh, D.H. 2008. Comparison of methods for determining specific-surface area of fine-grained soils. Geotech. Geol. Eng. 26: 121132. Baker, R. & Frydman, S. 2009. Unsaturated soil mechanicscritical review of physical foundations. Engineering Geology 106(12): 2639. Cerato, A.B. & Lutenegger, A.J. 2002. Determination of surface area of fine-grained soils be the ethylene glycol monoethyl ether (EGME) method. Geotechnical Testing J. 25(3): 314320. Chiappone, A., Marello, S., Scavia, C. & Setti, M. 2004. Clay mineral characterization through the methylene blue test: comparison with other experimental techniques and applications of the method. Can. Geotech. J. 41: 11681178. Churaev, N.V. & Derjaguin, B.V. 1985. Inclusion of structural forces in the theory of stability of colloids and films. J. Colloid Interface Sci. 103(2):542553. Cokca, E. 2002. Relationship between methylene blue value, initial soil suction and swell percent of expansive soils. Turkish J. Eng. Env. Sci. 26: 521529. Cokca, E. & Birand, A. 1993. Determination of cation exchange capacity of clayey soils by the methylene blue test. Geotechnical Testing J. 16(4): 518524. Derjaguin, B.V., Churaev, N.V. & Muller, V.M. 1987. Surface Forces. Plenum, New York. Frydman, S. & Baker, R. 2009. Theoretical soil-water characteristic curves based on adsorption, cavitation, and a double porosity model. Intl. J. of Geomechanics, ASCE, in print. Inparajah, D. & Wong, R.C.K. 2008. 1-D consolidation characteristics of kaolinite-bentonite mixtures with different pore fluid salinity. GeoEdmonton08: 334338. Iwamatsu, M. & Horii, K. 1996. Capillary condensation and adhesion of two wetter surfaces. J. Colloid Interface Sci. 182: 400406.
Iwata, S., Tabuchi, T. & Warkentin, B.P. 1995. Soil-water interaction: Mechanisms and applications. 2nd Ed., Lavoisier, France. Krahn, J. & Fredlund, D. 1972. On total, matric and osmotic suction. J. Soil Sci. 114(5): 339348. Liu, N. 2007. Soil and site characterization using electromagnetic waves. PhD. Thesis, Virginia Tech. Lutenegger, A.J. & Cerato, A.B. 2001. Surface area and engineering properties of fine-grained soils. Proc., 15th Intl Conf. on SMFE, 1: 603606. Melgarejo Corredor, M.L. 2004. Laboratory and numerical investigations of soil retention curves. PhD thesis, Imperial College of Science, Technology and Medicine, London. Medout-Marere, V. 2000. A simple experimental way of measuring the Hamaker constant A11 of divided solids by immersion calorimetry in apolar liquids. J. Colloid Interface Sci. 228: 434437. Meisana, C. 2006. Characterization of weathered clayey soils responsible for shallow ladslides. Nat. Hazards Earth Syst. Sci. 6: 825838. Mitchell, J.K. & Soga, K. 2005. Fundamentals of soil behavior. 3rd Ed. Wiley, New York. Or, D. & Tuller, M. 1999. Liquid retention and interfacial area in variably saturated porous media: upscaling from single-pore to sample-scale model. Water Resources Research 35:35913605. Parreira, A.B. & Goncalves, R.F. 2000. The influence of moisture content and soil suction on the resilient modulus of a lateritic subgrade soil. Geoeng 2000, Melbourne, Australia. Romero, E. 1999. Characterization and thermo-hydro mechanical behavior of unsaturated Boom clay: An experimental study. PhD Thesis, Universsita Politecnica de Catalonia, Barcelona, Spain. Santamarina, J.C., Klein, K.A., Wang, Y.H. & Prencke, E. 2002. Specific surface: determination and relevance. Can. Geotech. J. 39: 233241. Sivapullaiah, P.V., Guru Prasad, B. & Allam, M.M. 2008. Methylene blue surface area method to correlate with specific soil properties. Geotechnical Testing J. 31(6): 503512. Stenke, F., Toll, D.G. & Gallipoli, D. 2006. Comparison of suction measurement techniques for three clayey soils. Proc., 4th International Conf. on Unsaturated Soils, Phoenix, USA, 14511461. Tarantino, A. & De Col, E. 2008. Compaction behavior of clay. Geotechnique 58(3):199213. Tuller, M., Or, D. & Dudley, L.M. 1999. Adsorption and capillary condensation in porous media: Liquid retention and interfacial configurations in angular pores. Water Resour. Res. 35: 19491964. Tuller, M. & Or, D. 2005. Water films and scaling of soil characteristic curves at low water contents. Water Resour. Res. 41(9). Villar, M.V. 2007. Water retention of two natural caompacted bentonites. Clays and Clay Minerals 55(3): 311322. Yukselen, Y. & Kaya, A. 2008. Suitability of the methylene blue test for surface area, cation exchange capacity and swell potential determination of clayey soils. Engineering Geology 102: 3845. Yukselen, Y. & Kaya, A. 2006. Comparison of methods for determining specific surface area of soils. J. Geotech & Geoenvir. Engng, ASCE. 132(7):931936.
816
Numerical simulations of the thermomechanical behavior of clays on volumetric and deviatoric paths
M.S. Ghembaza & Z. Bellia H. Trouzine
U D L, F S I, Laboratoire Gotechnique, Matriaux et Environnement, Sidi Bel Abbs, Algrie U D L, F S I, Laboratoire Gotechnique, Matriaux et Environnement, Sidi Bel Abbs, Algrie I M B, Laboratoire Sciences et Modlisation, France
ABSTRACT: The aim of this paper is to simulate the thermo-mechanical behavior of saturated clays. In the first part, we present the model of Cui et al. (2000) whose principle is based on the volumetric and deviatoric thermomechanical simulation, the cited model show an decrease or an increase in deviatoric strength with temperature, in triaxial drained and undrained conditions. According to the researchers, the history of thermomechanical consolidation and the nature of soil may provide some explanation for this, the phenomenon has been dealt in the literature by (Hueckel & Pellegrini 1991, Cui et al. 2000, Graham et al. 2001, Cekerevac and Laloui 2004, Ghembaza et al. 2007). In the second part of this work, the results of the model (Cui et al. 2000) have been compared with the experimental results of (Del Olmo et al. 1996, Cekerevac and Laloui 2004, Ghembaza et al. 2007, Uchaipichat and Khalili 2009). 1 INTRODUCTION 1992) extended the Cam Clay model to unsaturated soils and swelling unsaturated soils, respectively. A new elastoplastic model for saturated soils submitted to temperature changes is proposed by Cui et al. (2000), with particular attention given to volume change behavior and the effects of the overconsolidation ratio (OCR). The model is based on experimental data provided by various authors for different saturated soils. Additional elastoplastic mechanisms are proposed to improve the understanding and modeling of the thermomechanical behavior of fine-grained soils. In this paper, we summarize the principal bases of the model of Cui et al. (2000) and we discuss numerical simulations of the thermo-mechanical behavior tests carried out by (Del Olmo et al. 1996, Cekerevac and Laloui 2004, Ghembaza et al. 2007, Uchaipichat and khalili 2009), using this same model. 2 THE ELASTOPLASTIC MODEL OF CUI ET AL. (2000)
In the last few decades, the effects of temperature on soil properties have become one of the major interests in the field of geotechnical engineering. The main reason has been industrialization and the growing number of nuclear facilities for power generation. The thick formations of clay that provide a low permeability have been proposed to host the disposal of high-level nuclear waste, often at several hundred meters underground. The heat flow generated by nuclear waste or any other high temperature source, such as high voltage cables and heat storages, must be dissipated in the surrounding clay. This can significantly affect the physical, chemical and mechanical properties of clay, in other words, this can lead to leakage of radioactivity. Therefore, the effects of temperature on soil and in particular on clay properties and on mechanical behavior are of great importance. Since 1980 a great number of thermo-mechanical models have been developed by different researchers to simulate the soil behaviour at elevated temperature. They are mainly based on the elastoplastic modified model of Cam Clay (Roscoe & Burland 1968) and the critical state concept. New types of modified models of Cam Clay have been published in the literature (Hueckel and Borsetto 1990) proposed an extension to temperature, whereas (Alonso and al. 1990, Gens and Alonso
The use of mathematical and constitutive models has been developed by many researchers such as Hueckel & Baldi (1990) in order to investigate the thermo-mechanical behavior of soils. These models are mainly an extension of the Modified Cam Clay model proposed by Roscoe and Burland
817
(1968). The Cam Clay model has been widely used for clay by many geotechnical experts due to its simplicity and because all parameters needed for modeling can be easily obtained through conventional triaxial tests in the laboratory. Cui and al.s model has been used for predictions of soil behavior during isotropic thermo-mechanical consolidation and the Cam Clay model has been employed for predictions of triaxial stress states (deviatoric loading). This model provides a new volumetric thermal plastic mechanism for predictions of plastic strains at higher OCR and over consolidation effects due to heating. The model has some additions to the extended Cam Clay model proposed by Hueckel & Baldi (1990) and is based on the following assumptions: the slope of the NCL, is independent of temperature. Heating may cause over consolidation in NC samples under constant load. The plastic contraction of NC samples due to heating is gradually converted to expansion at higher OCR. The applied load has no effect on the irreversible thermally induced volume changes of normally consolidated (NC) samples. The slope of unloading line () and the critical state parameter are constants and independent of temperature. Two volumetric plastic mechanisms are proposed by Cui et al. (2000). The first mechanism concern the elastic zone called LY (Loading Yield), defined by the exponential expression given by Hueckel & Borsetto (1990). pcT pco exp( o T ) (1)
higher OCR. This locus is called the HC curve and is defined by the following equation: p c1 pco exp(c2 T ) (3)
Where: c1 is the intersection of the HC curve with the p axis and c2 is a shape parameter. When the TY yield curve is reached during a heating test under a constant load, the thermo-plastic volumetric strains is defined by the following equation:
p vT exp( p T ) a p T 1
(4)
After derivation:
p d vT p [exp( p T ) a ] dT
(5)
With: a is the shape parameter close to unity. p is constant for normally consolidated soil (Demars and Charles 1982) and independent from the applied stress. For OCR 1/c1, p varies with OCR.When the TY yield curve is reached during a loading test under a constant temperature, a mechanical plastic strain must be defined by the following equation:
p d vTp 1
dp p
(6)
pco is the preconsolidation pressure and hardening parameter, the parameter o governs the shape of LY. The second mechanism is related to thermal hardening. This plastic mechanism, aimed at improving the prediction of the effects of OCR on the volume changes. The second yield locus, called thermal yield (TY), is activated by temperature increases in the range of higher OCR. The shape of TY is based on the experimental results of (Baldi et al. 1991). The equation of the TY yield locus is as follows: TcT (Tc To ) pco exp( p ) To (2)
In the absence of experimental data Cui et al. (2000) suggested that if the LY yield curve is reached by heating at constant load, the observed thermal plastic strain dp vpT would be similar to the plastic thermal strain on TY yield curve, dp vT. When the loading path crosses the LY curve, the plastic volumetric strain due to mechanical loading at constant temperature, dp vp can be evaluated by:
p d vp
dpcT pcT
(7)
The hardening rule is governed by two parameters, and pco . The movements of TY and LY yield curves are defined through the following hardening laws: For TY curve: d exp( p ) p d vT pTc To p [exp( p T ) a ]
Tc is a reference temperature corresponding to the intersection of TY curve with the Temperature axis, is a hardening parameter. Cui et al. (2000) have also defined another locus to implement the transition between expansion and contraction at
(8)
p d vTp 1
818
For LY curve: dpc 0 0 p d vpT pc 0 P (exp( p T ) a ) v p p 0 (d vp d vTp ) 0 v p k 0 d vT ( exp( T ) a ) P p
100
T [C] OCR=6
OCR=2
OCR=1
80
60
(9)
40
Exp , (Del Olmo and al. 19 9 6 ) Simul
The extension Cui and als model, to the triaxial state of stress is based on the critical state and the modified Cam Clay model Roscoe & Burland (1968). The yield locus in the qp plane is an ellipse and is defined by: f q2 M 2 p p pcT 0 The associated flow rule is defined by:
2 d vp M 2 p pcT p 2 q ds
VT%
20 -0.4
0.0
0.4
0.8
1.2
1.6
2.0
Figure 1. Volumetric strain simulation during cyclic heating on Spanish clay.
(10) probably due to the effect of OCR on the coefficient of thermal expansion. (11) 3.2 Simulation of CM clay (Cekerevac and Laloui 2004)
NUMERICAL VOLUMETRIC AND DEVIATORIC BEHAVIOR SIMULATIONS
To apply Cui et al. (2000) model, we simulated the volumetric behavior in non-isothermal conditions and the deviatoric behavior in isothermal conditions of saturated soils. The used model was validated using the experimental test results for Spanish clay (Del Olmo et al. 1996), CM (kaolin) clay Cekerevac & Laloui (2004), Sandy clay (Ghembaza et al. 2007) and Compacted Silt (Uchaipichat and khalili (2009). 3.1 Simulation of Spanish clay (Del Olmo et al. 1996) Consider the tests (Del Olmo et al. 1996), the sample was loaded to 6 MPa, maintained at 6 MPa (OCR 1), unloaded at 3 MPa (OCR 2) and at 1 MPa (OCR 6), and afterward subjected to a temperature cycle 229522C. The volumetric parameters are: pc 6 MPa, 0 3.93 103 1/C, 2 7.9104 106 1/C, a 0.9969, c1 0.508, c2 8.71 103 1/C, Tc 2000C, k 0. The result of simulation and experimental data is represented in Figure 1. In range of OCR 1 and 2, the model of Cui et al., (2000) presents a very good simulation. For OCR 6, there is a difference between simulation and experiment, this is
The thermo-mechanical volumetric simulation of CM clay is shown in Figure2. The samples underwent a heating path in drained conditions. After mechanical consolidation up to a mean effective stress of 600 kPa, unloading to different values of overconsolidation ratio (OCR 1, 1.5, 2.0, 6.0 and 12). The volumetric parameters are: pc 600 kPa, 0 7.6 104 1/C, 2 7.246 105 1/C, a 0.9611, c1 0.78, c2 0.07 1/C, Tc 200 C, k 0. The Cam Clay parameters are: For T 22C; 0.1, 0.026, * 0.0245, N 2.3274, v 0.25, M 0.80, 2.1092, * 2.2751. For T 90C; 0.1, 0.023, * 0.0245, N 2.2844, v 0.25, M 0.88, 2.0928, * 2.2321.
(*) Adjusted values (Ghahremannejad 2003). Figure 2 reveals that the thermal volumetric strain depends on stress history (OCR). Heating of the normally consolidated and lightly overconsolidated samples (NC and OCR 1.5) produced contraction. The sample (OCR 2) showed smaller thermal expansion. For OCR 6 and 12, the same phenomenon is observed in simulation of OCR 6 (Del Olmo et al. 1996), this confirms that OCR has a direct effect on the coefficient of
819
thermal expansion. However, the simulations of experimental data on drained triaxial shear test, at different temperatures (22 and 90C) are show in Figure3, they indicate increase of shear strength with temperature, these results reproduced very satisfactorily the experience, except the test of OCR 3 where simulation shows a highly overconsolidated behavior, while experience shows a
100 T [C] OCR=1 OCR=6 OCR=2 OCR=1.5
lightly overconsolidated behavior (Figure 3a). If we compare the experimental volume changes with those of the simulation (Figure 3b), no clear trend can be seen at different temperatures. 3.3 Simulation of sandy clay (Ghembaza et al. 2007)
80 OCR=1 60 OCR=6 40
Exp , Cekerevak & Lalo ui (2 0 0 4 ) Simul
An experimental study of thermal effects on mechanical behavior of remoulded sandy clay was performed by (Ghemabaza et al. 2007), the tests were performed under isothermal conditions. The Cam Clay parameters are: For T 22C; pc 900, 1250 kPa, 0.1, 0.0178, N 2.2899, v 0.25, M 1. For T 80C; pc 900, 1250 kPa, 0.1, 0.0178, N 2.2616, v 0.25, M 0.9. the numerical predictions and experimental data of undrained shear tests, normally consolidated at high temperatures are shown in Figure 4, the model predicted a small value of shear strength compared to
1200 1000 800 NC, pc=1,25MPa
20 -1.0
VT%
-0.6
-0.2
0.2
0.6
1.0
Figure 2. Thermal volumetric strain simulations during heating test (22 to 90C) on CM clay.
800 700 600 500 400 300 200 100 0 0 0 0 1 2 3 4 5 6 7 OCR=1
Simul-T2 2 Simul-T9 0 Cekerevak & Lalo ui (2 0 0 4 )-T2 2 Cekerevak & Lalo ui (2 0 0 4 )-T9 0 Simul-T2 2 Simul-T9 0 Cekerevak & Lalo ui (2 0 0 4 )-T2 2 Cekerevak & Lalo ui (2 0 0 4 )-T9 0
q [kPa]
OCR=1
q [kPa]
600 400 200 0
OCR=3
NC, pc=0,90MPa
Simul-T2 2 Simul-T8 0 (Ghemb aza and al. 2 0 0 7)-T2 2 (Ghemb aza and al. 2 0 0 7)-T8 0
(a)
5 5 10 10 15 15 20 25
1%
0 1000 800
10
15
20
25
1%
20 25
u [kPa]
NC, pc=1,25MPa
OCR=3
600 400 NC, pc=1,25MPa
NC, pc=0,90MPa
200
V%
(b)
0 0 5 10
Simul-T2 2 Simul-T8 0 (Ghemb aza and al. 2 0 0 7)-T2 2 (Ghemb aza and al. 2 0 0 7)-T8 0
NC, pc=0,90MPa
1%
15
20
25
Figure 3. Results of numerical simulations of triaxial tests on CM clay.
Figure 4. Results of numerical simulations of triaxial tests on sandy clay.
820
the experience, This can be explained either by, the thermal densification, or by the response of material that resembles to the overconsolidated sample type, So there may be a limitation of the Cui and als model to the overconsolidated behavior. 3.4 Simulation of compacted silt Uchaipichat and khalili (2009)
300 250 200 150 100 50 0 0 -2 0 0
q [kPa]
OCR=1.3
OCR=4
Simul-T25 Simul-T40 Simul-T60 Uchaipichat & Khalili (2009)-T25 Uchaipichat & Khalili (2009)-T40 Uchaipichat & Khalili (2009)-T60
The sample of compacted Silt undergoes a mechanical loading at different OCR, and submitted to the tests of heating-cooling. For more details concerning this test program see (Uchaipichat 2005, Uchaipichat & khalili 2009). The volumetric parameters are: pc 200 kPa, 0 3.42 104 1/C, 2 2.957 105 1/C, a 0.9899, c1 0.771, c2 0.017 1/C, Tc 150C, k 0. The Cam Clay parameters are: For T 25C; 0.09, 0.006, N 2.049, v 0.25, M 1.17, 1.997, * 1.991. For T 40C; 0.09, 0.006, N 2.044, v 0.25, M 1.17, 1.979, * 1.986. For T 60C; 0.09, 0.006, N 2.039, v 0.25, M 1.17, 1.968, * 1.981.
(*) Adjusted values (see Ghahremannejad (2003)). Figure 5 shows the volumetric simulation results, the phenomena of expansioncontraction in OC sample and contraction in the NC samples, are correctly reproduced.
10
15
20
s%
5 OCR=4 10 15
Simul-T25 Simul-T40 Simul-T60 Uchaipichat & Khalili (2009)-T25 Uchaipichat & Khalili (2009)-T40 Uchaipichat & Khalili (2009)-T60
20
OCR=1.3
V%
Figure 6. Results of numerical simulations of triaxial drained tests on compacted silt.
65
T [C] OCR=4
Exp , Uchaip ichat & Khalili (2 0 0 9 )
Simul
OCR=2
OCR=1.3
OCR=1
Moreover, other deviatoric simulations on drained oveconsolidated samples are realized. The results are shown on Figure 6 where computed stress-strain curves are plotted and compared with the experimental data. The overconsolidated samples (OCR 1.33 and 4) were sheared by keeping cell pressure at constant temperature, while increasing the deviator stress. The results of simulation reproduce the experience with satisfactory. They show a decrease of shear strength with temperature, for two values of OCR 1.33 and 4 (Fig. 6). 4 CONCLUSIONS
55
45
35
VT%
25 -0.20
0.00
0.20
0.40
0.60
0.80
Figure 5. Thermal volume change simulation under thermal cyclic loading on compacted silt.
The thermo-elasto-plastic model of Cui et al. (2000) was used for the prediction of volume changes of clays after applying the thermo-mechanical loading. In association with the modified Cam Clay model, the deviatoric response of clays was simulated. The predictions of the model are compared with experimental results at different simulations cases : thermo-mechanical consolidation, undrained and drained triaxial shear tests of normally (NC) and (OC) samples. The Cui and als model provided reasonable predictions of the ther-
821
mally induced volume changes for NC and lightly OC samples of clay during drained heating and subsequent cooling. In the case of higher OCR the model doesnt reproduce correctly the phenomenon of expansioncontraction; this is probably due to the effect of OCR on the coefficient of thermal expansion. Generally in saturated soils, application of modified Cam Clay model in thermo-mechanical conditions at higher OCR gives relatively poor predictions, because this model is insufficient to simulate this kind of behavior. PREFERENCES
Alonso, E.E., Gens A., & Josa A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40(3): 405430. Baldi, G., Hueckel, T., Peano, A., & Pellegrini, R. 1991. Developments in modelling of thermo-hydro-geomechanical behavior of boom clay and clay-based buffer materials. Commission of the European Communities, Nuclear Science and Technology, EUR 13365/1 and EUR 13365/2. Cekerevac, C., & Laloui, L. 2004. Experimental study of thermal effects on the mechanical behavior of a clay. International Journal for Numerical and Analytical Methods in Geomechanics 28: 209228. Cui, Y.J., Sultan, N., & Delage, P. 2000. A thermomechanical model for saturated clays. Canadian Geotechnical Journal, 37 (3): 607620. Del Olmo, C., Fioravante, V., Gera, F., Hueckel, T., Mayor, J.C. & Pellegrini, R. 1996. Thermomechanical properties of deep argillaceous formations. Engineering Geology 41: 87101.
Demars, K.R., & Charles, R.D. 1982. Soil volume changes induced by temperature cycling. Canadian Geotechnical Journal, 19: 188194. Gens, A., & Alonso, E.E. 1992. A framework for the behaviour of unsaturated expansive clays. Canadian Geotechnical Journal, 29: 10131032. Ghahremannejad, B. 2003. Thermo-mechanical behaviour of two reconstituted clays, Ph.D thesis, university of Sydney, Australia. Ghembaza, M.S., Tabi, S., & Fleureau, J.M. 2007. Effet de la temprature sur le comportement des sols non saturs sur les chemins de drainage and dhumidification. Canadian Geotechnical Journal, 44: 10641081. Hueckel, T., & Baldi, G. 1990. Thermoplasticity of saturated clays, experimental constitutive study. Journal of Geotechnical Engineering 116, No. 12: 17781796. Hueckel, T., & Borsetto, M. 1990. Thermoplasticity of saturated soils and shales: constitutive equations. Journal of Geotechnical Engineering 116, No. 1: 17651777. Hueckel, T., & Pellegrini, R. 1991. Thermoplastic modelling of undrained failure of saturated clay due to heating. Soils and Foundations 31, No. 3: 116. Roscoe, K.H. & Burland, J.B. 1968. On the generalized stress-strain behavior of wet clay. Symposium on Engineering Plasticity, Cambridge: 535610. Uchaipichat, A. 2005. Experimental Investigation and constitutive modelling of thermo-hydro-mechanical coupling In Unsaturated Soils, Ph.D thesis, University of New South Wales Sydney, Australia. Uchaipichat, A., & Khalili, N. 2009. Experimental investigation of thermo-hydro-mechanical behavior of an unsaturated silt. Gotechnique 59, 4: 339353.
822
Parameter calibration for hydro-mechanical modelling using numerical simulations of test results
R. Gmez, E. Romero, A. Lloret & J. Suriol C. Jommi
Universitat Politcnica de Catalunya, Barcelona, Spain Politecnico di Milano, Milano, Italy
ABSTRACT: The paper reports an experimental study and parameter calibration exercise centred on the hydro-mechanical response of an unsaturated soil. The study benefits from compacted silt samples retrieved from a fully instrumented experimental embankment, which was built under poorly compacted conditions to study the collapse response upon artificial flooding. In addition to the hydraulic characterisation, the experimental programme includes controlled-suction tests intended to follow similar wetting and drying paths to those undergone by the material under real conditions. The suction dependent elastoplastic BBM model is used to simulate selected experimental results related to the collapsible behaviour (its time evolution and its response with decreasing suction). Mechanical and hydraulic parameters of the model are determined by both direct experimental procedures and numerical back analysis. 1 INTRODUCTION 2 2.1 EXPERIMENTAL RESULTS Characterisation of the material and hydraulic properties
Volume changes of compacted soils induced by variations in their water content have many practical implications in the design and the service life of earth dams, river and canal dikes, and railway and road embankments. Field monitoring is essential to allow for a precise characterisation of the processes taking place with realistic assumptions, to assist in the validation of numerical tools, and to check the suitability of soil parameters obtained via laboratory tests. Despite its importance, the availability of instrumented case records of earth constructions is relatively scarce. In this regard, a fully instrumented embankment 5.4 m high was constructed at the CER (Centre dExperimentation Routire) facilities in Rouen (France) during the autumn of 2004 to study the earth-structure response in areas liable to flooding (LCPC 2007). Specifically, the embankment was built using poorly compacted silt (dry density between 1.3 and 1.5 Mg/m3) to study the collapse response that takes place during artificial flooding. This paper reports selected hydro-mechanical results of an experimental study centred on the wetting response of this poorly compacted silt. Some results were simulated numerically with the BBM constitutive model (Alonso et al. 1990) to validate the capability of this suction dependent elastoplastic model to reproduce the soil response. Simulation of a coupled hydro-mechanical collapse test were also run to back-analyse the hydraulic behaviour of the soil.
The main geotechnical properties and the ascompacted state of the silty material (d: dry density, e: void ratio, w: water content, Sr: degree of saturation, s: matric suction, and maximum compaction stress) are summarised in Table 1. Additional properties of the material and details of the retrieval of intact samples from the experimental embankment can be found in Gmez (2009) and Gmez et al. (2009). The water retention properties on drying at an approximate as-compacted void ratio of e 0.90 are shown in Figure 1. Data were obtained from controlled-suction paths, as well as from highrange tensiometer readings (Delage et al. 2008). Water retention data for drying at the ascompacted porosity were fitted to the van Genuchtens
Table 1. Main properties of the studied silty material. Particle size (mm) % 4.75 0.425 0.075 0.002 97 93 2730
Density Consistency of solids limits % s (Mg/m3) wL wP PI 2.66 State 3134 d (Mg/m3) e 1.40
20 1114 99
w Sr s (%) (%) (kPa) 50
max. stress (kPa) 105 5
0.90 17.0 50
823
1x100
1
e = 0.44 to 0.61
Tensiometer
MIP Controlled-suction kr =S r ; =5.96

0.8
Suction (MPa) 1x10-1
0.6
1x10
-2
0.4
e = 0.90
Controlled-suction Tensiometer Best fit
0.2
1x10
-3
0.3
0.4
0.5 0.6 0.7 0.8 Degree of saturation, S r
0.9
0 0.3 0.4 0.5 0.6 0.7 0.8 0.9 1
Figure 1. Water retention data on drying for different constant void ratios. Fitted curve Eq(1) for e 0.90.
equation (1980), in which the degree of saturation Sr is expressed as a function of the matric suction (ua - uw)
1 ua uw 1 Sr 1 P 0
Figure 2. Relative water permeability as a function of the degree of saturation. Experimental data from MIP and controlled-suction tests.
(1)
where ua is the air (gas) pressure, uw is the water (liquid) pressure, P0 is a reference air entry pressure at room temperature and is a shape parameter. It is important to remark that the retention curve is quite sensitive to the void ratio, as observed in the figure. This dependency of the water retention curve on porosity, which is more important in the low suction range (the range of interest of the present study), should be considered if accurate hydraulic simulations were envisaged (Romero et al. 1999, Della Vecchia 2009). The saturated water permeability was estimated to be around kw 2.2 106 m/s for a void ratio e 0.90. A Kozeny-Carmans model according to the following expression is used for the intrinsic permeability k k0
permeability. The model is based on the HagenPoiseuilles equation for laminar flow through a cylindrical capillary of known diameter. To simplify the calculation, only cross-sections with identical pore-size density functions are considered to be connected in a correlated way, according to the following expression (Romero et al. 1999) kr ( xm )
x x
xm
min
x 2 f ( x ) dx x 2 f ( x ) dx
xmax
min
(3)
which assumes a porous medium characterised by a pore size density function f(x) with pore sizes ranging from xmin to xmax. Good agreement between relative water permeability values using indirect methods (back-analysis from inflow and outflow data) and estimated values using MIP results may be observed in the figure. Test results were fitted to the following power expression kr ASr 1 . (4)
1 n
3 2
1 n0
3 n0
(2)
A summary of the calibrated hydraulic parameters referring to the previous expressions is presented in Table 2. 2.2 Controlled-suction tests The mechanical test presented here was focused on the evaluation of the compressibility of the compacted soil under loading at constant suction, as well as on the volume change response at constant stress during controlled-suction wetting and drying. The stress paths followed under oedometer conditions, which was performed in four stages, is presented in Figure 3. Stage AB represents a loading path from 20 kPa to 100 kPa at constant
where n is the porosity, n0 is the reference porosity and k0 is the intrinsic permeability at the reference porosity. Data for the variation of relative water permeability kr with degree of saturation Sr are plotted in Figure 2. Water permeability values for unsaturated states were obtained by back-analysis from inflow/ outflow data on controlled-suction steps and taking into account ceramic disc impedance effects. Mercury intrusion porosimetry (MIP) data were also used to indirectly estimate the relative water
824
Table 2. Hydraulic parameters of the poorly compacted silty material.

Void ratio, e
0.90
C D
A B
Retention curve (drying) P0 (MPa) 0.01
Intrinsic permeability k0 (m ) n0
Liquid phase relative permeability A 1 5.96
0.85
0.80
0.19 2.2 1013 0.474 1

100
50
0.75 Degree of saturation, S r (%)
E E
90 80 70 60 50 40
B D
Matric suction, (u a-u w) [kPa]
B A
20
10 C
5
5 20 10 Matric Suction, (u a-u w) [kPa]
50
100
Figure 4. Evolution of void ratio and degree of saturation vs. matric suction along the stress paths A-E in Figure 3.
E 10 100 Net vertical stress, (v-u a) [kPa]
20 50 200
Figure 3. Stress path followed under controlled-suction and controlled-load oedometer conditions.
matric suction of 50 kPa, which replicates the stress path undergone by the poorly compacted material during the construction of the embankment at constant water content of 17% (overburden depth around 5.4 m). The selected suction corresponds to the as-compacted value indicated in Table 1. Path BCDE stands for a wetting/drying/wetting cycle, with a final soaking stage, performed under constant vertical net stress. The cyclic hydraulic stage with suction varying between 50 kPa and 10 kPa simulates soil-atmosphere interaction undergone by the exposed material. These processes may induce water content changes such as those arising during rainfall and subsequent drying (evaporation). The final soaking stage to point E in Figure 3 replicates artificial flooding at maximum overburden depth. Figure 4 shows the evolution of void ratio e and degree of saturation Sr with matric suction, along the different stress paths indicated in Figure 3. The soil displays low compressibility on loading to a vertical net stress of 100 kPa at constant matric suction (path AB). The material is in a pre-yield condition, since the maximum compaction stress was estimated in 105 5 kPa at a matric suction of 50 kPa (Table 1). Changes in void ratio and degree of saturation are also small when suction cycles between 50 kPa and 10 kPa. Nevertheless, when suction decreased below 10 kPa, an important collapse occurred on
the poorly compacted material, as a consequence of water intake. This is consistent with the water retention data reported in Figure 1, in which an airentry value lower than 10 kPa can be estimated from the drying curve. During the flooding process, the degree of saturation of the material increased from about 55% to nearly 100%. From a mechanistic point of view, the collapsible response is modelled through the activation of a yield locus (loadingcollapse), which is described in the next section. 3 SIMULATION OF CONTROLLEDSUCTION TESTS
3.1 Mechanical parameters used in the simulation Under axis-symmetric stress states (p, q, s), where p is the mean net stress, q is the deviator stress and s is the matric suction, the BBM model (Alonso et al. 1990) postulates a yield function given by the following family of ellipses: q 2 M 2 p ps p0 p 0. (5)
In this equation, ps ks s is associated with the increase in apparent cohesion with suction, and p0 is the yield stress for isotropic stress conditions, which is related to the applied suction through
* s p0 p0 , c c p p
0
(6)
* , the yield stress for saturated conditions, where p0 is the hardening parameter, and (s) is the slope
825
of the virgin compression line for isotropic conditions. The latter is related to the suction through
s 0 1 r exp s r .
(7)
In the preceding equations, M, ks, (0), , pc, r and are model parameters; M is the slope of the critical state line, (0) is the slope of the virgin compression line for saturated conditions, is the slope of the (elastic) isotropic unloadingreloading paths, pc is a reference stress, r is a parameter controlling the unsaturated soil compressibility and provides the rate of change of (s) with s. The trace of the yield surface on the isotropic p:s plane is called the LC (Loading-Collapse) yield curve, because it represents the locus of activation of irreversible strains due to loading or wetting (implying a volumetric contractioncollapseof the structure). The model assumes isotropic hardening to be controlled by plastic volumetric strains (dvp ) through
* dp0 1 e d vp. * p0 0
(8)
A non-associated flow rule is adopted. Shear (d qp) and volumetric (d vp) plastic strain increments are related by the expression:
p d q
d vp
2 q , M 2 2 p ps p0
(9)
where is a constant related to M, and (0), defined in such a way that zero lateral strain is pree dicted for K0 loading. Volumetric (d v ) and shear e (d q) elastic strains induced by stress and suction changes inside the yield locus are given by the relationships:
boundary value problem, instead of referring to a fictitious material point element. This choice allows for tracking the whole unsaturated consolidation analysis, and hence for simulating the coupled hydro-mechanical response. On the other hand, the influence of the ceramic disc impedance on the time evolution of displacements could be taken into account. The analyses were run with CODE_BRIGHT (Olivella et al. 1996). A mesh of 80 linear elements for the 20 mm high soil sample and of 28 elements for the 7.15 mm high ceramic disc was adopted to simulate the entire test under axis-symmetric conditions. At the top of the sample constant atmospheric air pressure was imposed throughout the whole test. The load was changed accordingly to the experimental path followed. Suction was controlled through the water pressure value imposed at the bottom of the system. The ceramic disc was given a retention curve with an air entry value of 100 kPa, and a water permeability of kCDw 8.6 108 m/s, which were experimentally verified in the laboratory. Comparisons between model simulations and experimental data of the loading and wetting/ drying paths are shown in Figure 5 and in Figure 6. Porosity evolves through clear pre- and post-yield zones. The stiff pre-yielding response AD may be adequately reproduced, which corroborates the value of the maximum fabrication stress attained on compaction. Good match with the amount of final collapse could also be reproduced. As previously indicated (Figure 4), most of the collapse strain develops at suctions lower than 10 kPa (last part of path DE). In this mechanical model, plastic volumetric strains are generated when the wetting path drags the yield surface, affecting the * hardening parameter p0 . Test data and numerical simulations indicate that the previous wetting and
A A'
dp 1 dq dp; d e ; q 1 e p K 3G 3(1 2 ) ds e G K ; d v s 2(1 ) 1 e s patm

e d v
(10)
0.47
s=0.05MPa B' C' D' model: s=0.050.0360.0230.010.0360.01 BCD test: s=0.050.0360.0230.010.0360.01
s=0.05 B'
0.46 Porosity, n
where and s are the compressibility parameters against mean net stress and matric suction changes, respectively, G and K are the shear and the bulk moduli, is the Poissons ratio, and patm is the atmospheric pressure, which is added to avoid infinite elastic volumetric strains as suction approaches zero. 3.2 Numerical simulation of test results
0.45
0.44
0.43
test model
0.002 0.005
0.42
E' E
s=0
0.2
0.02 0.05 0.01 0.1 Net vertical stress, (v-ua) [MPa]
In contrast to usual practice, the test was simulated by means of a finite element analysis of the
Figure 5. Simulation of the evolution of porosity with vertical net stress.
826
0.47
C' C
D' D
A'
B' B
0.46 Porosity, n
0.45
0.44
test: ABCDE model: A'B'C'D'E' E' E
0.43
0.42 0
0.01 0.02 0.03 0.04 Matric suction, s=ua -uw (MPa)
0.05
Figure 6. Simulation of the evolution of porosity with matric suction. Table 3. Calibrated mechanical parameters for BBM. Elastic parameters 0.002 s 0.008 0.3
LC curve, which could develop collapsible strains at s 10 kPa. It is worth noting that the mechanical response given by the model is independent from the description of the hydraulic behaviour of the soil (water retention curve and water permeability). This does not hold true any more if the time evolution of collapse is analysed. If the retention curve for drying (depicted in Figure 1) is adopted in the numerical simulation, the whole consolidation stage, in which collapse occurs, lasts about 1 minute. The experimental time evolution of volumetric strains along the final soaking stage (DE), shown in Figure 7, demonstrates that collapse actually lasted slightly more than 10 minutes. To correctly reproduce the time evolution of collapse, a realistic estimation of the wetting branch of the water retention domain of the compacted soil, starting from a void ratio of e0 0.9, had to be adopted. Based on few experimental controlled suction data, the two curves represented in Figure 8
0
test k0= 2.2 10-13 m2 P0 = 0.0015 MPa k0= 5.0 10-13 m2 P0 = 0.0015 MPa
(0)
pc (MPa1) (MPa) M 0.01
e0
p0* (MPa)
Elastoplastic parameters
0.103 0.285 440
1.26 1
0.90 0.019
k0= 5.0 10-13 m2 P0 = 0.002 MPa
Table 4. Hydraulic parameters for silty material tested in the numerical simulations. Retention curve (wetting) P0 (MPa) 0.0015 0.0015 0.0020 0.19 0.19 0.19 Intrinsic permeability k0 (m2) n0 Liquid phase relative permeability A 1 1 1 1 5.96 5.96 5.96
8
2 5 20 50
10 Time, t (min)
100
2.2 1013 0.474 5.0 1013 0.474 5.0 1013 0.474
Figure 7. Time evolution of collapse strain at the end of the stress path DE in the controlled-suction oedometer test: experimental results and numerical simulations.
1x100 Drying Wetting Model, P0w = 0.002 MPa Model, P0w = 0.0015 MPa Suction (MPa) 1x10-1
drying paths (BCD) had not induced strain hardening (the initial position of the LC remains unchanged and the paths evolve inside the elastic domain). The calibrated constitutive parameters of the mechanical model are summarised in Table 3, together with the initial value of the hardening * parameter p0 . Some of them (M, , and (0)) had been already calibrated by means of direct procedures on different data sets (Gmez 2009). Elastic compressibility for suction variation, s, the parameters r and ruling (s), and suitable values for p0* and pc, were calibrated by means of the numerical back-analysis, to obtain a shape of the
1x10
-2
1x10-3 0.3 0.4 0.5 0.6 0.7 0.8 Degree of saturation, Sr 0.9 1
Figure 8. Water retention domain for the soil compacted at e 0.90.
827
were adopted. The two wetting curves simply differ from the drying curve by the air entry value, which had to be decreased by nearly one order of magnitude to catch the experimental data reported in the figure. The adoption of these curves allows for tracking correctly the degree of saturation, which dominates the relative permeability of the soil, and therefore the actual consolidation time. Best-fit with experimental data was obtained by slightly increasing the intrinsic hydraulic conductivity of the compacted soil to k0 5.0 1013 m . It is worth reminding that the initial value for the intrinsic permeability had been inferred from experimental data at much lower void ratios. Therefore, the contribution of the pores of larger dimensions to the hydraulic conductivity of the soil could have been slightly underestimated. 4 SUMMARY AND CONCLUSIONS
material level should be addressed, as the significant dependence of the water retention domain on void ratio suggests. Actually, the retention curves and the dimensions of the hysteretic domain do evolve along a general stress path implying significant changes of the void ratio. Models coupling the hydraulic and the mechanical responses of compacted soils, such as that proposed by Jommi & di Prisco (1994) and Romero & Jommi (2008), and further refined by Della Vecchia (2009) appear to be good candidates to enhance the predictive capabilities of numerical models. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Della Vecchia, G. 2009. Coupled hydro-mechanical behaviour of compacted clayey soils. Phd Thesis, Politecnico di Milano. Delage, P., Romero, E. & Tarantino, A. 2008. Keynote Lecture: Recent developments in the techniques of controlling and measuring suction. In D.G. Toll, C.E. Augarde, D. Gallipoli & S.J. Wheeler (eds.). Unsaturated Soils: Advances in Geo-Engineering: 3352. London: Taylor & Francis. Gmez, R. 2009. Caracterizacin hidro-mecnica del suelo del terrapln experimental de Rouen. MSc. thesis, Universitat Politcnica de Catalunya (in Spanish). Gmez, R., Romero, E., Lloret, A. & Suriol, J. 2009. Characterising the collapsible response of an in-situ compacted silt. In O. Buzzi, S. Fityus & D. Sheng (eds.). Unsaturated Soils: Experimental Studies in Unsaturated Soils and Expansive Soils, 1: 371376. London: Taylor & Francis. Jommi, C. & di Prisco, C. 1994. A simple theoretical approach to model the mechanical behaviour of partially saturated granular soils (in Italian). In Proc. Italian Conf. Il Ruolo dei Fluidi nei Problemi di Ingegneria Geotecnica. (C.N.R.), Italy. Mondov, 1(II): 167188. LCPC 2007. Remblais en zones humides et inondables. Consquences de linondation du remblai exprimental de Rouen et enseignements tirs pour la conception. Rapport final. Convention SNCF-LCPC-CETE Normandie Centre. Olivella, S., Gens, A., Carrera, J. & Alonso, E.E. 1996. Numerical formulation for a simulator (CODE_ BRIGHT) for the coupled analysis of saline media. Engineering Computations. 13: 87112. Romero, E., Gens, A. & Lloret, A. 1999. Water permeability, water retention and microstructure of unsaturated compacted Boom clay. Engineering Geology 54: 117127. Romero, E. & Jommi, C. 2008. An insight into the role of hydraulic history on the volume changes of anisotropic clayey soils, Water Resour. Res. 44, W12412: 116. van Genuchten, M. Th. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J., 44: 892898.
The data presented come from an experimental study on the hydro-mechanical response of a poorly compacted silt, which was retrieved from a prototype embankment built in Rouen. A controlledsuction stress path under oedometer conditions was performed, to replicate possible coupled hydraulic and mechanical history of the soil under in situ conditions. The experimental test was back-analysed by means of numerical simulations to verify the adequateness of some of the mechanical and hydraulic parameters, which characterise the behaviour of the soil and that were previously determined by direct tests. To model the mechanical behaviour of the compacted soil, the BBM model was adopted, which allowed for reproducing the test results accurately. By choosing a suitable form of the loading collapse curve, occurrence of significant collapse strains at suctions lower than 10 kPa could be matched well. The development of collapse strains at low suctions is a clear consequence of the low air-entry value of the water retention curve at the as-compacted void ratio. The study was complemented by a hydraulic characterisation (water retention and water permeability properties) of the as-compacted state. Important dependence on void ratio was observed for the water retention curve of the material at low suctions, which is the range of interest of the present study. The results presented show the relevant role played by the retention behaviour of the soil on the hydro-mechanical response of the system at the boundary value problem level. To further refine the study in view of the field applications, hydro mechanical coupling at the
828
Evaluation of a constitutive model for unsaturated soils: Stress variables and numerical implementation
N.A. Gonzlez & A. Gens
Technical University of Catalunya, Barcelona, Spain
ABSTRACT: This paper presents an evaluation of a constitutive model for unsaturated soils based on the BBM (Alonso et al. 1990). The focus of the paper is on the stress variables used and on the numerical algorithms adopted. Conventional stress variable approach (net stress and suction) as well as the approach that takes into account the degree of saturation (Bishops stress and suction) are examined. To solve the constitutive stressstrain equations, two stress integration procedures have been implemented, an explicit stress integration scheme with automatic substepping and error control techniques (Sheng et al. 2003) and a fully implicit stress integration scheme based on the Backward-Euler algorithm with substepping (Prez et al. 2001). A triaxial test on compacted Barcelona clayey silt performed by Barrera (2002) is analysed to examine the effects of stress variables and of numerical algorithms. Relevant features of different stress variable formulations to simulate experimental data are discussed. Particular attention is given to the strong dependency of Bishops stress on the shape of the soil water retention curve (SWRC). SWRC shape changes can lead to the appearance of an inflection point in the stress path (plane p-s) that may result, even for a simple test, in the elastic prediction of the stress path crossing the yield surface several times. 1 INTRODUCTION et al. 2003a,b). In implicit algorithms, all gradients are estimated at an advanced stress state (which is unknown) and then the resulting non-linear constitutive equations are solved by iteration (see, Prez et al. 2001; Vaunat et al. 2000). The relative performance of implicit and explicit methods is strongly dependent on the precise form of the constitutive model. For unsaturated constitutive models, the problem of the non-convexity of the yield surface at the transition between saturated and unsaturated states can significantly complicate the implementation of these models into finite element codes (Sheng et al. 2003; Sheng et al. 2008). In this paper both stress integration procedures are evaluated. The individual merits and useful features of stress variable approaches and the numerical algorithms to simulate a laboratory test are examined in this paper using a simulation of a laboratory tests on compacted Barcelona clayey silt performed by Barrera (2002).
There is agreement that at least two constitutive variables are generally required to represent adequately the full range of unsaturated soil behaviour, that is, including strength and deformation. Several review articles on the subject are available (Jommi 2000; Sheng et al. 2008; Nuth & Laloui 2008; Gens 2010). Conventional constitutive stress variable, namely net stress ( ij ij ua ij ) as well as the constitutive stress variable that takes into account the degree of saturation, commonly called Bishops stress or average stress ij ua ij Sr (ua uw ) ij ) are examined in ( ij this paper. In both formulations the second constitutive variable is the suction (s ua uw). ij are total stresses, ua the air pressure, uw the water pressure and ij the Kroneckerss delta. The selection of net stress or Bishops stress or other alternative as the constitutive variable remains at present a matter of convenience (Gens 2010). Incremental stress-strain equations for unsaturated soils can be solved by a wide range of explicit and implicit integration algorithms. Explicit algorithms, use the gradients of the yield surface and plastic potential at the start of the strain increment, and their accuracy can only be controlled by breaking up the strain increment into sub-increments, special automatic substepping and error control techniques have been proposed (see, Sloan et al. 2001; Sheng
2 2.1
FORMULATION Constitutive relations
The main features of the model used herein are: i) to follow the Barcelona Basic Model (BBM) characteristics to account for the behaviour of unsaturated soils. ii) to work with both types of
829
constitutive stresses: Net stress or Bishop stress. Suction is always adopted as the second stress variable. iii) Hydraulic hysteresis effects are not considered in the formulation. The yield function in the general stress space in terms of net stress and Bishops stress are defined by equations (1) and (2), respectively, g ( ) F 3J 2 M 2 ( p ps )( Pc p ) g ( 30) g ( ) F 3J 2 M 2 ( p )( Pc p ) g ( 30)
2 2
To complete the description, a relationship between degree of saturation and suction is required if Bishops stress is used as the constitutive stress. The expression suggested by van Genuchten (1980) is adapted for the SWRC, Sr
1 (s / a)
b c
(6)
(1)
where a,b,c, are constants. 2.2 Numerical integration algorithms The constitutive equations which characterize the elasto-plastic material can be briefly stated as follows, d d e d p d e,s d De (d d p d e,s ) G ( , k , s ) d m P dk dP0 0 d p p d p d (7)
(2)
where, p is the mean net stress, p is the mean Bishop stress, J is the square root of the second stress invariant of deviatoric stress tensor and g() is a function Lode angle. Preconsolidation pressure Pc is assumed to vary with suction and define the yield curve denoted as LC (loading collapse), P Pc pr 0 pr
0 s
(3)
where, P0 is the yield surface location at zero suction and is also the hardening parameter, pr is a reference mean stress, 0 is the slope of the saturated virgin consolidation line, s is the slope of the virgin consolidation line for an specific value of suction and is the slope of the unloading-reloading line (assumed independent of suction). s is assumed to vary with the suction according to,
s 0 (1 r )exp( s ) r
(4)
where, r and are material parameters. As noted in equation (1), net stress formulation requires an explicit variation of apparent cohesion with suction, defined as, ps kss, where, ks is a scalar variable. Isotropic hardening is controlled by the plastic volumetric strains (d vp ) through, P (1 e ) p dP0 0 d v 0 (5)
where, d, d e and d p are increments of the total, elastic and plastic strain tensors respectively and d e,s is the contribution of suction to increment of elastic strain tensor (only necessary in the net stress formulation). m is the flow vector, dk represents the increment of hardening parameters (in this case of P0) and d is the plastic multiplier. Note that in above equations, vector will be either net stress ( ) or Bishops stress ( ). For the stress state to remain on the yield surface, the consistency condition must be satisfied, F F F dF = ds dk 0 d s k
T
(8)
Solving equations (7) and (8) for the plastic multiplier d gives, d where, n F F G 1 ; ; ns ;m ;b s 3K s F P0 T m; T 1,1,1, 0, 0,0 P0 vp n T De d n s n T De b ds H n T De m
where, e is void ratio. Flow rule may be chosen to be associated or non-associated, through the incorporation of a constant related to M, 0 and (Alonso et al. 1990). The mechanical elastic behaviour is the same as in the Cam-clay models. Net stress formulation requires an explicit relation to consider the effect of suction on volumetric elastic strains, through the incorporation of the elastic stiffness parameter s.
(9)
H
830
Note that for Bishops stress formulation vector b 0. Combining equations (7) and (9), the constitutive equations give, d Depd W epds dk R epd Qds where, Dep De De mnT De H n T De m W ep De m (n s n T De b ) H n T De m (11)
R B
ep
1 Dep , k W ep , k , s s
s s
k1 Rep , k Q , k , s s
s s
(12)
(10)
where, Dep, Wep, Rep and Q are computed using equations (11). Using the above quantities, the stresses and hardening parameters at the end of the substep are 1 and k k1, respectively. These are then used to calculate a second estimate of the changes in stress and hardening parameters over the substep, namely, 2 D ep 1, k k1 s
n D
H n T De m
B
vp
Po
m Q=
B ( ns n T De b ) H n T De m
k2 R ep 1, k k1 s
W ep 1, s s s, k k1 s s Q 1, s s s, k k1 s s
A brief description of the integrations algorithms (explicit and implicit) is presented in the next sections. In the integration the infinitesimal increments in the above equations (denoted by d) are approximated with finite increments (denoted by ). 2.2.1 Explicit algorithm The more refined versions of the explicit algorithms (Sloan et al. 2001; Sheng et al. 2003) combine sub-stepping techniques with automatic sub-stepping control, error control and yield surface drift correction. An algorithm of this type has been implemented to solve the constitutive stress-strain equations at the Gauss point level. In this algorithm suction variable is treated as an additional strain component and it is assumed that it may be subincremented at the same rate as the other strain components. The substepping procedure automatically divides the increment of strain and suction into a number of substeps small enough to ensure that the desired integration accuracy is enforced. The scheme involves splitting the elasto-plastic strain step (1 ) and suction step (1 ) s into a series of smaller substeps, s Tn (1 ) and s s Tn (1 ) s (where 0 Tn 1), and using a modified Euler approximation for each substep. (1 ) and (1 ) s are the portions of the strain increment and suction increment, respectively, that are outside of the yield surface. The size of each substep is determined by estimating the error in the stress changes and comparing it to a user-defined tolerance, STOL. The procedure begins assuming that only one substep is necessary. Consequently Tn is set to unity and Tn is set to zero. A first estimation of the changes in stresses and hardening parameters at the end of the pseudotime step Tn are evaluated using a first order Euler approximation, as,
(13)
A more accurate estimate and the end of interval Tn is founded using the modified Euler procedure,
1 2 1 2 k k 1 2 k1 k2

(14)
A relative error measure is computed as,

Rn 2 1 k2 k1 1 mx , k 2
(15)
The current strain subincrement is accepted if Rn is not greater than STOL. If Rn STOL then the solution is rejected and a smaller step size is computed. After accepting or rejecting the current substep, the size of the next substep is calculated based on the estimated error and the set tolerance. The next pseudo-time step is found from the relation, Tn1 qTn where q is chosen so that, Rn1 STOL. A conservative choice for q is, q 0.9 STOL / Rn and it is also constrained to lie within the limits, 0.1 q 1.1, so that, 0.1Tn-1 Tn 1.1Tn-1. The end of the integration procedure is reached when the entire increment of strain and suction is applied so that Tn Tn 1. After a successful substep the yield surface consistency condition is verified. If it is violated a drift correction procedure (Potts & Gens 1985) is activated, which must ensure that the current state lies on the yield surface with a certain tolerance (YTOL). This correction changes both stress and internal variables but keeps the strain and suction increments unchanged. 2.2.2 Implicit algorithm A fully implicit stress integration scheme based on the Backward-Euler (BE) algorithm with
831
substepping (Prez et al. 2001) extended to unsaturated soil has been implemented. Integrating the constitutive equations with the BE methods, leads to an incremental algebraic format which is followed by a plastic corrector of the elastic trial stress violating the current yield surface. In this algorithm the plastic multiplier calculation is integrated with the internal variables updates and the incremental stress-strain relationship in a monolithic fashion. Time-integration equation with BE scheme yields the following non-linear local problem of the type R 0 (16):
R , k ,
(n 1)
the new iterative update of the eight variables is obtained, ( n 1) ( n 1) 1 J (n+1) , k ( n 1) , ( n 1) k ( n 1) R (n+1) , k ( n 1) , ( n 1)
(19)
Adding the iterative corrector to the old values of the independent variables yields the eight updates: (n+1) (n+1) (n+1) (n+1) (n+1) (n+1) k k k (n+1) (n+1) (n+1)
(n 1) De m (n 1) (n) De, s De e,s 0 (n 1) k T m (n 1) k (n) 0 k (n 1) p n+1 n+1 F , k , s n+1 0 (16) The unknowns of this local problem are the stresses (n1) and the hardening parameters k(n1) at time t(n1), and the plastic multiplier . e,s is required only for net stress formulation and is a known variable. As in the explicit algorithm, the elasto-plastic strain step and suction step will be subdivided in smaller steps, as, s Tn (1-) and s s Tn (1-) s (where 0 Tn 1), in case that no convergence is reached in the iterative process of residual minimisation. The non-linear system of equation (16) is solved by linearizing the residual and expanding it into a Taylor series, obtaining the following expression,
(20)
In some cases, when large strain increments are prescribed, the minimization of residual equations (17) is not possible after a given number of iterations. In this case, the strain and suction increments are reduced by Tn qTn, where q is chosen as 0.5. The end of the integration procedure is reached when the entire increment of strain and suction is applied so that Tn Tn 1. In order to start the iteration process, the elastic solution at the contact point with the yield surface is chosen:
(n 1) (n 1) 0 (n) 1 De ; k0 k (n) (n 1) s0 s (n) 1 s; 0 0
(21)
0 R ,k,
R ,k , 2 k O ,k ,
(17)
To solve the global problem with quadratic convergence it is necessary to use a consistent tangent matrix, n 1 PT J n 1 n 1

1
P
(22)
The gradient expression (,k,)R{,k,} is the Jacobian matrix J (eq.18),

J , k ,
(n 1)
e m I n D k m p nT
m k k m 1 p k F k De
De m k p m 0 t=t(n1)
PT (In 0 0) is the projection matrix on stress space. Important differences will be highlighted in section 4 between explicit and implicit algorithms, which will have influence on accuracy, robustness and efficiency of the stress-strain solutions for unsaturated soils. 3 VALIDATION
(18) Truncating after the first order terms, O[2], and solving the linearized system of equations
A triaxial compression test presented in Figure 1 on compacted Barcelona clayey silt performed by Barrera (2002), was selected to investigate the influence of the constitutive stress: net stress and
832
Degree of saturation
1 0.8 0.6 0.4 0.2 0
500
Option (a) (1) Option (a) (2) Option (b) (Barrera,2002)
400
Option (a) (1) Option (a) (2) Option (b) Barrera (2002)
s*Sr
300 200 100 0 0 200 400 600 800 1000
10
102 103 Suction (kPa)
104
105
Suction (kPa)
Figure 2. a) SWRC adopted for Silty Clay; b) Variation of the product sSr with suction.
900
Figure 1.
Stress paths followed in test A.

Suction (kPa)
800 700 600 500 400 300 200 100 0
A1
A2 A4
A5
Table 1. Material properties for Silty clay. Bishops stress Parameter 0 s ks M e0 P0 (kPa) pr (kPa) r a b c g Net stress 0.33 0.005 0.085 0.001 0.42 1.155 0.632 71 0.07 0.78 0.155 0.6 Option (a) (1,2) 0.33 0.005 0.085 1.155 0.632 71 0.03 0.80 0.12 286 1.6 0.80 (1)/ 1.20 (2) 0.6 Option (b)
A3
LC initial LC in A2 LC in A3
0 900
500
1000
1500
2000

800 700
A1 A2 A4 A5
(a)
0.33 0.005 0.085 1.155 0.632 71 1.5e8 1.2 0.095 56.8 1.3 0.23 0.6
Suction (kPa)
600 500 400 300 200 100 0 0 500

A3 LC initial LC in A2 LC in A3
1000
1500
2000
Mean Bishop's stress (kPa)

900 800 700
A1 A2 A4 A5
(b)
Suction (kPa)
600 500 400 300 200 100 0 0 500

A3 LC Initial LC in A2 LC in A3
1000
1500
2000
Mean Bishop's stress (kPa)
(c)
Bishops stress. The principle is to re-plot experimental data sets expressed in terms of classical net stress and suction planes into Bishops stress and suction planes. Test was performed starting from the same net stress state and suction ( p0 600 kPa and s0 800 kPa). An overconsolidated state was induced by a previous suction decreaseincrease cycle at p 600 kPa before shearing stage. Model parameters used in the numerical simulation are presented in Table 1. Parameters for net stress were taken from Barrera et al. (2002). For Bishops stress a new set of parameters (r, , pr) related to the LC curve have been calibrated. Options a(1) and a(2) use the same set of parameters (r, , pr) except for parameter c of the SWRC. Option (b) uses a different SWRC. SWRCs are presented in (Figure 2a) indicating the option used in each curve. Also, (Figure 2b) show the variation of the product sSr with suction.
Figure 3. Stress path followed in test A. a) Net stress; b) Bishops stress option (a2); c) Bishops stress option (b).
Figure 3 shows the stress paths in the constitutive stresssuction plane. Initial and final positions of the LC curves are also plotted in the figure. It is observed that the shape of the LC curve in the Bishops stress interpretation is similar to the net stress one, but enlarged. Whereas a drying or wetting path under constant net stress is a simple vertical line in the suction-net stress plane, the Bishops stress adopts a curved shape in the suction- Bishops stress plane. The curvature is closely linked to SWRC shape through the product sSr. SWRC employed in option (a) shows a lower product sSr than SWRC used in option (b), which is reflected in the curvature of stress path. In fact, it is observed
833
in (Figure 2b) for the option a(2), that the product sSr tends to decrease once a certain value of suction is reached (300 kPa) resulting in an inflection in the stress path and a reduction in the Bishops stress upon drying. Some authors (Nuth & Laloui 2008), have related this aspect to the residual state of saturation, where water phase is described as a film coating the particles rather than under the form of a connected reservoir. Consequently, particular attention must be paid for interpreting this limit state with the Bishops stress. Comparisons between simulations and experimental results of the wetting/drying cycle are shown in Figure 4. Net stress leads to a smaller reduction in volume. Similar results are obtaining using Bishops stress but option a(2) which shows a higher decrease in volume (collapse) due to wetting. This behaviour is generated by the effect of the product sSr, which increases during wetting from 800 kPa to 300 kPa and then reduces (Figure 2b). This leads to the stress path going outside of the LC curve from the beginning of wetting path (Figure 3b), inducing elasto-plastic strains. Once the product sSr begins to decrease, Bishops stress reduces and the stress path goes inside of the LC curve until it intersects again the LC at a low suction value (Figure 3b). Therefore, during wetting path the LC yield curve is intersected twice and numerical problems may appear concerning convergence. Figure 5 shows the results of shearing path. Similar results are obtained with the two options
-1% 0% 1% 2% 3% 4% 10 A3
drying path Experiment Net stress Bishop's option (a) (1) Bishop's option (a) (2) Bishop's option (b)
of constitutive stresses. However, option (a) of Bishops stress formulation predicts lower shear strength. This is a result of the influence of product sSr, which is not high enough to increase the strength due to increases in suction. For option (b) of Bishops stress the value of sSr is equivalent to the additional tensile strength (ps) required in net stress formulation and therefore the same shear strength is reached. 4 PERFORMANCE OF THE ALGORITHMS
Volumetric strain
wetting path
A2
A4 1 000
1 00
Suction (kPa)
Figure 4. Results of the wetting/drying cycle (path A2-A3-A4).

2000 1750 1500 TEST A
q (kPa)
1250 1000 750 500 250 0 0% 5% 10% 15%

Experiment Net stress Bishop's option (a)(1) Bishop's option (a)(2) Bishop's option (b)
Axial strain
Figure 5.
Results of shearing path (A4-A5).
Performance of numerical integration algorithms (explicit and implicit) is evaluated in terms of CPU time and the number of sub-increments required in each scheme. In all runs the yield surface tolerance is fixed at YTOL 108; this parameters is also used to control the convergence of the residual in the implicit algorithm. Control error tolerance of explicit algorithm varies from STOL 102 to 106. All CPU times presented are for an Intel Core Duo (2GHz) with 2GB of RAM. Results are presented in bar graphs where average values of the variables evaluated were computed for each stress path (Figure 1). Figure 6 shows a comparison between implicit and explicit algorithms for net stress and the option (a) of Bishops stress. It is observed that the wetting path (A2-A3) requires both a higher computational cost and higher mean number of sub-increments when Bishops stress and the implicit scheme are employed. This is because Bishops stress induces a high curvature of the stress path as a function of the SWRC; in particular it was noted that for option (a2) the LC yield curve is intersected twice during the wetting path (Figure 3b). As a consequence, the plastic corrector of implicit scheme has difficulties in returning to the yield surface and the requirement for sub-increments increases. Explicit algorithm is more efficient in this case because proceeds in an incremental fashion where all gradients are estimated at known stress states. Also, it is noted for path A2-A3 that the computational cost of net stress formulation is considerably lower than that of Bishops stress and differences between implicit and explicit schemes are minimised. It is also observed from Figure 6 that during stress path A1-A2 of isotropic load at constant suction, no significant differences are observed between implicit and explicit schemes. During drying path A3-A4, behaviour is elastic and no strain sub-incrementation is needed. Finally, during shearing path A4-A5, the explicit scheme demands a higher number of sub-increments than the implicit one. However, this tendency is not reflected in the CPU time, where explicit scheme spend slightly less CPU time than implicit one. This may be explained
834
2.0
Mean subincrements
CPU time (sec)
1.6 1.2 0.8 0.4 0.0 A1-A2 A2-A3
Implicit-Bishop's Explicit-Bishop's Implicit-net Explicit-net
20 16 12 8 4 0 Implicit-Bishop's Explicit-Bishop's Implicit-net Explicit-net
of sub-increments. At least in this case the shape of the SWRC has a little effect on the numerical efficiency during shearing stages.
A4-A5
A3-A4
A4-A5
A1-A2
A2-A3
A3-A4
Path
Path
CONCLUSIONS
Figure 6. Comparison of integration algorithms (STOL 104 for explicit scheme).

Mean Drift corrections
Mean subincrements
140 120 100 80 60 40 20 0 A1-A2 A2-A3 A3-A4 A4-A5

STOL=10-2 STOL=10-4 STOL=10-5 STOL=10-6
300 250 200 150 100 50 0 A1-A2 A2-A3 A3-A4 A4-A5 STOL=10-2 STOL=10-4 STOL=10-5 STOL=10-6
Important features of both types of constitutive stresses (net stress or Bishops stress) and of the numerical algorithms (explicit or implicit), have been noted: a. Net stress is a more simple and practical choice in terms of stress path representation than Bishops stress. However, it requires additional assumptions to take into account the shear strength increase with suction and the elastic volumetric strain due to changes in suction. Using Bishops stresses this features derive directly from the definition of the constitutive stress. b. In the Bishops stress formulation, a strong influence of the SWRC shape on compressibility and shear strength behaviour is observed. SWRC shape changes can lead to the appearance of an inflection point in the stress path (plane p-s) that may result, even for a simple test, in the elastic prediction of the stress path crossing the yield surface several times. c. In terms of the efficiency of numerical algorithms, explicit scheme is likely to be more robust than implicit scheme to solve the kind of complex stress path mentioned before. REFERENCES
Alonso EE, Gens A & Josa A (1990). A constitutive model for partially saturated soils. Gotechnique, 40, 405430. Barrera M, Romero E, Snchez M & Lloret A (2002). Laboratory tests to validate and determine parameters of an elastoplastic model for unsaturated soils. Int. Symp. on identification and determination of soil and rock parameters for geotechnical design. Jean-Pierre MAGNAN, 351358. Gens A (2010). Soilenvironment interactions in geotechnical engineering. Gotechnique 60, No. 1, 374. Jommi C (2000). Remarks on the constitutive modelling of unsaturated soils. Proc. Int. Workshop UNSAT, Trento, pp. 139153. Nuth M & Laloui L (2008). Effective stress concept in unsaturated soils: Clarification and validation of a unified framework. Int. J. Numer. Anal. Geomech, 32:771801. Prez A, Rodrguez A & Huerta A (2001). Consistent tangent matrices for substepping schemes, Comput. Methods Appl. Mech. Engrg. 190, 46274647. Potts D & Gens A (1985). A critical assessment of methods of correcting for drift from the yield surface in elastoplastic finite element analysis. Int. J. Numer. Anal. Methods Geomech., Vol. 9, pp. 149159.
Path
Path
Figure 7. Influence of STOL on sub-increments and drift corrections of explicit algorithm: Bishops stress (option a).
2.0
20
CPU Time (sec)
1.6 1.2 0.8 0.4 0.0 A1-A2 A2-A3
Mean subincrements
Implicit-option (a) Explicit-option (a) Implicit-option (b) Explicit-option (b)
16 12 8 4 0
Implicit-option (a) Explicit-option (a) Implicit-option (b) Explicit-option (b)
A3-A4
A4-A5
A1-A2
A2-A3
A3-A4
A4-A5
Path
Path
Figure 8. Comparison of integration algorithms: options (a) and (b) of Bishops stress (STOL 104 for explicit scheme).
by the fact that implicit scheme requires second derivatives of yield function and plastic potential and inversion of the Jacobian matrix. Figure 7 shows the influence of the error control tolerance (STOL) on mean number of sub-increments and drift corrections of explicit algorithm. As expected, the number of subincrements and drift corrections increases as STOL is decreased. Due to the fact that the stress error controls the strain sub-increments in proportion to the square root of STOL, their number increases by a factor of roughly 10 if STOL is reduced by an order of magnitude. Figure 8 shows a comparison between options (a) and (b) of Bishops stress. It will be observed that the computational cost during path (A2-A3) is considerably reduced using option (b). Mean number of sub-increments is also reduced using option (b) and in the case of the implicit scheme reduction is almost half. Improvements are explained by the fact that during wetting path the LC curve is only intersected once at a very low suction value (Figure 3c). Figure 8 also shows that in the shearing path (A4-A5) there are no important effects on either CPU time or number
835
Sheng D, Gens A, Fredlund DG & Sloan SW (2008). Unsaturated soils: From constitutive modeling to numerical algorithms. Computers & Geotechnics, 35:810824. Sheng D, Sloan SW, Gens A & Smith DW (2003). Finite element formulation and algorithms for unsaturated soils. Part I: Theory. Part II: Verification and application. Int. J. Numer. Anal. Methods Geomech., 27, 745790.
Sloan SW, Abbo AJ & Sheng D (2001). Refined explicit integration of elasto-plastic models with automatic error control. Engineering Computations; 18:121154. Vaunat J, Cante JC, Ledesma A & Gens A (2000). A stress point algorithm for an elastoplastic model in unsaturated soils. Int. J. Plasticity 16, No. 2, 121141.
836
Prediction of the behaviour of unsaturated soils using evolutionary polynomial regression: An incremental approach
A.A. Javadi, A. Johari, A. Ahangar-Asr & A. Faramarzi D.G. Toll
Computational Geomechanics Group, University of Exeter, Exeter, UK School of Engineering, University of Durham, Durham, UK
ABSTRACT: A number of constitutive models have been developed to describe the complex behaviour of unsaturated soils. Despite the significant recent developments in constitutive modelling of unsaturated soils, many of the existing models are unable to predict all aspects of behaviour of unsaturated soil behaviour under different loading conditions. In this paper a new approach is presented, based on evolutionary polynomial regression (EPR), for modelling of unsaturated soils. EPR is an evolutionary data mining technique that generates a transparent and structured representation of the behaviour of a system directly from data. The capabilities of the proposed EPR-based framework in constitutive modelling are illustrated by application to modelling the behaviour of unsaturated soils. EPR models are developed and validated using results from a comprehensive set of triaxial tests on samples of compacted unsaturated soils from literature. The main parameters contributing to the constitutive behaviour of unsaturated soils during shearing, namely, initial water content, initial dry density, mean effective stress with respect to pore air pressure, axial strain, suction, volumetric strain, and deviatoric stress are used in developing the EPR models. The developed models are used to predict different aspects of behaviour of unsaturated soils for conditions not used in the model building process. The results show that the proposed approach is very effective and robust in modelling the behaviour of unsaturated soils and provides a unified framework for constitutive modelling of unsaturated soils. The merits and advantages of the proposed approach are highlighted.
INTRODUCTION
Describing the mechanical behaviour of unsaturated soils based on a single effective stress equation, similar to the one proposed by Bishop and Donald (1961), has always been limited and has led to development of different models to describe the observed behaviour of these soils. Consequently, the mechanical behaviour of unsaturated soils has been the topic of numerous investigations in recent years. Fredlund and Morgenstern (1977) proposed the concept of state surface defined in terms of independent state variables such as ( ua) and (uauw), where is the total stress, ua is the pore-air pressure, and uw is the pore-water pressure. Using these state variables, suitable state surfaces were defined for the stressstrainstrength behaviour of unsaturated soils. Alonso et al. (1990) and Toll (1990) proposed critical state frameworks by considering the effect of net mean stress and
suction separately. Gens and Alonso (1992) developed a constitutive model of expansive soils based on the Barcelona Basic Model of Alonso et al. (1990). Among recent important contributions are the works of Alonso et al. (1990), Toll (1990), and Wheeler and Sivakumar (1995) within the independent stress state framework, and the contributions of Kogho et al. (1993), Bolzon et al. (1996), Loret and Khalili (2000 and 2002), Vaunat et al. (2000), Gallipoli et al. (2003),Wheeler et al. (2003), Borja (2004), Ehlers et al. (2004) and Khalili et al. (2008) based on the combined stress (sometimes called effective stress) concept. These models, though successful in presenting the critical state parameters in terms of degree of saturation or soil suction, but they are yet incapable of dealing with many aspects of unsaturated soil behaviour. As a result, several modifications have been and are being proposed to the existing constitutive models in many conference and journal publications to account for
837
these aspects of unsaturated soil behaviour. In this paper a new approach is presented for constitutive modelling of unsaturated soils using evolutionary polynomial regression. The capabilities of the proposed EPR-based approach in constitutive modelling are illustrated by application to modelling the behaviour of unsaturated soils. Results from a comprehensive set of triaxial experiments on samples of compacted unsaturated soils are used in development and validation of EPR models. It is shown that the proposed approach is very effective and robust in modelling the behaviour of unsaturated soils and provides a unified framework for constitutive modelling of unsaturated soils. 2 EVOLUTIONARY POLYNOMIAL REGRESSION
of data. An interesting feature of EPR is in the possibility of getting more than one model for a complex phenomenon. A further feature of EPR is the high level of interactivity between the user and the methodology. The user physical insight can be used to make hypotheses on the elements of the target function and on its structure (Equation (1)). Selecting an appropriate objective function, assuming pre-selected elements in Equation (1) based on engineering judgment, and working with dimensional information enable refinement of final models. Detailed explanation of the method can be found in Rezania et al. (2008). 3 DATABASE
Evolutionary polynomial regression (EPR) is a data-driven method based on evolutionary computing, aimed to search for polynomial structures representing a system. A general EPR expression can be presented as (Rezania et al. 2008) y F (X , f (X ), a j ) a0
j 1 n
(1)
where y is the estimated vector of output of the process; aj is a constant; F is a function constructed by the process; X is the matrix of input variables; f is a function defined by the user; and n is the number of terms of the target expression. The general functional structure represented by F(X, f(x), aj) is constructed from elementary functions by EPR using a Genetic Algorithm (GA) strategy. The GA is employed to select the useful input vectors from X to be combined. The building blocks (elements) of the structure of F are defined by the user based on understanding of the physical process. While the selection of feasible structures to be combined is done through an evolutionary process the parameters aj are estimated by the least square method. In this technique, the combination of the genetic algorithm to find feasible structures and the least square method to find the appropriate constants for those structures implies some advantages. In particular, the GA allows a global exploration of the error surface relevant to specifically defined objective functions. By using such objective functions some criteria can be selected to be satisfied through the search process. These criteria can be set in order to (a) avoid the overfitting of models; (b) push the models towards simpler structures; and (c) avoid unnecessary terms representative of the noise in data. EPR shows robustness and in every situation can get a model truly representative
Results from a set of triaxial tests on Lateritic gravel reported by Toll (1988) were adopted for the analysis. Table 1 indicates the range of basic soil properties in the database. Table 2 shows the initial condition of soil specimens and indicates whether the results of a particular test were used for training or testing of the EPR models. This database consists of results from 23 different unsaturated specimens, prepared using static or dynamic compaction. However, for the sake of consistency, only 14 specimens prepared with static compression were considered in this investigation. The remaining tests were not included in the database used as the specimens were prepared by dynamic compaction. The experimental results (graphs) presented by Toll (1988) were digitized which resulted in a database including a total of 5153 patterns that were used for training and testing of the EPR models. 4 DEVELOPMENT OF THE EPR MODELS
In development of EPR models a number of constraints can be implemented to control the constructed models in terms of the length of the equations, type of the functions used, number of
Table 1. Range of soil properties of specimens. Properties Initial water content (%) Dry density (Mg/m3) Degree of saturation (%) Suction (kPa) Axial strain (%) Deviatoric stress (kPa) Volume Strain (%) Mean effective stress with respect to pore air pressure (kPa) Range 1726.3 1.4421.716 46.595 9.5545.4 011.52 0930 7.50.35 23.9237.6
838
Table 2. Initial condition of soil specimens. Water content (%) 19.6 25.5 20.8 21.4 20.7 21 17 21.1 25.1 24.9 26 25 24.3 25.8 Dry density (Mg/m3) 1.442 1.632 1.531 1.551 1.646 1.489 1.474 1.587 1.508 1.506 1.706 1.702 1.708 1.705 Degree of saturation (%) 51.4 84.9 61.1 64.4 70.2 58.5 46.5 66.4 71.6 70.8 95 90.9 89 94.2 Initial suction (kPa) 384 4 149 22 105 256 450 186 11 26 5 12 78 54 Ua (kPa) 496 247 297 227 300 445 450 304 300 299 299 300 450 299 3 (kPa) 552 302 350 300 353 500 500 352 350 350 350 399 473 324 Neural network status Train Train Train Train Train Test Train Test Test Train Train Train Train Train
Sample MGU1 MGU2 MGU3 MGU4 MGU5 MGU6 MGU7 MGU8 MGU10 MGU11 MGU14 MGU15 MGU22 MGU23
terms, range of exponents, number of generations etc. As a result there is a potential to achieve different models for a particular problem which enables the user to gain additional insight into the problem (Rezania et al. 2008). Applying the EPR procedure, the evolutionary process starts with a constant mean of output values. By increasing the number of evolutions it gradually picks up the different participating parameters in order to form equations representing the constitutive relationship. Each model is trained using the training data and tested using the testing data. The level of accuracy at each stage is evaluated based on the coefficient of determination (COD) as the fitness function:
Table 3. Parameters involved in the EPR models*. Contributing parameters w, d, sr, a, (p ua), si, v , qi, a
i
Model output qi1 si1
vi 1
Notes: *a axial strain; (p ua) mean effective stress with respect to pore air pressure; sr degree of saturation, w initial water content, d dry density; S suction; v volumetric strain; q deviator stress; a axial strain increment.
COD 1
(Ya Yp )2
N
1 Ya Ya N N N
(2)
where Ya is the actual desired output; Yp is the output predicted by EPR and N is the number of data points on which the COD is computed. If the model fitness is not acceptable or the other termination criteria (in terms of maximum number of generations and maximum number of terms) are not satisfied, the current model should go through another evolution in order to obtain a new model. 5 CONSTITUTIVE MODELLING OF UNSATURATED SOILS USING EPR
A typical scheme to train most of the material models based on data mining techniques like neural
networks, for soils includes an input set providing the network with information relating to the current state units (e.g., current stresses and strains) and then a forward pass through the network yields the prediction of the next expected state of stress or strain relevant to an input strain or stress increment (Ghaboussi, 1998). Due to the incremental nature of soil stressstrain modelling in practical applications, this scheme has been utilised in this section. The EPR models have nine input parameters as summarised in Table 3. The first three input parameters namely, gravimetric water content, dry density and degree of saturation, represent the initial conditions of the soil specimens and the other parameters, namely, net mean stress with respect to pore air pressure, axial strain, deviatoric stress, suction and volumetric strain are updated incrementally during the training and testing based on the outputs from the previous increment. The output parameters are deviatoric stress, suction and volumetric strain
839
Table 4. COD values for EPR incremental models. Equation Deviator stress (Equation 6) Volume strain (Equation 7) Suction (Equation 8) COD values for training (%) 99.95 99.99 99.99 COD value for testing (%) 90.50 99.04 99.85
Deviatoric stress(kPa)
500 450 400 350 300 250 200 150 100 50 0 0 1 2 3 4 5 Axial strain(%) 6
Experiment EPR prediction
corresponding to the end of the incremental step for the three EPR models. The data is divided into training and testing sets (Table 2). One set is used for training to obtain the constitutive models and the other one is used for validation to appraise the generalisation capabilities of the trained models. Three separate models are developed for deviator stress (q), suction (s) and volumetric strain (v). After development of the EPR models, from the 15 resultant equations for deviator stress, 6 equations do not include the effect of all contributing parameters. Among the remaining equations the most appropriate and efficient one is selected as the final model. The same procedure is followed to choose the best fit constitutive equations for volumetric strain and suction. Equations 3, 4, 5 represent the incremental EPR constitutive models for deviator stress, volume strain, and suction respectively. Table 4 represents COD values for developed models. qi 1
Figure 1. Comparison between the incremental EPR model predictions with experimental data for deviatoric stress (MGU 6).
20 15
Volumetric strain
10 5 0 -5 -10 -15 -20 0 1 2 3 4 5 Axial strain(%) 6
Figure 2. Comparison between the incremental EPR model predictions with experimental data for volume strain (MGU 6).
In this section, the EPR models (equations 3, 4 and 5) are used to predict the entire stress paths, incrementally, point by point, in the q : a; s : a and v : a spaces. Results from a testing set of data
529.41 0.76 1.66 106 592.96 3 3 0.01v 0.98qi 0.01Sr2 48.78d 2 i Si ( p ua )3 ( p ua )2 a d

2 2.49w 2 3.16 10 5 a qi3 4124.48 a d 16.77 a qi a ( p ua ) w
228.94 a vi
a
372.99 a 18.05 a vi 423.93
(3)
vi 1
3 0.57( p ua ) 54.02 a d vi 4.82 10 4 a qi3 3 1.87 10 5 a Si v i w3 w2 a ( p ua )2
3 1.38 10 9 a d qi3
2.36 a w3 a
3 d Sr3
vi 0.001w 0.03
(4)
Si 1
55.23Si2 qi w3 Sr3
3 6322.18d qi
w3 ( p ua )
3 126.38 6.27Sr 13267.78 1.11 10 qi vi 7.88 10 4 w 3 2 a d Sr a d 2 652.65 3.43 10 3 ( p ua ) 0.9 96Si 8.42 10 5 Sr3 71.52 d
5.56 10 2 a qi3 Sr3
1881.71 a w qi
3 d Sr3 a
(5)
840
300 250
Suction (kPa)
200 150 100 50 0 0 1 2 3

Axial strain(%)
Figure 3. Comparison between the incremental EPR model predictions with experimental data for suction (MGU 6).
(MGU6) are used to evaluate the ability of the incremental EPR models to predict the complete behaviour of unsaturated soil during the entire stress paths. The values of water content, dry density and degree of saturation represent the initial conditions of the soil and are constant throughout the test. Other contributing parameters are updated in each incremental step, considering the values from the previous increment and the EPR models outputs in response to an axial strain increment. 6 DISCUSSION
stress, volume strain and suction in terms of initial values for these and other contributing parameters from the previous stage in response to an increment of axial strain in tri-axial test experiment are suggested. The results show that EPR constitutive models can capture the underlying relationships between various parameters directly from experimental data and predict the unsaturated soil behaviour with a very high precision. EPR constitutive models are also tested using data that were not used in the training stage of EPR model development process. The results show the excellent ability of the EPR models in generalizing the training to predict the behaviour of unsaturated soils under unseen conditions. It was shown that the (incremental) EPR models can be used to predict the complete stress paths in the q : a, v : a and s : a spaces incrementally, point-by-point. As the errors of prediction of the individual points are accumulated in this approach, and still the EPR models are able to predict the complete stress paths with a very good degree of accuracy is another representative of the robustness of framework used to develop constitutive models for unsaturated soils. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. (1990), A Constitutive Model for Partially Saturated Soils, Geotechnique, 40(3), pp 405430. Bishop, A.W. & Donald, I.B. (1961), The Experimental Study of Partly Saturated Soil in Triaxial Apparatus, Proc., 5th International Conference on Soil Mechanic and Foundation Engineering, Paris, 1, pp 1321. Bolzon, G., Schrefler, B.A. & Zienkiewicz, O.C. (1996), Elastoplastic soil constitutive law sgeneralized to partially saturated states, Geotechnique, 46(2), pp 279289. Borja, R.I. (2004), Cam-Clay plasticity. Part V: A mathematical framework for three-phase deformation and strain localization analyses of partially saturated porous media, Comput. Methods Appl. Mech. Engrg., 193, pp 53015338. Chateau, X. & Dormieux, L. (2000), The Behaviour of Unsaturated Porous Media in the Light of Micromechanical Approach, In Proceedings, Unsaturated Soils for Asia, Singapore, 1819 May. Edited by H. Rahardjo, D.G. Toll, & E.C. Leong. A.A. Balkema, Rotterdam. pp 95100. Ehlers, W., Graf. T. & Ammann, M. (2004), Nonlinear behaviour in the deformation and localization analysis of unsaturated soil, Proc. Int. Conf. From Experimental Evidence towards Numerical Modeling of Unsaturated Soils, Weimar, Germany, 2, pp 4152. Fredlund, D.G. & Morgenstern, N.R. (1977), Stress Strain Variables for Unsaturated Soils, Journal of Geotechnical Engineering, ASCE, 103(GT5), pp 4427466.
Current practice in geotechnical engineering, when dealing with unsaturated soils, often involves the use of subjective or empirical techniques to evaluate imprecise field or experimental data. These techniques often involve complex relationships between various parameters. The simplifying assumptions that are usually made in the development of the traditional methods may, in some cases, lead to very large errors. In this paper a new approach is presented for modelling of unsaturated soil behaviour using EPR. The proposed EPR models generate a transparent and structured representation of the system. An additional advantage of the EPR approach is that there is no need to assume a priori the form of the relationship between the input and output parameters. The explicit and transparent structures obtained from the proposed EPR method can allow physical interpretation of the problem which gives the user additional insight into the relationship between input and output parameters. 7 CONCLUSIONS
Evolutionary polynomial regression has been used to develop constitutive models to represent various aspects of unsaturated soil behaviour. Incremental models to predict the changes in deviator
841
Gallipoli, D., Gens, A., Sharma, R. & Vaunat, J. (2003), An elastoplastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour, Geotechnique, 53(1), 123135. Gens, A. & Alonso, E.E. (1992), A framework for the behaviour of unsaturated expansive clays, Canadian Geotechnical Journal, 29:10131032. Ghaboussi, J., Pecknold D.A., Zhang, M. & Haj-ali, R.M. (1998), Autoprogressive Training of Neural Network Constitutive Models, Int. J. Number. Meth. Engng., 42, pp 105126. Khalili N., Habte, M. & Zargargashi, S. (2008), A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses, Computers and Geotechnics, 35(6), 872889. Kogho, Y., Nakano, M. & Myazaki, T. (1993), Theoretical aspects of constitutive modelling for unsaturated soils, Soils and Foundations, 33(4), pp 4963. Loret, B. & Khalili, N. (2002), An effective stress elastoplastic model for unsaturated porous media, Mechanics of Materials, 34, 97116. Loret, B. & Khalili, N. (2000), A three-phase model for unsaturated soils, Int. J. Numer. Anal. Methods Geomech., 24, 893927.
Rezania, M., Javadi, A.A. & Giustolisi, O. (2008), An Evolutionary-Based Data Mining Technique for Assessment of Civil Engineering Systems, Journal of Engineering Computations, 25(6), pp 500517. Sun, D.A., Matsuoka, H., Yao, Y.P. & Ichihara, W. (2000), Three Dimensional Elasto Plastic Model for Soils, Proceedings of the Unsaturated Soils for Asia, Singapore, 1819 May. Edited by H. Rahardjo, D.G. Toll, and E.C. Leong. A.A. Balkema, Rotterdam. pp 153158. Toll, D.G. (1990), A Framework for Unsaturated Soil Behaviour Geotechnique, 40(1), pp 3144. Toll, D.G. (1988), The Behaviour of Unsaturated Compacted Naturally Occurring Gravel, Ph.D. Thesis, University of London. Vaunat, J., Romero E. & Jommi, C. (2000), An elastoplastic hydro-mechanical model for unsaturated soil, Proc. International Workshop on Unsaturated Soils, Trento, pp 121138. Wheeler, S.J., Sharma, R.S. & Busson, M.S.R. (2003), Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils, Geotechnique, 53(1), pp 4154. Wheeler, S.J. & Sivakumar, V. (1995), An elasto-plastic critical state framework for unsaturated soil, Geotechnique, 45(1), pp 3553.
842
A fully coupled hysteretic finite element model for hydro-mechanical analysis of unsaturated soils
N. Khalili, S. Zargarbashi & M.A. Habte
Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering, University of New South Wales, Sydney, NSW, Australia
ABSTRACT: The finite element formulation for fully coupled analysis of flow and deformation in unsaturated soils is presented. Both mechanical and hydraulic hystereses are considered. The elastic-plastic behaviour and mechanical hysteresis are captured using the bounding surface plasticity. The hydraulic hysteresis is accounted for through the soil water characteristic curve. Attention is also given to the interrelations between the effective stress and wetting and drying paths. Spatial discretization is achieved using the Galerkin technique and the time discretization using the finite difference approach. The application of the model is demonstrated using several numerical examples. 1 INTRODUCTION In its most common form, the effective stress is defined as (Bishop, 1959): net s (1)
Interrelation of hydraulic and mechanical behaviour in unsaturated soils is a subject of great interest in many geotechnical engineering problems. The plastic volumetric strain affects the soil water characteristics curve and causes a change in the degree of saturation. Wetting and drying cycle, on the other hand, increases the stiffness (Gallipoli et al., 2003, Wheeler et al., 2003) and causes irreversible volumetric strain (Alonso et al., 1995). Several constitutive models have been proposed for hydro-mechanical analysis of unsaturated soils. Notable contributions include the work of Vaunat et al. (2000), Jommi (2000), Buisson and Wheeler (2000), Wheeler et al. (2003), Gallipoli et al. (2003), Sheng et al. (2004), Tamagnini R (2004) as well as Khalili et al. (2008). However, only a few of these models take into account the simultaneous flow of air and water and their complex interaction with the deformation of the solid skeleton within a consistent elasto-plastic framework. In this paper, a fully coupled hydro-mechanical model is presented for flow and deformation analysis of unsaturated soils taking into account hydraulic as well as mechanical hysteresis effects. 2 EFFECTIVE STRESS
in which net pa is the net stress ( is total stress), is the effective stress parameter, attaining a value of one for saturated soils and zero for dry soils and s pa pw is the matric suction (pa and pw are pore air and water pressures, respectively) and is the identity vector. The incremental form of the effective stress equation is obtained through a simple differentiation of (1) as net s (2)
A superimposed dot indicates the rate of net p a is the incremental net stress, change. p a p w is the incremental suction, d(s)/ds s is the incremental effective stress parameter. The effective stress parameter, , is quantified using the model proposed in Khalili and Zargarbashi (2010). 3 GOVERNING EQUATIONS OF FLOW AND DEFORMATION IN UNSATURATED SOILS
The effective stress concept plays a central role in the present formulation. It defines the contribution of suction to the stress of the solid skeleton and serves as a platform for coupling of fluid flow and deformation fields in the unsaturated porous medium.
Satisfying the conservation equations of mass, the differential equations governing flow of water and air in unsaturated porous media may be written as (Khalili et al., 2008).
843
k k d p 1 d sVw div rw pw w g nwcw w w dt V dt w k k d p 1 d sVa div ra pa a g naca a a dt V dt a
(3)
Differential equations (8) and (9) along with (5) form the governing equations for coupled flowdeformation analysis of unsaturated soils. 4 PLASTICITY MODEL
(4)
in which k is the intrinsic permeability of the soil, and krw and kra are the relative permeability with respect to water and air, respectively. w and a are the dynamic viscosity of water and air, and g is the vector of gravitational acceleration. nw is the volumetric water content, na is the volumetric air content, and cw and ca are the coefficients of compressibility for water and air phases, respectively. V, Vw and Va are total volume, volume of water phase and volume of air phase respectively, and t is time. The deformation model of solid phase is in turn expressed as div D sym u pw 1 pa F 0
(5)
in which u denotes the displacement vector of soil skeleton and F is the body force per unit volume. The constitutive equations complementing the governing equations are provided through defining the volumetric deformations of water and air phases in terms of primary model parameters as (Khalili et al., 2008). V w v a11 p w a12 p a V V a v a21 p w a22 p a (1 ) V
S
The elastic-plastic behaviour of the solid skeleton is captured through the bounding surface plasticity (Dafalias and Herrmann, 1980). In this approach, plastic deformation occurs when lies on or within the bounding surface. This is achieved by defining the hardening modulus h a decreasing function of the distance between and an image point on the bounding surface. The image point is selected using a mapping rule such that the normals to the loading surface at the current stress point, and to the bounding surface at the image point are the same. The loading surface is a sub-yielding surface located inside the bounding surface. For the sake of clarity, the model is formulated in the p q plane; p 13 ( T ) is the mean normal effective stress, and q 3J2 is the deviatoric stress. J2 1 2 ( s T s ) is the second invariant of the deviator stress vector, s p. The corresponding work conjugates strains are volumetric strain v T and deviatoric strain
q 2 3 [( 13 v )T ( 13 v )] (in which is the strain vector).

4.1 Bounding surface
(6) (7)
Following the work of Khalili et al. (2005) and to maintain contact with the mechanics of saturated soils, the bounding surface is defined as q F ( p, q , pc ) Mcs p
N
ln pc p ln R
0
(10)
in which a11 a22 a12 a21 n sr |v are determined using the soil water characteristic curve of the soil (Khalili et al., 2008). v is the volumetric strain. Now, combining the governing equations for flow of air and water (34) and the constitutive equations (67) we have k k a11 p w a12 p a (8) div rw pw w g div u w k k a22 p a a21 p w div ra pa a g (1 )div u a (9) with, a11 cw nw a12 , a22 a21 ca na
where the superimposed bar denotes stress conditions on the bounding surface, Mcs is the slope of the critical state line (CSL) in the q p plane. The parameter pc controls the size of F and is a function of suction and plastic volumetric strain. The material constant R represents the ratio between pc and the value of p at the intercept of F with the critical state line in the q p plane. The material constant N controls the shape of F. 4.2 Loading surface
The conventional definition that is always located on a loading surface is applied. The loading surface is assumed to be of the same shape and homologous to the bounding surfaces about the centre of homology. For first time loading, the centre of homology is located at the origin of
844
stresses in q p plane. For cyclic loading, the centre of homology moves to the last point of stress reversal and the maximum loading surface through the point of stress reversal serves as a local bounding surface (Fig. 1). To maintain similarity with the bounding surface, the loading surfaces undergo kinematic hardening during loading and unloading such that they are tangent to the maximum loading surface at the centre of homology. The image point for cyclic loading is located sequentially by projecting the stress point onto a series of intermediate image points on successive local bounding surfaces passing through each point of stress reversal. The loading history of the soil is captured through the stress reversal points and the corresponding maximum loading surfaces. Within this context, the loading surface (f) takes the form q , q , p c f (p ) M p
cs N
4.3 Plastic potential The plastic potential (g) is defined by integrating an extension of Rowes stress-dilatancy relationship with respect to p and q as (Khalili et al., 2008)
A 1 AMcs ( p ) p q g ( p, q, po ) t 1 A 1 po
for A 1; q Mcs ( p ) ln( p ) for A 1 g ( p, q, po ) t p o (13) where po is the variable controlling the size of the plastic potential. The direction of plastic flow is therefore defined as m g g (14)
c ln p p ln R
0
(11)
p p, q c q q, p pc p , and where p [p, q]T is the kinematic hardening vector conc trolling the position of the loading surface. p is the isotropic hardening parameter controlling the size of the loading surface, and is determined by enforcing the constraint that the loading surface must be tangent to the local bounding surface at the centre of homology and pass through the current stress state . The unit normal vector at the image point defining the direction of loading is given using the general equation n f F f F (12)
In (13), the sign of t controls the direction of plastic flow, with t 1 corresponding to compressive loading (q 0) and t 1 for extensive loading (q 0). 4.4 Hardening modulus
Following the usual approach in bounding surface plasticity, the hardening modulus h is divided into two components. h hb hf (15)
q
+ M cs
Centre of homology
Loading Surface
Bounding Surface
where hb is the plastic modulus at on the bounding surface, and hf is some arbitrary modulus at , defined as a function of the distance between and . hb is determined by imposing the consistency condition at the bounding surface and incorporating the hardening effects of plastic volumetric strain and matric suction leading to hb mp F pc pc s p p v pc s v F (16)
Maximum Loading Surface
pc
n n
M cs
The modulus hf is defined such that it is zero on the bounding surface and infinity at the point of stress reversal. Following Khalili et al (2005) and Russell and Khalili (2006), hf for unsaturated soils is taken as p pc p pc s c hf t 1 km ( p ) p p c v s p v pc (17)
Figure 1. Mapping rule and loading surface for cyclic loading.
845
where q/p is the stress ratio, p is the slope of the peak strength line in the q p plane, and km is a material parameter controlling the steepness of the response in the q q plane. 5 ELASTO-PLASTIC STRESS-STRAIN RELATIONS
N I x1 B1 0 N I x2
0 N I x2
The elasto-plastic stress-strain relation for unsaturated soils is expressed as De mn T De ep De D h n T De m (18)
N I x1 N I x 1 and B3 N I x1
... T N ... , B2 I x1 ... ... . ...
N I x2
...
where De and Dep are the elastic stiffness matrix and the standard elasto-plastic stress-strain matrix of the soil, respectively. 6 FINITE ELEMENT FORMULATION
in which I 1,2, 3.. represents the number of nodes in the finite element. The nodal forces and nodal fluxes N T F d N T T d , are given by R

Qw N qw d ,

T
Qa N qa d , with

T
Applying the Galerkins approach, the spatial discretization of the governing equations (5), (8), and (9) is achieved as w 1 Cp a R Cp K u C
T
w 1 N p N u k k k k a , qw rw N pw , qa ra N p
Dep T
the boundary tractions and fluid influxes given by
N pa .
(19)
w H w pw a11 M p u (20)
The rate form of the discretized equations are integrated using the finite difference approach, such that the integration of an arbitrary function y over a time interval t is given by
t t t
a Qw a12 M p 1 C

T
1 yt yt t t y yt t y t dt
(22a)
w H a pa a21 M p u
y ytt yt (21)
(22b)
a Qa a22 M p
in which [K] is the element stiffness matrix, [C] is the coupling matrix, [M] is the mass matrix, [Hw] and [Ha] are flow matrices corresponding to the permeabilities of the water and air phases respectively, {u} is the vector of nodal displacements, {Pw} is the vector of nodal pore water pressures, {Pa} is the vector of nodal pore air pressures, {R} is the vector of nodal forces, {Qw} and {Qa} are vectors of nodal fluxes of the water and air flows respectively. The element matrices are evaluated using: [K ] [B1 ]T [Dep ][B1 ]d ,
where yt is the value of y at time t, and is a parameter controlling the type of interpolation. 0 for forward interpolation, 1 for backward interpolation, and 0.5 for central interpolation. Applying equation (22) to the primary variables in equations (19), (20) and (21) results in the following linearized form of the governing equations,
Ku Cpw (1 )Cpa R

(23) C
T
u a11 M pw
[ M ] [ N ]T [ N ] d , [ H w ] [Ha ]
kra [ B ]T [ k ][ B3 ] d , a 3
krw [ B ]T [ k ][ B3 ]d and w 3
H w pw pw t t a12 M pa Qw t Qw
with
(24)
846
1 C
u a21 M pw
H a pa pa t t a22 M pa Qa t Qa t
(25)
Equations (24) and (25) can be rearranged in a compact matrix-vector form as
Deviatoric Stress, q (kPa)
K T C 1 C T
C a11 M t H w a21 M
a12 M a22 M t H a
(1 ) C
(a)
Model Simulation
80
60
0 u R p w 0 1 t Qw t Q p 0 a a t

t Q w t t t H w p w t
40
20

Q a t t
H p a a t
(26)
0.002
0.004
0.006 0.008 Axial Strain, 1
0.01
0.012
(b)
NUMERICAL EXAMPLES
To demonstrate the application of the fully coupled model, the experimental data from a single cyclic test reported in the literature is simulated. No additional simulations could be included due to space limitations. Mun (2004) conducted undrained (constant water volume) cyclic triaxial tests on lightly compacted, unsaturated samples of silty sand to investigate the liquefaction potential of unsaturated soils. In these tests, the samples were compacted to a dry density of 1400 kg/m3 and moisture contents of 13 to 14% and then were isotropically consolidated to a net confining cell pressure of 120 kPa. Cyclic loading was applied through axial strain increments while controlling the peak deviatoric stress to a predetermined cyclic stress ratio. (The cyclic stress ratio is defined as the ratio of the maximum shear stress to the net confining stress.) Results of one test conducted at suction values of 80 kPa with cyclic stress ratios of 0.3 are presented in this section. As most of the parameters required in this model were not available in the data reported by Mun (2004), simulations were conducted using typical parameters for a silty sand. The elastic constant was assumed to be 0.004, and the parameters defining the shape of the bounding surface were taken as N 3 and R 7.5. The slope of the
(c)
0 0.002
Volumetric Strain, v
Model Simulation
0.004 0.006 0.008 0.01 0 0.002 0.004 0.006 0.008 Axial Strain, 1 0.01 0.012
(d)
Figure 2. Comparison of model simulation and experimental data for undrained cyclic triaxial test on an unsaturated sample of Texas silty sand at a constant suction of s 80 kPa and cyclic stress ratio of 0.30: (a) experimental q 1, (b) predicted q 2, (c) experimental v 1 and (d) predicted v 1 responses.
847
critical state line was taken as Mcs 1.32. Material parameters defining the isotropic compression line were (0) (s) 0.08 and N(0) N(s) 2.295. In this analysis, in line with the observations of Russell and Khalili (2006), the effect of suction hardening is assumed to be negligible and hence the same isotropic compression line was used the same for both the saturated and unsaturated states. The model constant A defining the plastic potential was taken as unity. Parameter km was taken as 140. The air entry and air expulsion values were assumed to be sae sex 10 kPa. The pore water pressure at the beginning of the test was 200 kPa while the pore air pressure was 280 kPa. The initial void ratio and degree of saturation were e 0.893 and Sr 0.34 for the samples tested at s 80 kPa. Figures 2a-d shows comparison between model simulation and experimental data. These figures compare q 1 and v 1 responses obtained from the model simulation against the experimental data. In all the results shown, the model simulations match the experimental data reasonably well. The model captures the successive stiffening of the sample with the progress of the cyclic loading and the mechanical hysteresis between unloading and reloading. It is interesting to note that all the samples were stable with no sign of tendency for liquefaction. During undrained loading of unsaturated soils, there is no volume change in the water phase but the total volume of the sample changes due to the compressibility of the air phase. This dramatically reduces the tendency for increase in pore water pressure minimizing the potential for liquefaction. 8 CONCLUSION
REFERENCES
Alonso, E.E., Lloret, A., Gens, A. & Yang, D.Q. 1995. Experimental behaviour of highly expansive double structure clay. In: Proceedings of the 1st international conference on unsaturated soils, Paris, Vol. 1, pp. 116. Bishop, A.W. 1959. The principle of effective stress. Teknisk Ukeblad 106 39: 859863. Buisson, M.S.R. & Wheeler, S.J. 2000. Inclusion of hydraulic hysteresis in a new elasto-plastic framework for unsaturated soils. In Experimental evidence and theoretical approaches in unsaturated soils (eds A. Tarantino and C. Mancuso), pp. 109119. Rotterdam: Balkema. Dafalias, Y.F. & Herrmann, L.R. 1980. A Bounding surface soil plasticity model. International Symposium on Soils under Cyclic and Transient Loading, Swansea. 335345. Gallipoli, G., Gens, A., Sharma, R. & Vaunat, J. 2003. An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour, Geotechnique 53; 123135. Jommi, C. 2000. Remarks on the constitutive modelling of unsaturated soils, Experimental Evidence and Theoretical Approaches in Unsaturated Soils, Proc. of Int. Workshop on Unsaturated Soil (Tarantino A and Mancuso C, eds), Balkema, Rotterdam, 139153. Khalili, N., Habte, M.A. & Valliappan, S. 2005. A bounding surface plasticity model for cyclic loading of granular soils. International Journal for Numerical Methods in Engineering, 63(14), 19391960. Khalili, N., Habte, M.A. & Zargarbashi, S. 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses, Computers and Geotechnics, 35: 872889. Khalili, N. & Zargarbashi, S. 2010. Influence of hydraulic hysteresis on effective stress in unsaturated soils, Gotechnique, in press. Mun, B. 2004. Unsaturated Soil Behavior Under Monotonic and Cyclic Stress States. PhD Thesis, Texas A & M University, Texas, USA. Russell, A.R. & Khalili, N. 2006. A unified bounding surface plasticity model for unsaturated soils. International Journal for Numerical and Analytical Methods in Geomechanics, 30(3): 181212. Sheng, D., Sloan, S.W. & Gens, A. 2004. A constitutive model for unsaturated soils: thermomechanical and computational aspects., Computational Mechanics; Vol. 33 (Number 6, May). Tamagnini, R. 2004. An extended Cam-clay model for unsaturated soils with hydraulic hysteresis. Gotechnique, 54:223228. Vaunat, J., Romero, E. & Jommi, C. 2000. An elastoplastic hydromechanical model for unsaturated soils, Experimental Evidence and Theoretical Approaches in Unsaturated Soils, Proc. of Int. Workshop on Unsaturated Soil (Tarantino A and Mancuso C, eds), Balkema, Rotterdam, 121138. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils, Geotechnique, 53(1): 4154.
The finite element formulation for a fully coupled constitutive model is presented for flow and deformation analysis of unsaturated soils including hydraulic and mechanical hystereses. The model is formulated incrementally using the effective stress principle. The elastic-plastic deformation behaviour is captured using bounding surface plasticity. Effect of suction on the stiffening of the soil response is taken by introducing it as an additional hardening parameter. Coupling between flow and deformation models is achieved through the effective stress parameter. A single set of material parameters is introduced for the complete characterization of the coupled constitutive model. The Galerkins approach and finite difference method are applied for spatial and time discretization of the governing equations. The application of model to a cyclic test form the literature is demonstrated. Good agreement is obtained between the simulation results and the test data.
848
A simple elasto-plastic model for unsaturated soils and interpretations of collapse and compaction behaviours
M. Kikumoto, H. Kyokawa, T. Nakai & H.M. Shahin
Nagoya Institute of Technology, Nagoya, Japan
ABSTRACT: Stress-strain behaviour of unsaturated soils seems rather different from that of saturated ones. It should however be described properly because soil may lie under an unsaturated state in the actual field. In the present study, a model taking account of the necessary characteristics of unsaturated soils is proposed by extending an existing model for saturated ones based on the Bishops effective stress. In the model, variation in the degree of saturation is estimated by a newly proposed model for water retention curve considering the effects of hydraulic hysteresis and density, and the increase in the degree of saturation is linked with downward movement of normally consolidation line in the mean effective stress vs. void ratio plane in order to describe the typical behaviours of unsaturated soils. In the present paper, mechanisms of consolidation, soaking and compaction behaviours of unsaturated soils are discussed through the simulations by the proposed model. 1 INTRODUCTION larger void ratio compared with the saturated one; the unsaturated sample however exhibits significant compression from a stress level of around 200 kPa, Sr then increases due to the volumetric compression (although suction is kept constant) and the compression line asymptotically approaches to the normally consolidation line of the saturated one (NCLsat). Meanwhile, Figure 2 shows the results of compression and soaking tests on air dried silt (Jennings and Burland, 1962). Although the air dried silts can stay in a looser region above the compression line of saturated soil during the compression process, they exhibit hydraulic collapse or stress relaxation behaviour until they come close to the NCLsat once samples are soaked under constant suction or void ratio. In the present study, a model for saturated soils (Nakai et al., 2010) is extended to one considering
0.80 Compression line of air-dried sample (8 tests)
It is usually indicated that suction increases the stiffness of soil. Unsaturated soils are hence able to exist in a looser state compared with saturated soil. Compression behaviour of unsaturated soil is compared with that of saturated one in Figure 1 (Honda, 2000), in which vertical component of the effective stress defined by Bishop (1959) as eq. (1) is plotted against void ratio e and degree of saturation Sr.
ij ij ua ij (ua uw ) ij ijnet Sr s ij
(1)
is a variable given as a function of Sr and is assumed to be equal to Sr here for simplicity. It is seen from Figure 1 that: the unsaturated sample initially shows relatively high stiffness and retains a
1.2
Void ratio e
Void ratio e
1.0 0.8 0.6 Sr 10
Sa
80
tur ate
Degree of saturation Sr [%]
Unsaturated sample (suction s = 73.5)
100
ds
0.75
am
ple
60
40 100 1000 Vertical effective stress v" [kPa]
Soaked at constant void ratio Soaked at constant 0.70 applied pressure Saturated sample Saturated sample (slurry) 0.65 0.1 1.0 10.0 Applied pressure [t/ft2]
30.0
Figure 1. Oedometer test under constant suction (drained water and exhausted air) on catalpo clay (after Honda, 2000).
Figure 2. One-dimensional compression and soaking tests on air dried silt (after Jennings and Burland, 1962).
849
the necessary characteristics of unsaturated soils. Unsaturated soil having a lower Sr is able to retain a larger void ratio so Sr is assumed to have a dominant effect on the position of the normally consolidation line in the proposed model. 2 ELASTOPLASTIC CONSTITUTIVE MODEL FOR SATURATED SOILS
The recoverable elastic change of void ratio ee is assumed to follow a usual elastic relationship as: e e ln
0
(2)
where is the swelling index. The total change of void ratio e and its plastic component are obtained: e e0 e ln
Outline of a model for saturated soils considering the effect of density and confining pressure (Nakai et al., 2010) is explained here. Figure 3 shows a typical e-ln relation of saturated, over consolidated soil. It is assumed in the model that: soil exhibits elastoplastic deformation even under over consolidated state and its state gradually approaches to the normally consolidation line (NCL) with increase in the stress level; soil having a larger value of a state variable , which is defined as the difference of void ratios of current state and normally consolidated state under same confining stress, shows stiffer behaviour. Figure 4 illustrates the variation of void ratio from the initial state I ( 0, e e0) to the current state P ( , e e). Here, eN0 and eN are void ratios on the NCL under same stress level.
0 0
(3)
e p ( e e e ) ( )ln
( 0 ) 0
(4)
where is the compression index. From eq. (4), a yield function f is introduced as equation (5). f ( )ln
( 0 ) e p 0 0
(5)
As the state variable may decrease with the development of the plastic deformation and finally converges to zero, eq. (6) is applied as an evolution law. d G()(dep) a(dep) (6)
eN
Void ratio e
NC
Lf
or
sat
monotonic decrease in (d< 0)
ura
ted
soi
Here, a is a material parameter controlling the effect of density. A one-dimensional constitutive relationship of saturated over consolidated soil is finally obtained from the consistency condition (df 0). d de (de p de e ) 1 G (7)
Logarithm of effective stress

Figure 3. Typical relationship between void ratio e and logarithm of effective stress in saturated over consolidated soil.
eN0
Void ratio e
0 e0 eN e I -e
NC L
As volumetric behaviours are mainly focused on in the present paper, formulation of the simplified, one-dimensional model has been explained. This model can however be extended to multi-dimensional one easily by adding a term of stress ratio to the yield function and assuming the associated flow rule. 3 MODEL FOR WATER RETENTION CURVE
for
sat
ura
ted
soi
ln(/0)
l
P 0 Logarithm of effective stress
Figure 4. Modelling of the volumetric behaviour of saturated, over consolidated soil.
It is well-known that the stress-strain behaviour of unsaturated soil is remarkably affected by the variation of Sr. A proper model for water retention curve is thus necessary to formulate a constitutive model for unsaturated soils. In this section, a model for water retention curve incorporating the effects of hysteresis and density is proposed.
850
Suction s s
Iw = 0
fd = 0 (drying) fc = 0 current state (s, Sr, Iw) Iw = 1 fw = 0 (wetting)
where is a material constant. From eqs. (9) to (11), a function fc representing the water retention surface through the current state (s, Sr, Iw) is given as follow. fc(s, Sr, Iw) IwFd (s) (1 Iw)Fw(s) Sr 0 (12)
Srw Sr Srd Degree of saturation Sr

Figure 5. Modelling of the hydraulic hysteresis.
Water retention curves trace different paths according to drying and wetting histories. Two curves in Figure 5 representing the highest and lowest boundaries of Sr, which are usually called as main drying and wetting curves, are firstly given by a classical equation for water retention curve proposed by van Genuchten (1980) as follows.
n fd Sr , s Smin Smax Smin 1 d s m
Sr (8)
Fd s Sr 0 fw Sr , s Smin Smax

n Smin 1 w s m
As any hydraulic state of unsaturated soil always locates on this surface fc 0, an incremental form of the water retention curve considering the influence of the hydraulic hysteresis is given by applying the consistency condition (dfc 0) to eq. (12). It is also indicated through the past experimental studies (e.g. Tarantino and Tombolato, 2005) that volumetric behaviour also influences the water retention characteristics and denser soil tends to retain higher Sr. The increase of Sr under a constant suction (Figure 1) may therefore be a consequence of the volumetric compression. In the present study, a modified suction s*, which is dependent on both ordinary suction s and void ratio e, is proposed to describe the effect of volume change. A tentative form of s* considering the effect of density is given as: e s* s Nsat
e
(13)
Sr (9)
Fw s Sr 0
Smin, Smax, n, m and are material parameters and the superscripts d and w denote the main drying and wetting curves, respectively. As any state of water retention lies between two main curves, a variable Iw defined as the ratio of interior division of the current state between two reference states on the main curves is introduced as an intermediary of the model. I w (Sr Srw )(Srd Srw ) 1 (10)
where e is a parameter controlling the effect of density and Nsat is the reference void ratio of saturated, normally consolidated soil under atmospheric pressure. By applying the modified suction s* to eq. (12) in stead of the ordinary suction s, the effect of the volume change can be automatically incorporated. Material parameters for two soils are summarised in Table 1. Calculated results by the proposed model are compared with experimental ones (Huang, 2005) in Figure 6. It is known that the proposed model accurately predicts the hysteretic water retention behaviour under complicated suction histories. It is also seen from Figure 7 showing the calculated water retention curves of samples of
Table 1. Parameters for water retention curve.
White Fujinomori silica sand clay 1.00 Smax 1.00 0.20 0.1 Smin 0.1 0.03 [1/kPa] parameters for van Genuchten model 0.3 0.10 [1/kPa] 0.7 0.1 m 4.7 2.0 n influence of suction histories 70.0 100.0 influence of void ratio 1.2 5.0 e
Here, Srd and Srw are degrees of saturation of the reference states on the main drying and wetting curves under current suction, respectively. Iw bound by 0 for the main wetting curve and 1 for the main drying curve provides a simplified, scalar representation of the current state of hydraulic hysteresis. It increases monotonically from 0 to 1 as Sr decreases, and vice versa. An evolution law for Iw is thus given as:
3 1 I w dSr when dSr 0 dI w 3 I w dSr when dSr 0
(11)
851
25 20
plot: experiment line: simulation
15 10 5
B D
0 0
A B C D A
20 40 60 80 Degree of saturation Sr [%]
C
100
Figure 6. Water retention curve of White silica sand (after Huang, 2005) and corresponding results of simulation.
25
eNC = 0.830
20
15 10 5
e = 0.830 e = 0.700 e = 0.635

20 40 60 80 Degree of saturation Sr [%] 100
Firstly, we propose to formulate the model based on the Bishops single effective stress defined by eq. (1). It has been indicated by the past experimental studies (e.g. Sivakumar, 1993; Khalili et al., 1998) that critical state line is uniquely defined in the Bishops mean effective stress vs. deviator stress plane regardless of the degree of saturation Sr. It has also been mentioned through the experimental evidences (Kawai et al., 2000) that, provided that the effective degree of saturation Sre (Karube et al., 1996) is applied as the effective stress parameter , a unique relationship between stress ratio of Bishops stress and dilatancy exists irrespective of Sr. It can hence be stated that a unique critical state friction angle and a unique stress-dilatancy relationship can be assumed in a model for unsaturated soils if it is formulated based on the Bishops effective stress. Jennings and Burland (1962) however pointed out that collapse phenomena of unsaturated soils due to soaking (e.g. Figure 2) cannot be explained by the Bishops single effective stress. An extra mechanism is accordingly added in the proposed model to describe the effect of the variation of Sr on the volumetric behaviour. Because soil having a lower Sr can retain larger void ratio and compression line gradually falls toward that of saturated one due to the increase in Sr, NCL of unsaturated soil is shifted downward (or upward) in the plane of effective confining stress '' and void ratio e depending on the increase (or decrease) of Sr in the proposed model. In order to represent the current changing position of the NCL, a variable is defined as vertical distance of the current NCL from that for the saturated one (Figure 8). The variable varies with the change of Sr and is assumed to be incrementally proportional to Sr with satisfying 0 at saturated condition: (Sr) l(1 Sr) (14)
Suction s [kPa]
Suction s [kPa]
00
Figure 7. Calculated water retention curves of White silica sands of different densities.
where l is a material parameter representing the vertical distance of the NCLs for dried and saturated samples in the compression plane. Figure 9
Nsat+l Void ratio e Nsat+ Nsat (Sr)
different densities that the proposed model properly describes the influence of density. In addition, as s* varies due to the variation of void ratio as well as suction, the increase in Sr induced by the volumetric compression such as shown in Figure 1 can be described. 4 ELASTOPLASTIC CONSTITUTIVE MODEL FOR UNSATURATED SOILS
NC NC
NC Lf or
Lf
or
uns sat
dri rat
Lf
atu
ed
soi soi
or
ed
l (
=
ura
ted
l ( =
l) l(1 -S )
soi
= 0)
l (
In this section, the one-dimensional model for saturated soil introduced in section 2 will be extended to one incorporating the unsaturated soil behaviour.
1 Pa Logarithm of effective stress ''
Figure 8. Effect of Sr on the position of NCL and its description through the state variable (Sr).
852
eN0
Void ratio e
Table 2. Parameters for stress-strain characteristics.
e0 eN
0
ln(/0)
NC L( 0 = -e 0) NC L( e = P ) 0'' '' Logarithm of effective stress ''
Nsat a l
0.104 compression Index 0.01 swelling Index 0.83 reference void ratio of saturated, normally consolidated sample under '' = 98kPa effect of density and confining pressure 100 effect of the degree of saturation on the 0.5 position of NCL
1.0 0.8 Void ratio e 0.6 0.4 0.2 100 Degree of saturation Sr[%] 90 80 70 60
Initial state NC L
sat
Figure 9. Modelling of the volumetric behaviour of unsaturated, over consolidated soil.
(A, B, C, D) (a) s = 294 [kPa] s = 196 [kPa] s = 98 [kPa] s = 49 [kPa] 100 Suction loading/Drying Initial state (A)s = 49kPa (B)s = 98kPa (C)s = 196kPa (D)s = 294kPa (b) s = 294 [kPa] s = 196 [kPa] s = 98 [kPa] s = 49 [kPa] 10000 1000 Net stress net [kPa] 10000
shows the variation of void ratio of an over consolidated, unsaturated soil from the initial state I (" "0, e e0, 0) to the current state P (" ", e e, ). The change in void ratio -e of an over consolidated, unsaturated soil is given as equation (15). e e0 e ln
( 0 ) ( 0 ) 0
(15)
Comparing eqs. (15) with eq. (3), the term (0 ) representing the effect of Sr through the vertical shift of the NCL is added. Assuming a nonlinear elastic relation given by eq. (2), the plastic component of the change of void ratio ep is obtained and a yield function f of the general elastoplastic theory is consequently introduced. f ( )ln
100
( 0 ) ( 0 ) e p 0 0 (16)
1000 Net stress net [kPa]
Figure 10. tion tests.
Simulations of constant suction consolida-
From the consistency condition (df 0) and eqs. (6) and (14), a one-dimensional constitutive relationship for unsaturated over consolidated soils is obtained. de (de p de e ) d d dSr G ( ) 1 dSr 1 G ( ) (17)
Loading condition of the one-dimensional model is simply given by assuming that the plastic component of the variation of void ratio is always compressive. 5 SIMULATIONS
Three series of simulations are performed here. The analyses have been carried out using parameters for Fujinomori clay listed in Tables 1 and 2.
First, consolidation tests have been conducted on initially over consolidated sample (e0 0.8) (Figure 10). In the calculation, suction s is increased from zero to prescribed values (49, 98, 196, 294 kPa) under constant net stress net of 49 kPa and consolidation tests are then carried out under constant s. It is indicated that the proposed model can describe the typical unsaturated soil behaviours (e.g. Figure 1) that unsaturated samples initially exhibit relatively high stiffness, they are able to stay in a region above the NCLsat, but their compression lines finally converges to the NCLsat with the increase in Sr. It is also known that the proposed model for water retention curve can predict the increase in Sr due to volumetric contraction though the suction remains constant. It is also known that the proposed model can consider the experimental tendency reported by Wheeler and
853
1.0 0.8 Void ratio e
NC
sat
consolidation
0.9
Sr
73.0 73.2 73.7 (a) 72.9 76.1 net = 49 [kPa] 0.6 80.5 net = 294 [kPa] net = 588 [kPa] net = 1017 [kPa] 0.4 net = 1960 [kPa] net = 2940 [kPa] net = 9800 [kPa] 0.2 100 1000 Net stress net [kPa]
Initial state drying process com p
acti
Void ratio e
0.8 0.7 0.6
on (
A (w = 21.5%) B (w = 23.2%) C (w = 24.9%)

con stan t wa ter con
unloading
tent
98.8
Final state 0 200 Initial state

unloading Final state
(a) 400 600 Total stress [kPa] 800
10000 9800
100
Degree of saturation Sr [%]
Degree of saturation Sr
90
ecr
eas
e in
s)
90 80 70 60
80
2940 1960 1017 588 net 294 49 50 100 150 200 Suction s [kPa] 250 300
drying
soa
kin
100
consolidation
g (d
(b)
70
(b) 0
200
Figure 11. Simulations of constant suction consolidation and subsequent soaking tests (initial void ratio e0 0.8).
0.14 0.12 0.10
.9
67.3 66.7 68.6 73.1 78.8 73.7 76.1 73.2 80.5 Sr before soaking
t stan con on ( ent) i t c t n pa com ater co w 400 600 Total stress [kPa]
A (w = 21.5%) B (w = 23.2%) C (w = 24.9%) 800
Figure 13. Simulations of compaction tests under constant water content (max 800 kPa, w 21.5, 23.2, 24.9%).
Decrease of void ratio e during soaking
compression
0.08 0.06 0.04 0.02 0.00 -0.02

66.5
= e0
0.8
73.0 98.8 98.7 10000
100 72.9 expansion
1000 Net stress net [kPa]
Figure 12. Decreases in void ratio of dense and loose samples (e0 0.8, 0.9) due to soaking process.
Sivakumar (1995) and Kayadelen (2008) that soil having lower Sr shows stiffer behaviour. Second, soaking tests have been performed on two kinds of samples (initial void ratio e0 0.8, 0.9). In the simulation, suction s is firstly increased from zero to 294 kPa under constant net stress net of 49 kPa, net is then increased to prescribed value (49, 294, 588, 1017, 1960, 2940, 9800 kPa) under constant s of 294 kPa and samples are finally soaked by decreasing s to zero. It is seen from Figure 11 that the proposed model is able to simulate typical soaking collapse behaviour such as shown in Figure 2. The volume changes due to soaking process are plotted against net stress level in Figure 12. The proposed model properly predicts the magnitude of the hydraulic collapse considering
the effects of stress level and density. The proposed model can also describe the soaking-induced expansion of heavily over consolidated soil reported by Sun et al. (2007). Third, compaction behaviour has been investigated (Figures 13 and 14). Suction s is firstly increased under constant applied stress of 50 kPa until prescribed water content w is achieved. Because compaction of soil is generally regarded as the expulsion of entrapped air without significant drainage of void water, total stress is increased from 50 kPa to predetermined maximum value (max 100, 200, 400, 800, 1600, 3200, 6400 kPa) with keeping constant water content and back again in this calculation. Figure 13 shows compaction behaviours of three samples of different water content w. It is seen from this figure that: soil sample having larger water content is more compressible in the beginning stage of compaction; soil is hardly compressed regardless of its water content once it is saturated as the mean effective stress hardly increases. Results of 80 compaction tests, in which water content w and maximum applied stress max are varied, are summarised in terms of dry density and water content in Figure 14. It is pointed out from this figure that the proposed model can express typical compaction curves having optimum water content and it can also consider the transition of compaction curves owing to the increase in the applied pressure.
=0
854
2.0 1.9
Applied maximum stress
Dry density d [g/cm3]
1.8 1.7 1.6

B
50 kPa 100 kPa 200 kPa 400 kPa

Z er oa ir v o id
C
800 kPa 1600 kPa 3200 kPa 6400 kPa
stress space to determine the plastic strain components. Using modified stress tij (Nakai and Mihara, 1984) instead of ordinary stress ij, the influence of intermediate principal stress on the deformation and strength of soils is suitably and automatically considered. REFERENCES
1.5 1.4 1.3 15
sc u rv e S r = 100
80 %
Sr
=6
0%
70 %
90 %
20 25 Water content w [%]
30
Figure 14. Compaction curves for several kinds of applied maximum stresses (final states of compaction tests).
CONCLUSION
A simple model incorporating necessary descriptions of several aspects of constitutive behaviour of unsaturated soils has been developed. The main issues in modelling the unsaturated soil behaviour are: a) volumetric behaviour related with the variation in Sr; b) hydraulic behaviour associated with saturation, suction, hydraulic hysteresis or void ratio changes. The proposed constitutive model is formulated using the Bishops effective stress and normally consolidation line which is shifted downward or upward in the void ratio e and logarithm of confining pressure plane depending the increase or decrease in Sr of saturation. A simple model for water retention curves considering the effects of hydraulic hysteresis and void ratio has also been proposed and applied. Validity of the proposed model is checked through a series of simulations of consolidation tests, soaking tests and compaction tests on unsaturated soils. It is indicated through the simulations that the proposed model suitably describes typical behaviours of unsaturated soils such as soaking collapse phenomena. In addition, the proposed model can also express the wellknown compaction curves of soil including their transition due to the increase in the applied stress. It is easy to extend the proposed one-dimensional elastoplastic model to the multi-dimensional one considering the ordinary stress tensor ij and strain tensor ij. The yield function f in Equations (5) and (16) need to be formulated using mean stress p and deviator stress q (or stress ratio q/p), and flow rule is assumed in the ordinary
Honda, M. 2000. Ph.D Thesis, Kobe univ. (in Japanese). Bishop, A.W. 1959. The principle of effective stress. Tecnisk Ukevblad 39: 859863. Jennings, J.E.B. and Burland, J.B. 1962. Limitations to the use of effective stresses in partly saturated soils. Gotechnique 12(2): 125144. Nakai, T. et al. 2009. Simple and unified method for describing various characteristics of geomaterials. J. Applied Mech. 12: 371382 (in Japanese). Van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44: 892898. Huang, H.C. et al. 2005. A novel hysteresis model in unsaturated soils. Hydrol. Process. 19: 16531665. Tarantino, A. and Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique 55(4): 307317. Sivakumar, V. 1993. A critical state framework for unsaturated soils. Ph.D. Thesis, Univ. of Sheffield. Khalili, N. et al. 2004. Effective stress in unsaturated soils: review and new evidence. Int. J. Geomech. 4(2): 115126. Kawai, K. et al. 2000. Behavior of unsaturated soil and water characteristics in undrained shear. Research paper of research center for urban safety and security, Kobe Univ. 4: 231239 (in Japanese). Karube, T. et al. 1996. The relationship between the mechanical behavior and the state of pore water in unsaturated soil. J. Geotech. Eng. No.535/III-34: 8392 (in Japanese). Vanapalli, S.K. et al. 1996. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Gotechnique 49(2): 143159. Sugii, T. et al. (2002) Relationship between soil-water characteristic curve and void ratio, Proc. 3rd Unsaturated Soils Conf., 209214. Wheeler, S.J. and Sivakumar, V. 1995. A elasto-plastic critical state framework for unsaturated soil. Gotechnique 45(1): 3553. Kayadelen, C. 2008. The consolidation characteristics of an unsaturated compacted soil. Environ. Geol. 54: 325334. Sun, De'an et al. 2007. Collapse behaviour of unsaturated compacted soil with different initial densities. Can. Geotech. J. 44: 673686. Nakai, T. and Mihara, Y. 1984. A new mechanical quantity for soils and its application to elastoplastic constitutive models. Soils and Foundations 24(2): 8294.
855
An introduction of jumped kinematic hardening rule to elastoplastic model for unsaturated geo-materials
Y. Kohgo
Tokyo University of Agriculture and Technology, Fuchu, Tokyo, Japan
ABSTRACT: An extended subloading surface model on the base of jumped kinematic hardening rule concept was proposed. In the concept, centers of similarity of subloading surfaces to the normal yield surfaces may jump to the reversed stress points whenever reverses of loadings occur. The model was introduced to the generalized elastoplastic model for unsaturated geo-materials proposed by the author. Oedometer and triaxial compression tests for an unsaturated rockfill were tried to be simulated using the elastoplastic model. The elastoplastic model could simulate the hystereses of loading and unloading, and different stress-strain curves with different suction values. However more experimental simulations are necessary to verify the model, especially in the shear process. 1 INTRODUCTION conduct simulations of some laboratory tests for a rock material to investigate performances of the model. 2 JUMPED KINEMATIC HARDENING RULE
It is very important to estimate permanent deformations of soil structures such as embankment dams due to earthquake motions in the performancebased design. These soil structures are normally constructed by compacting geo-materials. The compacted geo-materials are usually under unsaturated conditions. Then the elastoplastic models used in the design should be capable to predict properties of unsaturated geo-materials. However few research works concerned with constitutive modeling for unsaturated geo-materials have been conducted (Oldecop and Alonso, 2001; Kohgo 2007a). Kohgo et al. (1993b, 2007a) have proposed elastoplastic models based on the cyclic plasticity concept. However, as the models were described on the base of the original subloading surface model (Hashiguchi, 1980), the model could not express hysteresis loops of stress-strain relationship due to cyclic loads, and then could not estimate correctly the permanent deformations. Hashiguchi (1989) proposed an extended subloading surface model with a combined hardening rule to overcome the shortages of the original subloading surface model. However, it was very complicate and difficult to apply the model to the real boundary value problems. Then we have proposed a simplified model that belonged to the extended subloading surface model (Kohgo et al. 2007b). In the model, a new hardening concept named jumped kinematic hardening rule was used. The concept was incorporated into the generalized elastoplastic model for unsaturated geo-materials proposed by Kohgo et al. (2007a). Here we will
In the following, compression stresses are assumed to be negative. In the subloading surface model, loading surfaces named subloading surface are newly introduced within yield surfaces (normal yield surfaces) defined in classic plasticity theory. Then smooth transition from elasticity to plasticity can be established. The subloading surfaces are defined to be similar to the normal yield surfaces. The center of similarity is fixed the origin of stress spaces in original subloading model, while it can move within the yield surfaces in the extended subloading surface model. In this model, a new hardening rule for the center of similarity is introduced. We will name the hardening rule jumped kinematic hardening rule. In the rule, the following assumptions are used: 1) the centers of similarity can move whenever reverses of loadings occur, and 2) the points moved are consistent with the stress points at which reverses of loadings occur. Other assumptions: 3) the current stress point is always on the current subloading surfaces, and 4) the subloading surfaces keep the similar shape to the normal yield surfaces, are the same as those of the original extended subloading surface model. Figure 1 shows the schematic explanation of this concept. For simplicity, subloading and normal yield surfaces are assumed to be elliptical. It is also
857

m The similarity ratio Rn is defined as (see Fig.2)

m Rn
m m ijn ijn m ijn m
m (0 R n 1),
(1)

where tensor ijn coordinate tensor of the origin of local coordinate system for subloading surface (back stress), and superscript m and subscript n express number of reverses of loading and number of load increments, respectively. The judgement whether the reverse of loading occurred is conducted by the increment of similarity ratio. If the increment is negative, the reverse of loading will occur. Then,
m dR n 0

Reverse of loading,
(2) (3)

m m m dR n Rn Rn1, m dRn
increment of similarity ratio. where As the center of similarity should be always present within the normal yield surface, it may be adjusted using the following equation (see Fig. 3):
m ijn 1 m m I cn 1 ijn m I cn
Figure 1. Explanation of jumped kinematic hardening subloading surface model. (a) along the 0th loading path, (b) along the 1st loading path, and (c) along the 2nd loading path.
(4)
J2 1/2 Normal yield surface

m ijn
postulated to ignore the hardening and softening of the normal yield surface with loadings. The vertical and horizontal axes in Fig. 1 are the square root of the second stress invariant of deviator stresses (J2) and the first stress invariant (I1), respectively. Supposing that the loading starts from the origin (ij0: coordinate of origin) that is also the center of similarity. The initial subloading surface (SL0) may expand with the loading. Here the tensor ij0 is the coordinate of the center of similarity and the superscript of the tensor denotes number of reverses of loadings. If the loading reversed at the point with the stress tensor ij1, the new center of similarity should be consistent with the stress tensor: ij1 ij1 (see Fig. 1(a)). This operation may be continuously done whenever the reverses of loadings occur. This operation means that once a reverse of loading occurs, the existing subloading surface disappears at the reverse point and the new subloading surface (SL1 or SL2 shown in Figures 1(b) and (c)) may expand from the reverse point with the reversed direction loading. When the stress point encounters the normal yield surface, the calculation of plastic stresses will be conducted under the manner of the classic plasticity theory. The generalized return mapping algorithm (Ortiz and Simo, 1986) is used in the calculation process.
0 ij
m ijn
m ijn
Stress point Subloading surface

m I cn
m ijn m ijn
-I1
Conjugate point
Figure 2.
Back stresses and conjugate stresses.
J2 1/2
Normal yield Surfaces Similarity centers
ijmn+1
ijmn
ijmn
ijmn+1
-I1
Loading path
Subloading surfaces
Figure 3. Relative movements between normal yield surfaces and loading surfaces when the normal yield surfaces shrink during unloading processes.
858
where Ic yield stress (see Figs. 2). The relationship between the conjugate point and the back stress is given by the following equations (see Fig. 2),
m ijn m ijn m Rn
strain tensor; d pij increment of plastic strain q f tensor; n ij unit normal vector to yield ij and n function and plastic potential function at conjugate point. 3 MODELING OF SUCTION EFFECTS
(5) (6)
m m ijn ijn , m ijn
ij conjugate effective stress; ijn m current where m current effective effective stress tensor; and ijn stress tensor with respect to the local coordinate system. The hardening modulus H under isotropic hardening is assumed to be: ln R m , H H h n (7)
hardening modulus at the conjugate where H point, which lies on the normal yield surfaces and has the same outward normal as the current stress point does, shown in Figure 2; and h material is given by parameter. The value of H f , H p ij ij (8)
Kohgo et al. (2007a) discussed the difference of physical mechanism on unsaturated behavior between rockfills and soils. Consequently the mechanical properties of unsaturated geo-materials may be formulated by considering two suction effects: 1) an increase in suction increases effective stresses, and 2) an increase in suction enhances yield stresses and affects the resistance to interparticle slides (plastic deformations). The first suction effect controls changes of the shear strength for lightly overconsolidated or normally consolidated samples due to suction, volume reductions due to an increase in suction and swellings due to a decrease in suction. The second suction effect controls both changes of the shear strength for heavily overconsolidated samples due to suction and volume reductions due to a decrease in suction, namely saturation collapse (Kohgo et al., 1993a, 2007a). The first suction effect postulates to be estimated by the following effective stress equations.
p where ij plastic strain tensor; f normal yield function; and plastic potential function. The elastoplastic stress can be calculated on the basis of the generalized plastic theory (Dafalias and Hermann, 1982) using the stress gradients at conjugate stress point. The process is as follows: e p d ij D ijkl d kl d kl , p p ij d ij d n ,
ij ij ij ueq ,
ueq ua s (s se), a s* ueq ua se e s * ae s uauw, s* sse,
(14) (15)
s se ,
(16) (17) (18)
(9) (10)
d
e f q ab cd H Dabcd n n
f e kl n Dklqr d qr
(11)
f n ij
f ij
ff kl kl ij
12
(12)
q n ij
where 'ij effective stress tensor; ij total stress tensor; ij Kronecker delta; ueq equivalent pore pressure; ua pore air pressure; uw pore water pressure; s suction; s* effective suction; se air entry suction; ae material parameter; and brackets denote the operation z 0 at z 0 and z z at z 0. The second suction effect can be evaluated by formulating the state surface that expresses elastoplastic volume change behavior (Kohgo et al., 1993a). Here the following equations are used to estimate the state surfaces (Kohgo et al., 2007a). e * log(p) *, when s* s*m, (19)
kl kl
12
(13)
where d ij increment of effective stress tensor; Deijkl elastic modulus tensor; dij increment of
*
*f1s* , s* a*1
(20)
859
0 * e01
e
0 01
(21)
f *cs I1 f1 * I1
g ( ) J2
J2
0,
(24) (25) (26) (27)
when s* s*m,
* *m
0 * e02
f2 ( s s m ) , ( s * s * m ) a*2
* m 0 e02 *
g ( )
K * 0, J2 a 2b2 0, g ( )2
(22) (23)
f2 b2 ( I1 I 0 )2 a 2 g ( )
3 sin 3 cos sin sin
where * slope of e-log (p) curves; * void ratio of e-log (-p) curves at p unity; e void ratio; p mean effective stress; and are values of * and * at saturation, respectively; s*m effective suction where the state surface has the minimum or maximum slope values: * *m and * *m at s* s*m; and e001, e002, *f1, *f2, a*1 and a*2 material parameters. 4 GENERALIZED ELASTOPLASTIC MODEL FOR UNSATURATED GEOMATERIALS
where Lode angle; internal friction angle at failure with respect to effective stress; and I0, a, b, K*, P2, * and *cs are defined in Figure 4(a). The values of I0, a, b, K*, P2 are function of Ic. They are given by, I0 Ic/(1 R), a I0 Ic RI0, b *cs I0, K* (*cs *)I0 P2 ( *cs *)I 0 * 2 sin cs 3 (3 sin cs ) 2 sin 3 (3 sin ) (28) (29) (30) (31) (32) (33) (34)
The model described here is modified based on Kohgos model (Kohgo et al., 2007a). At first we have to define normal yield and loading surfaces. This model has two normal yield surfaces as shown in Figure 4(a). One is the Mohr-Coulomb type failure surface (f1) and the other is the elliptical cap model with corners (f2). Both are connected on the critical state line (f). They are:
*cs *
J21/2 Critical state line f

f1
1
where R material parameter and cs internal friction angle at critical state with respect to effective stress. Kohgo et al. (1993b) presented the details. The following two elliptical plastic potential functions are adopted as shown in Figure 4 (b). They are connected on the phase transformation line.
2 2 2 1 b *2 I1 P 1 a1 J 2 a1 b * 0, 2 2 2 2 2 b *2 I1 P 1 a2 J 2 a2 b * 0, 2
cs
(a)
f2 Ic
(35) (36)
K* P2
b
I0
a =-RI0
-I1
J21/2 Phase transformation line

1
where a1, a2, b* and P1 are defined in Figure 4(b). They are P1 Ic /(1 R1), a1 P2 P1, a2 P1 Ic, b* ptP1, 2 sin pt pt at 1 2 3 , 3 (3 sin pt ) (37) (38) (39) (40) (41) (42)
pt
(b)
2
1
b* P1 P2 a1 a2=-R1P1
Ic
-I1
Figure 4. Generalized elastoplastic model. (a) Normal yield functions, (b) and plastic potential functions.
pt
2 sin pt 3 (3 sin pt )
at 1 2 3 ,
860
pt
tan pt 9 12 tan2 pt
at plane strain condition, (43)
Table 1. rockfill.
Material parameters used for an unsaturated
where pt phase transformation angle; and R1 material parameter. Supposing this is an isotropic hardening model with plastic volumetric strain vp as a hardening parameter, the yield stress Ic may be evaluated by means of the state surface concept (Kohgo et al., 1993a). Obtaining the value of Ic, we can easily calculate I0, a, b, K*, a1, a2, P1, P2 and b* by using the relationships shown in Figure 4.
0.9
SIMULATION OF LABORATORY TESTS

Void ratio
(a)
0.8
Some laboratory tests are simulated using the elastoplastic model proposed here. Finite element method was used for the simulations. Tangential model (Kohgo, 2008) was used to model the soil water retention curve (SWRC). More details about the simulations can be seen in the reference (Kohgo et al 2007a). Here elastic properties are postulated to be evaluated as follows: K
0.7
0.6
0.5
0.4
Sr0=25% Experiment Sr0=25% Predict Predict
0.3 1 10 100 1000 10000
G Gi p p,
2.3 1 e0
Vertical stress (kPa)
Ki ,
(44) (45)
Void ratio
0.9
(b)
0.8
where K bulk modulus; G shear modulus; slope of e-log (p) curves at unloading; e0 initial void ratio; and Ki, Gi and p material parameters. The material parameters identified are shown in Table 1. The parameters concerned with unsaturated properties can be directly obtained from the triaxial compression tests and oedometer tests. The details are shown in the reference (Kohgo et al., 1995). The parameters for SWRC are the same as those used in the reference (Kohgo et al. 2007a). Simulation results of oedometer and triaxial compression tests for an unsaturated rockfill are presented in Figures 57. Figure 5 shows the simulation results of oedometer tests. The test results for two specimens with different initial degree of saturation (Sr0 25 and 48%) were simulated. In these figures, symbols with broken lines and thick solid lines denote experimental and simulation results, respectively. The both simulation results well expressed the whole void ratio-vertical stress relations of experimental ones Figures 6 show those enlarged unloading and reloading cyclic parts. Figure 6(a) and (b) are associated with Figures 5(a) and (b), respectively. The unloading and reloading cycles of experimental tests drew smaller loops than the simulations did. The simulation results drew smooth loops of unloading and
0.7
0.6
0.5
Sr0=48% Experiment
0.4
Sr0=48% Predict
0.3 1 10 100 1000 10000
Figure 5. Simulation results of oedometer tests for an unsaturated rockfill.
reloading, and expressed smooth transitions to the normal consolidation lines. Figure 7 shows the simulation result of a drained triaxial compression test. The specimen had a confining pressure 3 300 kPa. Figures 7(a) and (b) show the deviator stress q vs. axial strain a and volumetric strain v vs. a relationships, respectively. In these figures, broken lines and thick solid lines express experimental and simulation results, respectively. The whole q-a relation of the experimental test was well simulated. The slope of the unloading-reloading line of the experimental test was almost consistent with those of the simulation. Meanwhile in the v-a relationship, amount of
861
0.75
(b) (a)
0.7
0.65
0.6
Sr0=25% Experiment Sr0=25% Predict
0.55 1 0.7 10 100 1000 10000
Vertical stress (kPa) (d) (b)
0.65
0.6
0.55
Sr0=48% Experiment Sr0=48% Predict

0.5 1 10 100 1000 10000
was proposed. In the model we adopted a new hardening concept named jumped kinematic hardening rule. The rule was incorporated into a generalized elastoplastic model with subloading surface proposed by the author. In the concept, centers of similarity of subloading surfaces to the normal yield surfaces may jump to the reversed stress points whenever reverses of loadings occur. So the anisotropic behavior due to cyclic loadings can be evaluated by the concept. The cyclic properties can be modeled by only one parameter h. that was adopted in the previous model (Kohgo et al. 2007a). Simulations of oedometer and triaxial compression tests for an unsaturated rockfill were conducted to verify the model. The simulation results in oedometer tests could successfully express cyclic properties. However in the triaxial compression test, some more investigations concerned with plastic potential functions and hardening procedures will be necessary. The verifications were only restricted for a few laboratory tests. We will try to apply the model to other kinds of geo-materials. REFERENCES
Dafalias, Y.F. & Herrmann, L.R. 1982. Bounding surface formulation of soil plasticity, Chapter 10 in Soil MechanicsTransient and Cyclic Loads, edited by Pande, G.N. and Zienkiewicz, O.C., John Wiley & Sons. Hashiguchi, K. 1980. Constitutive equations of elastoplastic materials with elastic-plastic transition, J. Appl. Mech. (ASME) 47, 266272. Hashiguchi, K. 1989. Subloading surface model in unconventional plasticity, Int. J. Solids Struct. 25, 917945. Kohgo, Y. 2008. A hysteresis model of soil water retention curves based on bounding surface concept, Soils and Foundations 48(5), 633640. Kohgo, Y., Asano, I. & Hayashida, Y. 2007a. An elastoplastic model for unsaturated rockfills and its simulations of laboratory tests, Soils and Foundations 47(5), 919929. Kohgo, Y., Nakano, M. & Miyazaki, T. 1993a. Theoretical aspects of constitutive modeling for unsaturated soils, Soils and Foundations 33(4), 4963. Kohgo, Y., Nakano, M. & Miyazaki, T. 1993b. Verification of the generalized elastoplastic model for unsaturated soils, Soils and Foundations 33(4), 6473. Kohgo, Y., Nakano, M. & Miyazaki, T. 1995. Closuer Verification of the generalized elastoplastic model for unsaturated soils, Soils and Foundations 35(1), 173174. Kohgo, Y., Hayashida, Y. & Asano, I. 2007b. A cyclic plasticity model for unsaturated soils, Proc. of the 3rd Asian Conf. on Unsaturated Soils, UNSAT-ASIA 2007, Nanjing China, Science Press, 365370. Oldecop, L.A. & Alonso, E.E. 2001. A model for rockfill compressibility, Geotechnique 51(2) 127139. Ortiz, M. & Simo, J.C. 1986. An analysis of a new class of integration algorithms for elastoplastic constitutive relations, Int. J. Numer. Meth. Eng. 23, 353366.
Void ratio
Void ratio
Figure 6. Enlarged unloading and reloading cyclic parts in Figure 5.

1200
(a)
Deviator stress (kPa)
1000 800 600 400 200 0 0

0
Simulation Experimental
10
12
14
Axial strain (%)

(b)
-2
Simulation Experimental
-4
-6
-8 0 2 4 6 8 10 12 14
Axial strain (%)
Figure 7. Simulation results of triaxial compression tests for an unsaturated rockfill, (a) deviator stress vs. axial strain and (b) volumetric strain vs. axial strain.
volume change (reduction) simulated was larger than that of the experimental test. The both v-a relationships during the cyclic expressed a different manner. This point should be more investigated. 6 CONCLUSIONS
An elastoplastic model, which can express unsaturated geo-material behavior during cyclic loadings,
862
An effective stress based model for the dependency of a water retention curve on void ratio
D. Man
Charles University in Prague, Czech Republic
ABSTRACT: The paper presents a model for the dependency of a water retention curve (WRC) on void ratio. The approach is based on the effective stress principle for unsaturated soils and several underlying assumptions. The model describes the unsaturated soil behaviour along the main drying and wetting branches of WRC, it therefore does not incorporate the effects of hydraulic hysteresis. It leads to the dependency of a water retention curve (WRC) on void ratio, which does not require any material parameters apart from the parameters specifying WRC for the reference void ratio. Its predictive capabilities are demonstrated by comparing predictions with the experimental data on several different soils.
INTRODUCTION
Water retention curve (WRC), which quantifies the dependency of a degree of saturation Sr on suction s in unsaturated soils, is not unique for the soil of particular granulometry and mineralogy. WRC depends on suction path, it is thus different for wetting and drying processes. In addition, the main drying and wetting branches of WRC depend on the actual void ratio e. The importance of considering the dependency of WRC on e has currently been well recognised and it has been incorporated into a number of recently proposed constitutive models for hydraulic behaviour of unsaturated soils, among other by Sun et al. (2008), Gallipoli et al. (2003), Nuth and Laloui (2008) and Wheeler et al. (2003). In these models, the dependency of WRC on e is controlled by an appropriately chosen empirical relationship controlled by additional model parameters. In this paper, an alternative to these relationships is developed on the basis of several fundamental assumptions. No additional parameters are needed to describe the dependency of WRC on e, which simplifies the parameter determination procedure. Hydraulic hysteresis is not considered in the present derivations; the proposed model may be considered as suitable for description of the main wetting and drying branches of WRC. Nonetheless, it may easily be incorporated into a more advanced model predicting the hysteretic hydraulic behaviour. More details on the proposed approach are given by Man (2010).
ADOPTED DESCRIPTIONS OF THE STRESS STATE AND WATER RETENTION CURVE
2.1 Stress state description Central to the proposed approach is the adopted description of a stress state within unsaturated material. It now becomes generally accepted that two stress measures are needed for proper description of the stress state within unsaturated soil. A tensorial stress measure describing the averaged action of external forces and fluid pressures on the soil skeleton and a scalar stress measure quantifying the stiffening effect of water menisci on the skeleton. The tensorial stress measure may be in general written as uw1 (1 )ua1 net s1 (1)
where is a total stress, ua is the pore air pressure and uw is the pore water pressure, net is the net stress defined as net ua1 and s is matric suction s (ua uw). is the Bishop factor. The incremental form of (1) is written as u 1 w1 (1 )u a net s where (2)
d ( s) ds
(3)
Different formulations for the factor are adopted by different researchers. One of the
863
possible approaches selects the tensorial variable such that it forms a frame within which the unsaturated soil behaviour can be uniquely described. The additional scalar stress variable is needed to control the size of the state boundary surface. Such a tensorial stress measure can then be seen as an equivalent to the effective stress in saturated materials. Probably the most popular formulation for the factor is Sr. It was, however, derived from different theoretical considerations rather then from the actual observation of the unsaturated soil behaviour. Suitability of different stress measures was studied by Khalili and Khabbaz (1998) and Khalili et al. (2004). Based on experimental data on volume change and strength of overconsolidated soils they have shown that the most suitable expression for that would satisfy the aforementioned properties of effective stress in unsaturated soils does not appear to be the one with Sr. They proposed a formulation which does not involve the degree of saturation: 1 s e s for s se for s se
valid for Sr Sres, where Sres is the residual degree of saturation. Se is the air-entry or air-expulsion suction as in Eq. (4). Similarly to the adopted expression for the effective stress factor , Eq. (6) neglects the effects of hydraulic hysteresis. It may therefore be considered as appropriate for representation of the main drying and wetting branches of the WRC. In (6), both factors Se and p depend on the soil type and on void ratio. 3 BASIC RELATIONS
Considering the existence of generalised elastic and plastic potentials defined in terms of effective stress for unsaturated soils, Loret and Khalili (2000) and Khalili et al. (2008) derived the following constitutive formulations for the pore water volume (Vw) and pore air volume (Va) changes: v a11 u V w w a12 u a V v a21 u V a (1 ) w a22 u a V (7) (8)
(4)
where Se is the air entry value of suction (or air expulsion for wetting processes) and is an empirical coefficient. It was further shown that the bestfit value of the exponent 0.55 is suitable to represent the behaviour of different soil types. can thus be considered as a material independent constant. Se depends on the soil type and on the void ratio, though in the original formulation it is for simplicity considered as constant. Eq. (4) leads using (3) to the following expression for the incremental effective stress factor , which will be used as a basis for developments presented in this paper: 1 (1 ) 2.2
in which V is total volume of a soil element, V is the rate of soil skeleton volumetric v V strain, aij are material parameters and is the effective stress rate factor from Eq. (5). The definition of Sr implies rVv w SrV v S V and therefore w V r Sr v nS V (10) (9)
for s se for s se
(5)
n Vv /V is porosity. The degree of saturation Sr depends on suction s, void ratio and on the suction path. We may therefore write r S Sr S r v s s v (11)
Water retention curve
Unlike the exponent of the effective stress factor in (4), the WRC for a given void ratio depends significantly on the soil type and the granulometry. Many different mathematical relationships for the WRC with variable complexity are available throughout the literature. In this paper, a simple formulation formally similar to that by Brooks and Corey (1964) will be used: 1 Sr s p e s for s se for s se
which can be substituted into (10): w S S V Sr n r v n r s v s V Comparing (12) with (7) we have (12)
Sr n
Sr S Sr e r v e
(13)
(6)
where e n/(1 n) is void ratio. Substituting (13) into (11) yields the following general expression for the rate of the degree of saturation Sr:
864
r S
Sr Sr s e s e
(14)
The first term in (14) quantifies the WRC at constant void ratio, the second term evaluates the dependency of Sr on void ratio at constant suction. Eq. (14) was originally derived by Loret and Khalili (2000). 4 ADDITIONAL MOTIVATION FOR THE PRESENT WORK
constant air-expulsion value se0, with the factor defined as in Eq. (5), imposes incorrectly 1 for all states with s se0, therefore also for states with sr 1 (Fig. 1a). This problem may be overcome in two ways. First, e-dependency of se may be considered. A suitable relationship is presented in Sec. 5. Second, the effective stress factor from Eq. (4) may be, considering the adopted formulation for WRC (6), equivalently expressed in terms of sr
Sr
( / p )
(15)
Eq. (14) can be used to calculate the rate of Sr from rates of void ratio and suction. In the following it is shown that using the rate form for calculation of changes of Sr with s and e (Eq. (14)) in combination with the effective stress formulation of Eq. (4) with constant Se, and WRC formulation of Eq (6) with constant p, leads to incorrect results. Consider two specimens of the same void ratio e0 and air-expulsion suction se0 at two different suction levels, shown in Figure 1a. Both the two specimens lie on a single wetting branch of a water retention curve corresponding to e0 (states A and B). The specimens are first subject to a void ratio increase at constant suction to a void ratio e1 (paths A-A and B-B), followed by a suction decrease at constant void ratio e1 up to a full saturation (paths A-A and B-B). In the following it is assumed that the main drying and main wetting branches of WRC are only dependent on void ratio and independent on the stress history. Based on this assumption, the two paths A-A and B-B should lie on the same wetting branch of WRC corresponding to the void ratio e1, as shown in Fig. 1b. Instead, direct application of Eq. (14) for calculation of changes of Sr with constant p predicts that the two specimens follow different paths in Sr vs. s plane during suction decrease (Fig. 1a). The anticipated response may be obtained if void ratio dependent value of p is considered. Note that Fig. 1b presumes p1 p0 for e1 e0. This dependency is implied by derivations presented subsequently in Sec. 16. In addition to the problems described above, application of the effective stress principle (4) with
ln Sr
Sr=1
which eliminates se from the effective stress equation. Eq. (15) may be seen as a link between the effective stress concept by Khalili and Khabbaz (1998) from Eq. (4) and the theoretically derived expression Sr. As will be shown further, studies of the behaviour of different soils showed that ( /p) 1. The exponent /p may thus be seen as an empirically revealed manifestation of simplifying assumptions adopted in the theoretical approaches leading to the derivation of Sr. Eq. (15) is indeed valid for the main wetting and drying branches of WRC, provided that the double-logarithmic expression for WRC (6) is considered as appropriate. This does not directly imply its applicability for sr states on the hydraulic scanning curve. 5 QUANTIFICATION OF THE DEPENDENCY OF WRC ON VOID RATIO
In this section, solution for the dependency of WRC on e is presented. The formulas are applicable for unsaturated states, where s se and therefore (1 ). The derivations are detailed in Man (2010). It has been shown that the rate of the suction at air entry/expulsion se can be calculated by e s with
se e e psu
(16)
B A p0 1
=1 although S <1 r
ln Sr
Sr=1 e=e0
A= B A A p1 B
e=e0
A A 1 p1 B
psu
p0 e ln su 0 su 0 ln 0 su e0
1
0 su (17)

e=e1
p0 1
e=e1
1 s e1 s e0
p0 1 B
B ln (s)
where 0su (se0/se). se0 and p0 are values of se and p corresponding to the reference void ratio e0. Dependency of p on void ratio and suction is given by
ln (s)
s e1 s e0
(a )
(b)
Figure 1. Theoretical experiment demonstrating motivation for the present work. (a) response by Eq. (14) without e-dependent se and p; (b) response by the proposed model.
p
p0 e ln 0 0 ln 0 e0
1
0

(18)
with 0 (se0/s).
865
Knowledge of se from (18) and p from (17) may be used to calculate the value of Sr for given void ratio and suction using Eq. (6). In fact, Eqs. (1618) describe a state surface in the s vs. e vs. Sr space, which is depicted in Figure 2 for Pearl clay parameters from Table 1. Constant-void-ratio cross-sections through this surface represent water retention curves, shown in Figure 3a. Figures 2 and 3a also demonstrate how increasing void ratio leads to the decrease of the se value. Figure 3b shows the influence of e and s on the slope p of the WRC. Two ways may be followed in the case the rate form of Sr is required (Eq. (14)). The derivative Sr/s may either be found numerically from (6), or it can be substituted by pSr/s if the dependency
Sr [-] 1 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0 50 100 s [kPa] 150 1.5 200 2 0.5 1 e [-]
of p on s is neglected. This dependency is not significant for void ratios not significantly different as compared to the reference void ratio e0 (see Figure 3b). Note that the model includes Sr as a special case. In this case p0 and Eq. (18) simplifies to p p0 (i.e., p is independent of suction and of void ratio). The Equation (16) may then be integrated analytically, with resulting se se0e0/e. 6 DETERMINATION OF PARAMETERS
The model requires three material parameters. Parameters se0 and p0 may be found directly by a bi-linear representation of the water retention curve in the lnSr vs. ln(s/se0) plane, as shown in Figure 4. Parameter e0 is void ratio corresponding to the approximated WRC. Obviously, this approach to model calibration assumes that the void ratio changes due to suction changes during the measurement of WRC are negligible. In the case the variation of e cannot be neglected, the calibration of p0 requires coupling of a proposed hydraulic model with some suitable mechanical model for unsaturated soils, evaluation of e for different
ln Sr 0 measured WRC
Figure 2. Predicted state surface in the s vs. e vs. Sr space.
model approximation
1 0.9 0.8 Sr [-] 0.7 0.6 0.5 0.4 0.3 0 20 40 60 80 100 120 140 160 180
e=2 e=0.8
p0 1
0
s [kPa]
ln (s/se0)
(a)
90 80 70 60 50 40
(a)
0.4 0.38 0.36
e=2
p [-]
Sr [%]
0.34 0.32
e=0.8
experiment se0=20 kPa, p0=0.15 se0=10 kPa, p0=0.15 se0=30 kPa, p0=0.15 se0=20 kPa, p0=0.10 se0=20 kPa, p0=0.20 10 pnet [kPa] 100
0.3
20
40
60
80
100 120 140 160 180
s [kPa]
(b)
Figure 4. (a) Direct determination of parameters se0, p0 and e0 by a bi-linear representation of the WRC; (b) calibration by the trial-and-error procedure using Sr vs. s vs. e data (experimental data by Sharma 1998).
(b)
Figure 3. (a) The dependency of WRC on void ratio; (b) dependency of the slope p of the WRC on s and e for reference values p0 0.38, se0 15 kPa and e0 1.75.
866
suction states and comparison of the experimental WRC with an appropriate cross-section through the Sr:s:e state surface. Alternatively, the parameters may be found by a trial-and-error procedure using Sr vs. s vs. e data coming from laboratory experiments. In principle, any experiment with monotonous path may be used for this purpose. For example compression or shear tests at constant suction, wetting/drying tests at constant net stress, constant water content experiments, etc. Note that the value of e0 may be selected arbitrarily to be in the range of reasonable void ratios for the given soil. The calibration using the trial-and-error procedure is demonstrated in Fig. 4b, which shows pnet vs. Sr diagram of constant suction isotropic loading-unloading test on bentonite/kaolin mixture by Sharma (1998) and predictions with the proposed model with e0 1.2 and variable se0 and p0. Note that only one test is shown in Figure 4b for clarity. Proper calibration using the trial-and-error procedure requires considering more tests with variable suction and/or void ratio. 7 COMPARISON WITH EXPERIMENTAL DATA
100 90 80 S r [%] 70 60 50 40 30 1 10 s [kPa]
experiment p0=0.10 p0=0.18 p0=0.25
100
1000
Figure 5. Wetting branch of WRC of HPF type quartz silt, experimental data by Jotisankasa et al. (2007), direct calibration of the proposed model with e0 0.7, se0 3 kPa and variable p0.
100 80 60 40 20 0 7-10-T 7-10-H 7-10-D 7-10-K 7-10-G 1000 10000
tot
S r [%]
experiment 10 100
The proposed dependency of WRC on void ratio has been evaluated with respect to a wide range of different soils by Man (2010). All predictions were obtained using Eqs. (16-18). The predicted degree of saturation Sr was calculated directly from the experimentally measured e and s without a need to couple the present hydraulic model with a mechanical constitutive model for partly saturated soils. Parameters obtained are summarised in Table 1. It is interesting to note that for all studied soils the value of p0 is significantly lower than 0.55, demonstrating empirically that the exponent /p from Eq. (15) takes values significantly higher than unity. In the following, two examples of predictive capabilities of the proposed approach will be given. 7.1 HPF quartz silt
(a)
100 80 60 40 20 0
ax [kPa]
S r [%]
model 10 100
7-10-T 7-10-H 7-10-D 7-10-K 7-10-G 1000 10000 tot ax [kPa]
(b)
Figure 6. Results of suction-monitored oedometric tests at constant water content by Jotisankasa et al. (2007) (a), compared with model predictions of Sr using parameters calibrated directly from WRC (b).
Jotisankasa et al. (2007) performed a set of constant water content oedometric tests with monitored
Table 1. Parameters of the proposed model calibrated on the basis of different experimental data (Man 2009). soil Pearl clay HPF quartz silt Speswhite kaolin bentonite/kaolin mix. se0 [kPa] 15 3 65 20 p0 [] 0.38 0.18 0.3 0.15 e0 [] 1.75 0.7 1.4 1.2
suction on a mixture of 70% silt of HPF type (consisting mainly of angular quartz grains), 10% kaolin and 20% London clay. The authors made available the WRC measured by standard filter paper technique at zero vertical stress, which allows us to calibrate the proposed model using the direct approach, demonstrated in Fig. 4a. Figure 5 shows the wetting branch of the WRC starting from compacted state at s 1000 kPa and void ratio e0 0.7. Figure 5 also shows the WRC predicted by the proposed model with e0 0.7, se0 3 kPa and variable p0. The model curve for p0 0.18 reproduces well the experimental data, indicating suitability of Eq. (6) for the present purpose.
867
100 90 80 70 Sr [%] 60 50 40 30 20 0 experiment 200 400
w=0.215 w=0.236 w=0.254 w=0.275 w=0.299 w=0.311
of WRC on void ratio has been formulated. The model does not require any model parameters apart from parameters specifying WRC for the reference void ratio. Good match between observed and predicted behaviour indirectly supports the applicability of the effective stress concept for unsaturated soils.
1200
600 800 s [kPa]
1000
ACKNOWLEDGMENT The author wish to thank to Prof. Nasser Khalili for valuable discussions on the subject and to Dr. Alessandro Tarantino for providing experimental data on Speswhite kaolin. Financial support by the research grants GACR 205/08/0732 and MSM 0021620855 is greatly appreciated. REFERENCES
(a)
100 90 80 70 Sr [%] 60 50 40 30 20 0 model 200 400 600 s [kPa] 800 1000 1200 w=0.215 w=0.236 w=0.254 w=0.275 w=0.299 w=0.311
(b)
Figure 7. (a) Results of compaction tests on Speswhite kaolin by Tarantino and De Col (2008) plotted in Sr vs. s graphs for constant water contents. (b) predicted results.
The oedometric tests by Jotisankasa et al. (2007) were performed at constant water content conditions, therefore both suction and void ratio varied during compression. Figure 6a shows the observed dependency of Sr on vertical stress for five different tests. The predictions are shown in Figure 6b. The model is in a good match with the experiments, although tests with variable void ratio were not involved in model calibration. 7.2 Speswhite kaolin
Tarantino and De Col (2008) studied the behaviour of Speswhite kaolin under static compaction at seven different water contents with continuous measurement of suction. The experimental results are in Fig. 7a shown in terms of Sr vs. s plots for different water contents. Figure 7b shows predictions by the proposed formulation with parameters calibrated by means of the trial-and-error procedure (described in Sec. 3). The proposed state surface represents well the measured behaviour for virgin loading. As expected, the model is due to the absence of hydraulic hysteresis less successful in predicting the behaviour upon unloading-reloading cycles. 8 CONCLUSIONS
Based on several underlying assumption including existence of generalised elastic and plastic potentials defined in terms of effective stresses for unsaturated soils, a new model for the dependency
Brooks, R. & Corey, A. (1964). Hydraulic properties of porous media. Hydrology paper No. 3, Colorado state University. Gallipoli, D. & Wheeler, S.J. & Karstunen, M. (2003). Modelling the variation of degree of saturation in a deformable unsaturated soil. Gotechnique (1), 105112. Jotisankasa, A., Ridley, A. & Coop, M. (2007). Collapse behaviour of compacted silty clay in suction-monitored oedometer apparatus. Journal of Geotechnical and Geoenvironmental Engineering ASCE (7), 867877. Khalili, N., Geiser, F. & Blight, G.E. (2004). Effective stress in unsaturated soils: review with new evidence. International Journal of Geomechanics (2), 115126. Khalili, N., Habte, M.A. & Zargarbashi, S. (2008). A fully coupled flow-deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Computers and Geotechnics (6), 872889. Khalili, N. & Khabbaz, M.H. (1998). A unique relationship for for the determination of the shear strength of unsaturated soils. Gotechnique (2), 17. Loret, B. & Khalili, N. (2000). A three-phase model for unsaturated soils. International Journal for Numerical and Analytical Methods in Geomechanics, 893927. Man, D. (2010). Predicting the dependency of a degree of saturation on void ratio and suction using effective stress principle for unsaturated soils. International Journal for Numerical and Analytical Methods in Geomechanics, 7390. Nuth, M. & Laloui, L. (2008). Effective stress concept in unsaturated soils: Clarification and validation of a unified framework. International Journal for Numerical and Analytical Methods in Geomechanics, 771801. Sharma, R.S. (1998). Mechanical behaviour of unsaturated highly expansive clays. Ph. D. thesis, University of Oxford. Sun, D.A., Sheng, D., Xiang, L. & Sloan, S.W. (2008). Elastoplastic prediction of hydro-mechanical behaviour of unsaturated soils under undrianed conditions. Computers and Geotechnics, 845852. Tarantino, A. & De Col, E. (2008). Compaction behaviour of clay. Gotechnique (3), 199213. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. (2003). Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Gotechnique, 4154.
868
Rotational multiple yield surface framework and the prediction of stress-strain response for saturated and unsaturated soils
M.J. Md Noor & I.B. Mohamed Jais
Faculty of Civil Engineering, Universiti Teknologi MARA, Shah Alam, Malaysia
ABSTRACT: Anisotropic soil settlement model known as Rotational Multiple Yield Surface Framework characterises soil volume change behaviour from the standpoint of the soil stress-strain relationship. From the stress-strain relationship the soil mobilised shear strength behaviour in saturated and unsaturated conditions is derived in the form of Mohr-Coulomb curved-surface envelope in the space of :( ua) : (ua uw). Then within this stress space the anisotropic volume change behaviour is characterised from the interaction between the applied stress represented in the form of Mohr circle and soil state of shear strength, , in the form of the curved-surface mobilised shear strength envelope. Anisotropic compression is triggered whenever the applied stress state exceeds the strength and the process is stopped when the state of stress equilibrium is reached. This is the condition when Mohr circle just touches the strength surface envelope. The framework utilises the concept of multiple yield surface which is represented by rotation of the mobilised soil shear strength envelope about suction axis towards shear strength envelope at failure as the soil structure is compressed. The rotation of the mobilised shear strength envelope cum the yield surface is to mark the increase in the soil strength which is the direct incorporation of the soil hardening law into the framework. The advantage of this framework is that it directly links the role of shear strength and applied stress in the soil anisotropic volume change response. Consolidated drained triaxial tests have been conducted on saturated and unsaturated specimens of tropical residual soil. The framework prediction of the soil stress-strain response in saturated and unsaturated soil conditions is presented. The close agreement between the experimental and the predicted curves demonstrate the applicability of the framework. 1 INTRODUCTION Settlement is always associated with the increase in effective stress either through the increase in the total stress or the dissipation of the excess pore water pressure. However, the concept cannot explain the occurrence of settlement under constant load like the inundation settlement which is popularly known as wetting collapse. To make the situation more complex, wetting collapse is taking place under effective stress decrease (Jennings and Burland, 1962). Before inundation or infiltration the soil is in partially saturated condition where the pore water pressure is negative. When inundated the negative pore water pressure at any arbitrary point within the soil mass changed from being negative to positive. Thence there is an increase in the pore water pressure and hence there is a subsequent decrease in the effective stress according to the effective stress equation as follows; ' uw (6)
Conventionally, foundation settlement is quantified using equations which adopt the concept of effective stress of Terzaghi (1936 and 1943). The concept of effective stress has been the most fundamental concept in soil mechanics. These are like the following equations to predict settlement in clay and sand. sclay qB 1 2 I p, E
Steinbrenner (1934)
(1)
sclay
p p Cc H , log 0 1 e0 p0
Terzaghi (1943)
(2)
sclay
o 1qB 1 2 , Janbu et al., (1956) Eu

H ln C
o
(3)
ssand
, De Beer and Martens (1951) o (4) Iz z , Schmertmann et al., (1978) E (5)
ssand C1C2qn
Therefore wetting collapse is taking place under effective stress decrease. This type of settlement cannot be quantified by any of the settlement equations like Equations 1, 2, 3, 4 and 5 since in this case there is no increase in the effective stress. Thus wetting collapse or inundation settlement is considered as
869
Figure 1. Curved-surface Mohr-Coulomb envelope of Md.Noor and Anderson (2006) for limestone gravel.
complex soil volume change behaviour. Another soil complex volume change behaviour is the massive settlement near saturation as reported by Tadepalli et al. (1992) and Blanchfield and Anderson (2000). These complex characteristics volume change behaviour has caused a great difficulty to researchers to formulate a close form volume change framework. In fact it is not just the increase in effective stress that is triggering the settlement. It can be triggered by the reduction in the shear strength. Inundation is in fact reducing the soil shear strength when suction decreases and this aspect of strength reduction is demonstrated by the shear strength behaviour of limestone gravel reported by Md.Noor and Anderson (2006) as shown by the curved surface Mohr-Coulomb envelope in Figure 1. As suction approaches zero there is a steep drop in the shear strength. This shear strength attribute has a significant role on the volume change behaviour with regards to inundation. The reduction in shear strength thereby mobilised the sliding between the soil particles under the existing constant total stress which brings further soil compression. An anisotropic framework known as Rotational Multiple Yield Surface Framework has been introduced by Md.Noor and Anderson (2007) to characterize the soil volume change behavior by this concept. The framework is derived from the soil stress-strain response. This is an extended concept of effective stress which is the concept of effective stress and shear strength interaction. The framework applies the true soil elastic-plastic response and it has the capability to predict settlement. 2 CONCEPT OF EFFECTIVE STRESS AND SHEAR STRENGTH INTERACTION
ever when the magnitude of one of the variables has changed settlement would be triggered. This is when the driving variable overcomes the resisting. Consider an arbitrary point, A, underneath a foundation as shown in Figure 2. Point A is subjected to a state of stress derived from the applied stress coming from the column load. This will produce the state of normal and shear stresses at point A. The magnitudes of the normal and shear stresses at point, A, vary with the plane inclination angle, . They are best represented by a Mohr circle in the extended Mohr-Coulomb space, i.e., : ( ua) : (ua uw) as shown in Figure 3. The condition here is assumed to be fully saturated when the Mohr circle is sitting on the net stress axis. If the effect of pore water pressure has been subtracted from the normal stress then the Mohr circle will be representing effective stress state in any direction. Effective stress is the settlement driving variable. On the other hand, the state of mobilised shear strength at point A can be represented by the mobilised curved-surface Mohr-Coulomb shear strength envelope as shown in Figure 3. The magnitude of the mobilised shear strength depends on the the volumetric compressive state. The minimum mobi mob which increases as the lised friction angle, min soil undergo compression is directly related to the
Figure 2. Interaction between applied shear stress and soil shear strength along an incline failure plane at an arbitrary point A underneath a foundation in opposing particles sliding towards soil denser state.
(u a -u w )
Mobilised saturation
curved-surface
shear strength envelope at

' min
i
mob
Effective stress Mohr circle
In a stable condition the effective stress which is the settlement driving variable and the shear strength which is the settlement resisting variable are in equilibrium. In this condition there is no foundation downward movement i.e., no settlement. How-
(-u a )
Figure 3. The Mohr circle representing the state of applied stress and the curved-surface envelope representing the state of shear strength.
870
soil volumetric state. Note that the mobilised shear strength is lesser than the shear strength at failure. The equilibrium of the stress at point A in Figure 2 is not necessarily depending on the maximum normal stress or the maximum shear stress, but it relies on the optimum combination of the two according to Mohr (1900). The limiting stress on the Mohr circle is actually the state of stress represented by point A in Figure 3. The applied shear stress is resisted by the shear strength, i along the same plane as illustrated in Figure 2. If the applied shear stress exceeds the resisting shear strength which is when the Mohr circle extends above the strength envelope then there will be sliding along the plane and this will result in particle rearrangement towards a denser state which produces further settlement. Settlement or the compression of soil structure will cause an increase in the soil shear strength. When the packing of the soil particles are closer there will be a higher resistance against particle sliding. This is how the mobilised shear strength is increased with soil compression. As settlement is increased with soil compression or also referred as soil hardening, the developed strength will compensate the applied stress and at one point another stress equilibrium condition will be reached. At this point the settlement will be stopped. By the concept of effective stress and shear strength interaction settlement can be triggered through the increase in the effective stress or through the reduction in the shear strength. The settlement causes by stress increase is known as loading collapse while wetting collapse is the result of shear strength reduction. This concept is implemented in an anisotropic soil settlement framework known as Rotational Multiple Yield Surface Framework. 3 ROTATIONAL MULTIPLE YIELD SURFACE FRAMEWORK
at failure to represent a higher state of strength. When the envelope and the Mohr circle resume the point contact, the state of stress equilibrium is reinstated and settlement will be stopped. There are two ways that settlement can be triggered. First, is by increasing the vertical stress which enlarged the diameter of the Mohr circle until the Mohr circle extends above the existing shear strength envelope (refer Figure 4a). This type of settlement is known as loading collapse. Secondly, is by wetting which reduces the suction. This drives the Mohr circle towards net stress axis (refer Figure 4b). In other words, the Mohr circle moves towards the lower part of the envelope since shear strength decreases non-linearly relative to suction. During the process if the Mohr circle extends above the strength surface envelope, settlement will again be triggered and subsequently force the surface envelope to rotate to represent soil hardening. The settlement produces by this process is known as wetting collapse. This settlement is triggered by the reduction in shear strength. During the wetting process the state of effective stress remains constant since the reduction in suction does not affect the diameter of the Mohr circle. This is because suction is an isotropic tensor. Its changes affect the vertical and the horizontal stresses equally and so the diameter of the Mohr circle remains constant during the process of wetting. These are how the framework justifies qualitatively the occurrence of loading and wetting collapse. More importantly
Consider a Mohr circle and curved-surface envelope soil shear strength model plotted in the extended Mohr-Coulomb space i.e., : ( ua) : (ua uw) as shown in Figure 4. In the state of stress equilibrium the Mohr circle and the mobilised shear strength envelope are at a point contact. If the Mohr circle lies lower than the mobilised strength envelope it means that the driving effective stress is lower than the strength. In this condition the soil skeletal structure is still stable i.e., there will be no settlement. However when the Mohr circle extends above the strength envelope, this is a state of stress imbalance at any arbitrary point underneath the footing. The effective stress is greater than the strength. In this condition settlement will be triggered. However strength is simultaneously increases with soil compression. This is indicated by the rotation of the envelope towards the shear strength envelope
Figure 4 Change in stress state and settlement (a) Loading collapse (b) Wetting collapse.
871
it can justify the occurrence of settlement under effective stress decrease i.e., wetting collapse. Notice that the anisotropic settlement is indicated by the rotation of the mobilised shear strength envelope towards higher overall internal friction angle, . The position of the mobilised shear strength envelope is representing a certain degree of compression or soil volumetric state. This needs to be true for the framework to be valid. There must exist a unique relationship between the parameter that defines the position of the mobilised shear strength envelope (i.e., minmob ) and the degree of compression (i.e., axial strain, a) for any magnitude of effective stress that the tests were conducted. This has been substantiated in Section 4. Full 100% compression is achieved when the soil reaches fully interlocking position where there is no more room for rearrangement of particles. At this state shear plane will be developed and this is the state that shear strength at failure is achieved. At the same time the mobilised shear strength envelope is acting as the yield surface. This is because whenever the stress state represented by a Mohr circle exceeded the yield limit by extending above the surface envelope, the soil will yield by compressing its structure. Since the Rotational Yield Surface Framework applies the curved-surface envelope Mohr-Coulomb soil shear strength model then the framework is using the real soil shear strength behavior in characterizing the volume change. Consider the condition that the mobilised strength envelope is always at one point contact with the active Mohr circle, when the Mohr circle is driven towards the net stress axis by wetting as shown in Figure 4b, there is a substantial rotation of the mobilised shear strength envelope to indicate massive soil compression. The massive rotation is due to the steep drop in shear strength as suction approaches zero. This is how the framework explains the occurrence of massive settlement near saturation as being reported by Tadepalli et al. (1992) and Blanchfield and Anderson (2000). 4 LABORATORY ANISOTROPIC LOADING COLLAPSE TESTS AT SATURATION AND PARTIAL SATURATION
38 mm diameter and 76 mm height. The soil is classified as very silty SAND where there are 70% particles of sand size and 29% is silt size and only 1% is clay size. Two set of test series were conducted. The first test series is on saturated specimens and the second test series is on partially saturated specimens of 5% moisture content. The effective stresses and net stresses applied were 50, 100 and 200 kPa. Non axis translation technique using conventional triaxial cell was applied in the consolidated drained triaxial tests on partially saturated specimens. The pore air pressure was set to atmospheric zero. The magnitude of suction for the triaxial tests on partially saturated specimens is unknown. In fact it can be determined from the soil-moisture characteristic curve by correlating the suction that corresponds to the 5% moisture content. Thence the cell pressure applied in the test on partially saturated specimens of 50, 100 and 200 kPa corresponds to the applied net stress for the test. The stress-strain curves for the tests on saturated and partially saturated specimens are shown in Figure 5(a) and (b) respectively. Essentially the deviator stresses at failure for the partially saturated specimens are higher than those of saturated specimens of the same effective or net stress. The extra strength is produced due to the presence of suction under zero pore air pressure and naturally exist negative pore water pressure. This testing technique without application of axis translation is called unconfined triaxial test by Bishop and Blight (1963). They claimed that the pore-water in the soil is able to withstand high tensions (100 kPa) without rupturing the meniscus. The shearing stage
450.0
Effective stress 200 kPa (Prediction)

400.0
350.0
Deviator Stress ( 1 ' - 3 ') kPa
Effective stress 200 kPa (Saturated triaxial test)

300.0
250.0
Effective stress 100 kPa (Prediction)
(a)
200.0
Effective stress 100 kPa (Saturated triaxial test) Effective stress 50 kPa (Saturated triaxial test) Effective stress 50 kPa (Prediction)
150.0
100.0
50.0
0.0 0.00
5.00
10.00
15.00
20.00
0 25.00
Axial Strain a (%)

600.0
Effective Net stress stress 200 200 kPa kPa (Prediction) (Prediction)
500.0
The anisotropic loading collapse is in fact the specimen compression during shearing in a consolidated drained triaxial test. Since the specimen is allowed to expand laterally during the application of the deviator stress while the confining cell pressure is maintained constant, the specimen is considered to be subjected to anisotropic stress and volume change conditions. Consolidated drained triaxial tests were conducted on remoulded saturated specimens of granitic residual soil grade VI from Rawang, Malaysia. The specimens dimensions are
Deviator Stress ( 1 ' - 3 ') kPa
400.0
Net stress 200 kPa (Unsaturated triaxial test) Net stress 100 kPa (Prediction)
300.0
(b)
Net stress 100 kPa (Unsaturated triaxial test)

200.0
Net stress 50 kPa (Unsaturated triaxial test) Net stress 50 kPa (Prediction)
100.0
0.0 0.00
5.00
10.00
15.00
20.00
25.00 0
Axial Strain a (%)
Figure 5. Deviator stress versus axial strain curves (a) saturated (b) partially saturated specimens.
872
500
Shear strength (kPa)
Failure envelope at partial saturation

400
300
Failure envelope at saturation
200
100
93 75 100
0 100 200 300 400 500 600 700 800 900 1000
Effective/Net stress (kPa)
Figure 6. Curvi-linear shear strength envelope at failure for the saturated and partially saturated conditions.
of these tests is actually under anisotropic loading collapse condition. The magnitude of axial strain is taken to represent the soil deformation state. The soil structural compression is driven by the increase in the deviator stress which in turns increases the applied effective or net stress to the specimens. This is indicated by the enlargement of the effective or net stress Mohr circle during the application of the deviator stress. Thence the stress-strain curves obtained are under anisotropic stress conditions. The shear strength envelopes at failure for the saturated and partially saturated condition are deduced from the maximum deviator stresses. The effective stress Mohr circles at failure and the deduced shear strength envelope at failure are as shown in Figure 6. Essentially the envelope shows a curvi-linear shape where the transition effective stress is 100 kPa, transition shear strength is 75 kPa and the minimum internal friction angle at failure, min f is 27.5. The elevation of the shear strength envelope at partial saturation above the envelope at saturation by 18 kPa is actually the apparent shear strength for the 5% moisture content. 5 MOBILISED SHEAR STRENGTH ENVELOPES AND UNIQUE RELATIONSHIP
Figure 7. Mobilised shear strength envelopes and shear strength envelope at failure for (a) saturated (b) partially saturated conditions.
30
25
Minimum Mobilised Friction Angle
20
15
10
Unique relationship ' Axial min mob strains, % 0 0 0.6 5 1.13 8.9 1.875 12.9 2.864 16.9 4.04 20.9 6.59 24.9 12 27 19 27.5
0 5 10 15
Deduced unique relationship
Fully Saturated Partially Saturated Unique relationship

20 25
Axial Strain (%)
Figure 8. The unique relationship minmob a for the compression of saturated and partially saturated specimens.
The mobilised shear strength envelopes for various axial strains determined from the stress-strain curves for consolidated drained triaxial tests on saturated and partially saturated specimens are shown in Figure 7(a) and (b) respectively. The Mohr circles were drawn from deviator stress that correspond to a specific value of axial strain for tests at effective stresses and net stresses of 50, 100 and 200 kPa. This is followed by the drawing of the mobilised strength envelope. These mobilised strength envelopes are marked with the corresponding axial strain of which they represent as shown in Figure 7(a) and (b). Then the relationship between axial strain, a% and minimum mobilised friction angle, minmob as marked in Figure 7(a) and (b) are plotted as shown in Figure 8. Apparently the two graphs overlap to indicate that the curved mobilised shear strength
envelope for the saturated and the unsaturated specimens rotate simultaneously according to the axial strain during the application of the deviator stress in the shearing stage irrespective of the magnitudes of the effective stress or the net stress applied in the tests. The deduced unique relationship is as shown in Figure 8. This unique relationship is applied for the prediction of the stress-strain response at effective and net stresses of 50, 100 and 200 kPa. 6 PREDICTION SOIL STRESS-STRAIN RESPONSE AT SATURATION AND PARTIAL SATURATION
The advantage of having the unique relationship minmob a is that it proves that there is a simultaneous rotation of the strength envelope about suction axis during the compression of the specimens in the shearing stage. This is irrespective of the value of effective or net stress applied. This also substantiates the effect of soil hardening in the course of the compression. The existence of this unique relationship is the keystone of the Rotational Multiple Yield Surface Framework. Without the existence
873
of this unique relationship the framework would not be applicable. The most important advantage of having this unique relationship is that it can be used to predict the soil stress-strain behaviour at any effective or net stress. Since there is a unique relationship as shown in Figure 8, then the mobilised shear strength envelopes in Figure 7(a) and (b) can be applied to predict the soil stress-strain response under saturated and partially saturated (i.e., 5% m.c.) conditions respectively. The Mohr circles in Figure 7(a) and (b) are drawn tangential to the respective mobilised strength envelope. From the corresponding values of the % axial strains and the major principal stresses, 1 the magnitude of the deviator stress, can be determined when the cell pressure and the pore water or the pore air pressure are known. The predicted deviator stress and the corresponding % axial strain for the compression of the saturated and partially saturated specimens obtained from the predicted Mohr circles in Figure 7(a) and (b) are shown by the dotted data points in Figure 5(a) and (b) respectively. The predicted data points show a good agreement with the experimental data points. The same procedure can be applied to predict the soil stress-strain response at any other effective stress or net stress for saturated or partially saturated soil conditions. This is the main advantage of having the curvi-linear mobilised shear strength envelope. Hence the foundation settlement at any depth (i.e., any effective or net stress) can be calculated using the Rotational Multiple Yield Surface Framework in either saturated or partially saturated soil conditions. 7 CONCLUSIONS
explain the occurrence of various soil complex volume change behaviours like inundation settlement and massive settlement near saturation. Most important of all, the Rotational Multiple Yield Surface Framework applied the anisotropic stress and volume change conditions which replicates the true field condition and has elegantly incorporated the role of shear strength in governing soil volume change behaviour. ACKNOWLEDGEMENTS This research has been carried out with the financial support from the Fundamental Research Grant provided by the Ministry of Higher Education of Malaysia through Research Management Institute, Universiti Teckologi MARA, Shah Alam, Malaysia. REFERENCES
Bishop, A.W. & Blight, G.E. (1963). Some aspects of effective stress in saturated and unsaturated soils. Geotechnique, 13(3), 177197. Blanchfield, R. & Anderson, W.F. (2000). Wetting collapse in opencast coalmine backfill. Proceedings of the ICE Geotechnical Engineering, London: 139149. De Beer, E.E. & Martens, A. (1951). Method of computation of an upper limit for influence of heterogeneity of sand layers on the settlement of bridges. Proceedings 4th International Conference on Soil Mechanics and Foundation Engineering, London, pp. 275278. Janbu, N., Bjerrum, L. & Kjaernsli, B. (1956). Veiledring ved losning av fundermenteringsoppgaver. Norwegian Geotechnical Institute Publication No.16, Oslo. Md.Noor, M.J. & Anderson, W.F. (2006), A comprehensive shear strength model for saturated and unsaturated soils. Proceedings of the 4th International Conference on Unsaturated Soils, ASCE Geotechnical Publication No. 147, Carefree, Arizona, USA, 2, 19922003. Md.Noor, M.J. & Anderson, W.F. (2007) A Qualitative Framework for Loading and Wetting Collapses in Saturated and Unsaturated Soils. Proc. 16th South East Asian Geotechnical Conference, Kuala Lumpur, Malaysia. Mohr, O. (1900) Welche Umstande Bedingen die Elastizitatsgrenze und den Bruch eines Materiales? Zeitschrift des Vereines Deutscher Ingenieure, Vol. 44, 15241530, 15721577. Schmertmann, J.H., Hartman, J.P. & Brown, P.R. (1978). Improved strain influence factor diagrams. Proceedings, American Society of Civil Engineers, Vol. 104(GT8), pp. 113 1135. Steinbrenner, W. (1934). Tafeln zur Setzungsberechnung. Die Strasse, Vol. 1, pp. 121124. Tadepalli, R., Rahardjo, H. & Fredlund, D.G. (1992). Measurement of matric suction and volume change during inundation of collapsible soil. ASTM Geotechnical Testing Journal, 15(2), 115122. Terzaghi, K. (1936). The shear resistance of saturated soils. Proceedings for the 1st. International Conference on Soil Mechanics and Foundation Engineering (Cambridge, MA), 1, 5456. Terzaghi, K. (1943). Theoretical soil mechanics. New York, Wiley Publications.
Conclusions that can be drawn from these are: There is a unique relationship between the minimum mobilised friction angle, minmob and the % axial strain for the anisotropic compression of saturated and partially saturated specimens. This is representing that the mobilised curved-surface shear strength envelopes at saturation and partial saturation rotate simultaneously as one surface about suction axis to represent the increase in the axial strain and strength irrespective of the magnitude of effective or net stress applied to the test. The Rotational Multiple Yield Surface Framework applies the real soil shear strength behaviour which is the curved-surface envelope. With that it is able to make a good prediction of soil stressstrain response in both saturated and partial saturated conditions at any effective or net stress. By considering the mobilised shear strength envelope as the yield surface and its rotation towards shear strength envelope at failure as representing anisotropic compression, the Rotational Multiple Yield Surface Framework can qualitatively
874
Laboratory modelling: Settlement due to groundwater fluctuation in partially saturated soil

M.J. Md Noor, I.B. Mohamed Jais & J.D. Nyuin
ABSTRACT: Inundation settlement or wetting collapse is triggered by the rise of groundwater table. The complexity of this soil volume change behaviour is that it is taking place under effective stress reduction. In partially saturated soils, the void spaces are filled with both air and water, therefore suction force exists between the soil particles resulting to the increase in shear strength. When the soil is inundated the effect of suction is diminished and the soil shear strength reduced to the shear strength at saturation. The effect of this reduction in strength is anticipated to have the prime influence on this settlement behaviour. The role of shear strength in governing soil settlement is incorporated in Rotational Multiple Yield Surface Framework. The framework is able to model settlements due to loading and wetting. A laboratory scale tests of a model footing were conducted to study the effect of groundwater fluctuation on the soil settlement in tropical residual soil. Three types of test have been conducted: (1) the soil is loaded to failure (2) the soil is loaded and inundated (3) the soil is loaded, inundated and subsequent lowering of water table. The framework is used to predict the soil settlement behaviours. The results show good agreement between the predicted and the modelled settlement. These result shows that the Rotational Multiple Yield Surface Framework is able to model soil settlements in both conditions; loading and wetting for saturated and unsaturated soil conditions. 1 INTRODUCTION The existing soil volume change concepts which are based purely on effective stress concept of Terzaghi (1943) are not comprehensive enough to explain the various soil complex volume change behaviours. These are like inundation settlement (Blanchfield and Anderson, 2000) which is taking place under effective stress decrease or massive wetting collapse settlement near saturation (Alonso et al., 1990). A comprehensive soil volume change framework has been developed from anisotropic stress condition known as Rotational Multiple Yield Surface Framework by Md.Noor and Anderson (2007) which characterizes soil volume change behaviour from the interaction between effective stress and shear strength. The framework is derived from the soil anisotropic stress-strain response from triaxial tests on saturated and unsaturated specimens. The framework considers the soils actual yield surface which is a curved-surface extended Mohr Coulomb shear strength envelope of Md.Noor and Anderson (2007). 2 ANISOTROPIC SOIL SETTLEMENT MODEL
The behaviours of partially saturated soils in general appear to be relatively complex in comparison to those of fully saturated soils. Several attempts have been made in the past in understanding the partially saturated soil behavior. Generally, the design procedures of foundation are still applying a worst-case scenario where the effect of settlement due to wetting is incorporated by considering the ground water table to be at the ground surface or the soil to be fully saturated. However, in the field, settlement still takes place when the ground water table rises. This study was aimed to investigate the settlement behaviour due to fluctuation of water table in partially saturated soil. 1.1 History of soil settlement models
Using the effective stress concept, settlement is always associated with the increase in effective stress. This can occur either through the increase in total stress or dissipation of the excess pore water pressure. Inundation settlement or wetting collapse is settlement due to the rise of ground water table. This behavior cannot be explained based on effective stress concept of Terzaghi (1936) or using the effective stress equation of Bishop (1959) because settlement is taking place under effective stress decrease (Jennings and Burland, 1962) which makes the behaviour complex.
A qualitative soil anisotropic volume change framework for saturated and unsaturated soils known as Rotational Multiple Yield Surface Framework was introduced by Md. Noor & Anderson (2007) to
875
explain the soil behaviour due to loading and wetting. The framework takes the curved-surface mobilized shear strength envelope as the soil yield surface. Whenever the Mohr circle representing the stress state extends higher than the envelope the soil will yield. Loading collapse is settlement caused by the increase in loading as illustrated in Figure 1. When a soil is subjected to an increase in loading, the mobilised shear strength envelope rotates to a new position which is tangential to the enlarged Mohr circle to indicate soil compression. The rotation of the mobilised shear strength envelope is represented by the increase in the minimum mobilized friction angle, minmob as illustrated in Figure 1 and this is taken as the the indication of the increase in axial strain, a during the application of deviator stress in during shearing stage in a consolidated drained triaxial test. Wetting collapse is settlement triggered due to inundation. This is a complex phenomenon since it is occurring under no load increase. Even critical state approach faced some restrictions in explaining this behaviour as reported by Wheeler et al. (2003). Figure 2 illustrates the interaction between mobilised shear strength envelope and the Mohr circle during wetting which reduces suction, resulting in inundation settlement. A constant load is applied to the soil at a specific suction is represented by the Mohr circle located at the most right. When the soil is wetted during the rise of ground water table, suction reduces causing the Mohr circle to move towards the net stress axis which is towards the lower part of the envelope and thereby extends higher than the envelope. This causes the mobilized shear strength envelope to rotate about suction axis to reinstate a new position with the Mohr circle touching tangentially the envelope. The degree of rotation of the mobilized envelope is representing the magnitude of the axial compression. Consolidated drained triaxial tests were conducted on saturated specimens of granitic residual
Figure 2. Progression of the effective stress Mohr circle due to the loss of suction causing the mobilised shear strength envelope to rotate during wetting collapse (Md. Noor & Anderson, 2007).
soil at different effective stresses to obtain the stressstrain curves and deduced the shear strength envelope at failure. The unique relationship between minmob and a was then plotted from the stress-strain curves at identical axial strain for tests at different effective stresses. The unique relationship curves for different effective stresses nearly overlap producing similar trends as shown in Figure 3. The relationship minmob a is best represented by Equation 1, where is named as the soil coefficient of anisotropic compression (Jais and Md Noor 2009). takes the value of 0.335, 0.39 and 0.24 for effective stress of 50, 100 and 200 kPa respectively. min f minmob ln min f a
(1)
The shear strength envelope at failure is obtained from the three triaxial tests on saturated specimens. Using the unique relationship of the soil, the settlement can be predicted from the initial minimum mobilized friction angle, minmobinitial and the final minimum mobilized friction angle, minmob as shown in Figure 4. In drained trifinal axial test condition where the cell pressure and the pore water pressure are maintained constant then the minor principle stress, 3 is constant. Each of the mobilized strength envelopes represents a specific magnitude of axial strain; therefore, the difference in the axial strains, a can be applied to determine the settlement the soil of a known thickness, D. The change in axial strain is given by: ln
min f
min
mob final
Figure 1. Rotation of the mobilised shear strength envelope as the soil is loaded enlarging the effective stress Mohr-circle (Md. Noor & Anderson, 2007).
a
min
ln
min f
min
min
mobinitial
(2)
876
Table 1.
Basic properties of soil samples. Description 46.17% 31.00% 8.57% 14.27% 2.72 41% 30% 11% 15%
Parameters Gravel content Sand content Silt content Clay content Specific gravity Liquid limit Plastic limit Plasticity index Moisture content
Figure 3. Unique relationship; minmob a for saturated granitic residual soils at effective stresses of 50 kPa, 100 kPa and 200 kPa.
700
600
Rawang is reddish brown in colour which can be classified as very silty sandy GRAVEL based on the particle size distribution. 4 SIMULATION OF SETTLEMENT FROM PHYSICAL MODELLING OF FOOTING
500
Shear Strength, kPa
400
Failure Envelope
300
Final Strength Envelope Initial Strength Envelope
200
100
80 110
0 100 200 300 400 500 600 700 800 900 1000
uw) kPa Effective Stress, ( -
Figure 4. Soil shear strength response based on initial and final mobilized shear strength envelope of the soil deduced from the shear strength envelope at failure.
The settlement can be calculated as follows: s aD (3)
Where, D is the thickness of the soil layer. By substituting Equation 2 into Equation 3, the proposed settlement equation deduced from the change of axial strain is given by:
s D min f minmob final min f minmobinitial (4) ln ln min min f f 100
Types of tests performed on compacted granitic residual soil were: (1) the soil is loaded to failure (2) the soil is loaded and inundated (3) the soil is loaded, inundated and subsequent lowering of the water table. A model footing of 200 mm 200 mm size and 50 mm in thickness was used in the laboratory physical modeling. The overall set-up is shown in Figure 5a and the closer view of the model footing and its instrumentation is shown in Figure 5b. The soil sample was prepared in a cylindrical metal tank with the diameter of 600 mm and a height of 1200 mm. Two drainage valves were fitted at the bottom of the tank to allow the water table to be raised and lowered. The height of the water table was monitored using an externally attached transparent pipe connected to the bottom of the tank. The loads were applied using an electrically operated and mechanically controlled loading system with a load cell attached at the plunger. Four linear variable displacement transducers (LVDT) were installed at each corners of the model footing to measure the settlements due to loading and wetting. 4.1 Test series applied in the physical model
SOIL PROPERTIES USED IN THE MODEL
Remoulded granitic residual soil of weathering grade VI obtained from a cut slope at Rawang, Selangor was used in this study. The soil sample was obtained at about 1 meter below the ground surface using a shovel. The soil sample was quickly placed in dark tight plastic containers and stored in the laboratory until it is required for testing. Table 1 shows the properties of the soil samples. The granitic residual soil sample obtained from
The soil sample was taken from plastic containers and then oven-dried for 24 hours. Then, the soil was prepared in the calibration tank and compacted under moisture content of 15% in layers of 100 mm thickness using a 2.5 kg hand compactor with 50 tampings every layer. The total thickness of the compacted soil was 1000 mm. 4.1.1 Test series1: Loading In this test, the loads applied were 50 kPa, 100 kPa, 150 kPa, 200 kPa, 250 kPa, 300 kPa, 400 kPa,
877
Figure 6. Loading stages and settlements of the footing.
Table 2. Loading and settlement of the physical model. Load (kN) 6.31 9.44 12.56 15.69 18.81 25.06 31.31 37.56 43.81 50.06 Stress (kPa) 100 150 200 250 300 400 500 600 700 800 Settlement (mm) 10 28 46 62 77 106 133 153 172 177
Figure 5. Physical model setup for simulating loading and wetting collapse settlements in a calibration tank.
500 kPa, 600 kPa, 700 kPa and 800 kPa. Each load applied was maintained until settlement is constant before the following load increment was applied. Figure 6 illustrates the axial displacements versus load applied for the loading test series. Notice that the settlement graph is level before the next load increment was applied. This is to ensure that the soil has fully responded to the applied stress before the next stress increment is applied. In other words, the loading collapse settlement is fully mobilized in every loading stage. Figure 6 illustrates the loading stage and settlement experienced by the soil. Table 2 represents the load and settlement values from the physical model. 4.1.2 Test series 2: Loading and wetting The loads were applied starting from 50 kPa and then load increments were applied similar as in the first test series. At 300 kPa, the loading
was maintained and the settlement was noted as loading settlement. At this stage, the drainage valves were opened and water was allowed to flow through the bottom aggregate layer into the soil sample. The rising of the water level was monitored periodically in the transparent pipe. The drainage valves were closed when the water level reached the soil surface. The resultant settlement was noted as inundation settlement. The result of this test is shown in Figure 7. Figure 7 illustrates the stages of the physical model during loading and wetting at 300 kPa. During the loading stage, settlement value observed was 70 mm. At this stage, the load was then maintained and the groundwater table was raised to the foundation level causing the footing to settle at 5 mm. This settlement is known as wetting collapse or inundation settlement. 4.1.3 Test series 3: Loading, wetting and subsequent lowering of water table This test series is actually the continuation of test series 2. At the end of full inundation in test series 2, the water table was lowered by opening the drainage valves until there was no further discharge observed at the outlet pipe. This was followed by another inundation and subsequent lowering of water table.
878
700
Axial displacement due to wetting at 300kPa

25.00 300.00 280.00 260.00 20.00 300kPa 240.00 220.00 200.00 180.00 160.00 140.00
600
500
Shear Strength, kPa
Load (KN)
15.00
400
Failure Envelope
300
Final Strength Envelope
10.00
Stop wetting Start wetting (1)
120.00
Start Draining
100.00 80.00 60.00 40.00 20.00
200
5.00
Wetting Collapse (5mm) Loading Collapse (70mm)

0 0.5 1 1.5 2 2.5 3 3.5 4 4.5 5 5.5 6 6.5 7 7.5 8 8.5 9
Initial Strength Envelope

100
LOAD (KN) LVDT (mm)
0.00 9.5 10 10.5 11 11.5 12 12.5 13 13.5 20.00
75
0 0
0.00
Time ('000 sec)
100
100
200
300
400
500
600
700
800
900
1000
uw) kPa Effective Stress, ( -
Figure 7. Wetting and drying stages simulated in the physical model.
The resulting settlement due to lowering of water table was noted as drying settlement. The result is shown in Figure 7. Apparently the graph shows there is no subsequent settlement due to the lowering of the water table when the soil has experienced a complete wetting collapse before hand. 5 PREDICTION OF SETTLEMENTS USING ROTATIONAL MULTIPLE YIELD SURFACE ENVELOPE
Figure 8. Settlement prediction for loading collapse of granitic residual soil at Rawang, Selangor using Rotational Multiple Yield Surface Framework of Md Noor and Anderson (2007).
700 SETTLEMENT ANALYSIS 600 Layer Thickness 1.00 = Pressure, q = 140. kP Change in Suction80.0 = 4.07 Settlement = mm m
500 Shear Strength kPa
400
Failure Envelope for Net stress at failure
300
200
Final Strength Envelope Initial Strength Envelope
100
Figure 8 shows the prediction of loading settlement applying the Rotational Multiple Yield Surface Framework for the granitic residual soil. The initial applied vertical stress of 160 kPa, represented by Mohr circle A considering ko condition with an initial mobilized friction angle, minmobinitial of 9.61. When an applied vertical pressure of 250 kPa is applied to the sample, the Mohr circle is enlarged and is represented by Mohr circle B, rotating the mobilized shear strength envelope to a new position with the minimum mobilized friction angle, minmob final of 23.5. The predicted settlement analysis gave 43.75 mm whereas the results given in the physical model was recorded as 46 mm. This show that the prediction using Rotational Multiple Yield Surface Framework is in good agreement with the physical model conducted. 5.1 Loading and inundation settlement
75 100 100
C
200 300 400 500 600 Net Stress ( - ua) kPa 700 800 900 1000
(a)
700
600
500 Shear Strength kPa
400
300
200
100
E
0 0
D100
200
300
400 500 600 Suction (ua - uw) kPa
700
800
900
1000
(b)
The initial suction established from the soil water characteristic curve is 80 kPa for partially saturated granitic residual soil at 15% moisture content. This value is used to predict inundation settlement using Rotational Multiple Yield Surface Framework. Figure 9 illustrates the rotation of the mobilized shear strength envelope at a constant net stress where there is a change of suction from 80 kPa to zero (full saturation). Consider an applied vertical pressure of 300 kPa is applied to the soil represented by Mohr circle C
Figure 9. The rotation of the mobilized shear strength envelope at constant effective stress of 300 kPa when suction is reduced from 80 kPa (i.e., line D) to zero at line E (a) viewed at shear strength-net stress plane (b) shear strength-suction plane.
and a minimum mobilized friction angle, minmob of initial 10.94 as shown in Figure 9(a). Notice that upon inundation the mobilized shear strength envelope rotates and at full saturation the envelope cuts the origin. By looking at the shear strength-suction plane in Figure 9(b) the initial suction of 80 kPa is represented by a Mohr circle as line, D touching the initial shear strength envelope. When suction is reduced to zero at full saturation the final position
879
Table 3. Settlement predictions and physical model test results at an applied pressure of 300 kPa. Settlement (mm) Method Rotational multiple Yield Surface Framework Physical Model Loading 43.57 46 Wetting 4.07 5
and this phenomenon explains inundation settlement as reported by Alonso et al. (1990) and Blanchfield and Anderson. (2000) The Rotational Multiple Yield Surface Framework of Md. Noor and Anderson (2007) can simulate theoretically the settlement due to loading and wetting. The predictions of loading and wetting settlements using Rotational Multiple Yield Surface Framework are in good agreements with the physical model results obtained in the laboratory. ACKNOWLEDGEMENT This research was conducted with the financial support from the Fundamental Research Grant provided by the Ministry of Higher Education of Malaysia and managed by Research Management Institut, Universiti Teknologi MARA, Shah Alam, Malaysia. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990, A constitutive model for partially saturated soil. Geotechnique, 40(3), 405430. Bishop, A.W. 1959. The principle of effective stress. Teknisk Ukeblad, 106(39), 859863. Blanchfield, R. & Anderson, W.F. 2000. Wetting collapse in opencast coalmine backfill. Proceedings of the ICE Geotechnical Engineering, London: 139149. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsaturated soils. Canadian Geotechnical Journal, 15(3), 313321. Jennings, J.E. & Burland, J.B. 1962. Limitations to the use of effective stresses partly saturated soils. Geotechnique, 12(No.2): pp.125144. Md. Noor, M.J. & Anderson, W.F. 2006. A comprehensive shear strength model for saturated and unsaturated soils. Proceedings of the 4th International Conference on Unsaturated Soils, ASCE Geotechnical Publication No. 147, Carefree, Arizona, USA, 2, 19922003. Md. Noor, M.J. & Anderson, W.F. 2007. A Qualitative Framework for Loading and Wetting Collapses in Saturated and Unsaturated Soils. Proc. 16th South East Asian Geotechnical Conference, Kuala Lumpur, Malaysia. Mohamed Jais, I.B. & Md. Noor, M.J. 2009. Establishing unique relationship between minimum mobilised friction angle and axial strain for anisotropic soil settlement model. Asia Pacific Conference on Unsaturated Soil. Newcastle, Australia. Terzaghi, K. (1936). The shear resistance of saturated soils. Proceedings for the 1st. International Conference on Soil Mechanics and Foundation Engineering (Cambridge, MA), 1, 5456. Terzaghi, K. 1943. Theoretical soil mechanics. New York, Wiley Publications. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. (2003). Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Geotechnique, 53(1), 4154.
of the Mohr circle is represented by the line E. The mobilized shear strength envelope rotates giving a final minimum mobilized friction angle, minmob final of 12.93 causing the soil to settle at 4.07 mm as shown in Figure 9(b). Table 3 presents the results for loading and wetting collapse settlement for the granitic residual soil from prediction and laboratory modelling. 5.2 Effect of lowering of water table after wetting collapse
The third series of the modeling is to find out whether there is any further settlement due to the lowering of the water table after the soil experiences wetting collapse. As shown in Figure 7, no settlement is observed as the water is drained out from the model. As water table is lowered there was a reduction in the pore water pressure underneath the footing. Thence there must be an increase in the effective stress there. Therefore the Mohr circle which is representing the effective stress state there must be slightly driven to the right. In doing so it would not extend above the current mobilized shear strength envelope as the envelope is getting higher towards the right. This explains the reason for no settlement due to this lowering of the water table. However if the load is increased right after the wetting collapse, would there be any settlement? The increase in the load would make the Mohr circle C in Figure 9(a) becomes bigger. This would certainly cause it to extend above the current mobilized shear strength envelope. And this thereby would trigger another settlement. This is how the Rotational Multiple Yield Surface Framework explains the occurrence of settlement by considering the interaction between the Mohr circle and the current mobilized shear strength envelope. Any rotation of the mobilized envelope is to represent settlement. This happens whenever the Mohr circle extends higher than the current mobilized shear strength envelope. 6 CONCLUSION
Based on the results and observation, it can be concluded that Settlement occurs due to wetting of partially saturated soil as observed in the physical model
880
Hydraulic hysteresis coupled to volume change effects in unsaturated soil behavior

M. Morvan, H. Wong & D. Branque
Universit de Lyon, Dpartement Gnie Civil et Btiment, Ecole Nationale des Travaux Publics de lEtat, CNRS, Vaulx-en-Velin, Lyon, France
ABSTRACT: A new unsaturated soil model was recently developed by our team, based on the concept of bounding surface plasticity. This model only requires 13 parameters for its definition and can simulate a few behaviors typically observed on unsaturated fine sands which more classical models fail to reproduce. These include the post-peak softening, change from contractant to dilatant behavior during shearing. Since the first version of this model, a few improvements have been made. In particular, the hysteretic effects of cyclic variations of suction and water content and the existence of air-entry and residual suction are now taken into consideration. The theoretical formulation is based on classical concepts of hardening plasticity, combining with a few empirically deduced formulae. More importantly, the effects of porosity change on water retention characteristics are accounted for. The latest version of the proposed unsaturated soil model accounting for hydraulic hysteresis is validated using existing experimental data. 1 INTRODUCTION model developed by us. Another model of this kind has been proposed by (Wheeler et al. 2003) but the determination of the constants seems to be a real problem. (Sheng et al. 2004) (Tamagnini 2004) presented a method to incorporate hysteresis to their model. Our development takes inspiration from the previous work of (Khalili et al. 2008), (Sun et al. 2007), (Sun et al. 2008) and (Nuth & Laloui 2008) on hydro-mechanical coupling. 2 HYDRO-MECHANICAL COUPLING
Unsaturated soils are a three phase mixture: grains water and air. The presence of two fluids generate tension surface effects that can be expressed as a function of the suction s ua uw, which is the difference between pore air pressure ua and pore water pressure uw. It acts as an apparent attraction between grains. We have already developed a model to represent the basic behaviour of the skeleton of unsaturated soils (Wong et al. 2009). This model is based on bounding surface plasticity, using the plastic driver of Bardet (1986) and associated with the concept of effective stress for unsaturated soils. During wetting or drying, a curve of the water content w or saturation degree Sl in terms of suction can be drawn. Such curve is named retention curve. It has been shown that the retention curve in wetting and in drying are different leading to hysteresis during cyclic variations. As a result, to the same suction can correspond two distinct values water content hence saturation degrees. It is well-known that suction changes can lead to irreversible strains. Furthermore (Gallipoli et al. 2003) showed the importance to take into account the effect of saturation degree in the mechanical behaviour. Recent studies (Sun et al. 2007) show that the retention curve actually depends on void ratio e and on mean pressure p. The mechanical state therefore affects the hydaulic behaviour of soils. The purpose of this paper is to incorporate this hydro-mechanical coupling to an unsaturated soil
2.1 Effective stress It has been well-established that to describe the behaviour of unsaturated soils, two independent stress variables were needed. The combination of (a) the difference between total stress and air pressure, called net stress ua , and (2) the difference between pore air and pore water pressures, called suction s ua uw, is the most currently used. A few other combinations were also proposed. For example, Wheeler, Sharma and Buisson (2003) introduced the couple of stress variables defined by equations (1) and (2). The porosity n and the saturation degree Sl appear directly in the expressions of these variables.
* ij ij ij [Sl uw ua (1 Sl )]
(1) (2)
s* n[ua uw] ns
881
Similar to the approach adopted by these authors and following our original model (Morvan et al. 2009), we prefer to use an unsaturated effective stress instead of the net stress. This effective stress is a generalisation of Terzaghis classic definition to unsaturated soils. It ensures continuity when approaching full saturation. A few different expressions for the effective stress have been proposed. In (Bishop 1959), the following version was presented:
ij ij ij [ ua (ua uw )]
(3)
where is a parameter varying between 1 for fully saturated state and 0 for completely dry state. In the sequel, the above definition of effective stress will be adopted with Sl. In other words:
ij ij ij [ ua sSl ]
(4)
Figure 1. Intrinsic shape of soil water retention curves for loam: (Sl ln s) relationship at constant void rations data from (Sugii, Yamada and Kondou, 2002) interpreted by (Nuth and Laloui 2008).
The effective stress alone cannot describe fully the behavior of unsaturated soils. Suction s will be taken as the second stress variable. 2.2 Retention curves
We choose to define the retention curve in the (s,w) plane, w being the water content, for two reasons: Water content is an experimental data directly measurable whereas the saturation degree has to be calculated by accounting for the volumetric strain. Under undrained conditions (relative to pore water), water content will remain constant whereas the saturation degree will vary. This leads to simplifications. During monotonic drying, starting from full saturation or wetting, starting from a very dry state, the hydric behaviour is defined by two main retention curves. w s wsat s
Figure 2. Wettingdrying cycles performed on Pearl clay specimen compacted at different void ratios under isotropic net stress of 20 kPa and e0 1.08.
s sini se. For hydric loadings along one of the two master curves, differentiation of equation (5) gives: ds de ds dw w e s de s (6)
(5)
where s sae is the air-entry suction for drying and s sex the air-expulsion suction for wetting, at the limit of full saturation, while wsat is the water content at full saturation. 2.3 Hydro mechanical coupling
For hydric unloadings or reloadings along a scanning curve in-between the two master curves, the slope of the curve is lower (Sun et al. 2007), as illustrated in Figure 2. In this case we postulate the following relation: dw s ds s (7)
Figure 1 gives an experimental illustration of the dependence of the air-entry suction on the void ratio. In general, s depends on e while the distance between the master drying and wetting curves will remain constant on the (Ln(s), w) plane. During a suction increase, void ratio changes inducing changes of s, so that at the current state
Using the trace of (4) and the classic relation: Sl w Gs e (8)
882
it can be deduced that: dp dp w s sw Gs ds Gs dw 2 Gs de e e e (9)
where p and p are the classic mean total stress and mean effective stress. Combining with the constitutive equations relating the skeleton strain increments to their stress increments, we obtain the coupled equations: de de(dp, dq, ds ), dw dw(dp, dq, ds ) (10)
Figure 3. model. The radial mapping algorithm in Bardets
Based on the experimental observations reported by (Huang et al. 1998) an increment in the void ratio is supposed to induce a proportional change of both the air-entry and and air-expulsion suction, so that: se k3e, k3 0 (11)
The conjugate strains are the volumetric p and deviatoric q strains, defined by:
Figure 1 shows that the air entry suction se is decreasing with the void ratio e, so the parameter k3 has to be negative. The above equations on the hydric behaviour will be incorporated into the unsaturated soil (Morvan et al. 2009) model previously developed by our team in the next paragraph. 3 3.1 MECHANICAL BEHAVIOUR Bounding surface plasticity
p 1 2 2 ; q
2 2 3 1
(13)
We assume that all strain components can be decomposed into an elastic and a plastic part:
p e p p e p p ; q q q
(14)
Bardets model has been adapted to unsaturated soils, the bounding surface equation is given by:
One of the aims of our model was to obtain a progressive transition between elastic and plastic behaviors, so we decide to use bounding surface plasticity concept. In this type of elasto-plasticity, the plastic modulus depends on the distance between the current stress state and its image-point, obtained by projection on a surface called bounding surface. Further details on bounding surface plasticity can be found in Dafalias (1986), Dafalias & Hermann (1986), Manzari & Dafalias 1997, Crouch et al. 1994, Russel & Khalili 2004, Yu 2006). 3.2 BDNS a new unsaturated soils model
f( p, q, p p , s)
with
p A q 2 M A 1 (15) (16)
p A A ( p p ,s) l1 ( s )A( p ) l2 ( s )
The model developed in this paper is based on Bardets model (Bardet 1986). In this model, there is no purely elastic region and hence only one surface, the bounding surface. We work in the cylindrical symmetry and adopt the classic notation of triaxial variables (p,q), where p has already been defined, q is the deviatoric stress and is the stress ratio. To summarize: q 1 p ( 1 2 2 ); q 1 2; p 3 (12)
where l1, l2 are functions of suction, A is the hardening variable depending on the plastic volumetric strain. M is the slope of the critical state line at full saturation and is considered constant in the (p,q) plane. There is not enough experimental data at present to define precisely the suction-dependence of A. Hence formulas covering the most general case are conjectured. The image points are obtained by radial mapping: p A ; q xMA ; x with: q Mp (17)

1 ( 1) 1 x 2 ( 2 ) 1 ( 1)2 x 2
(18)
883
The definition of the plastic modulus and of the plastic multiplier used with the consistency relation give us:
p f d p Hbd f p p
p d q
1 ( n p dp nq dq ns ds ) nq H
(29)
Elastic strains are defined by classical relation: d e p 2(1 ) dq dp ; d e q (1 e0 ) p 9(1 2 ) 1 e0 p (30)
(19)
The gradient of f is made of three terms:
f f f f ; ; p q s
(20)
where and are two material parameters, respectively the volumetric deformability and the Poisson's ratio. e0 is the initial void ratio. Using experimental data, we assume the following forms for the functions l1 and l2 defining A. l1(s) 1 k1 (sSl se); l2(s) 0 (31)
After few calculations, the normalize components of this gradient of f are given by:
f p M( 1) np f g
f x( 1)2 q nq f g f ns s f where: g M( 1)2 2 A f
(21)
The suction-dependence of the critical state is chosen as: (s) 0 k2 (sSl se) 4 VALIDATION OF THE MODEL (32)
(22)
(23)
(24)
To validate our model, we use two different materials. First to make sure that all the unsaturated soils mechanical basic features are taken into account, we consider triaxial tests at constant suction. For this comparison we use the data of Russel & Khallili (2006), that present systematic post-peak softening and transition between contractancy and dilatancy. Then we consider the hydro-mechanical coupling by comparing the simulation and experimental data of a wetting drying test. To do so we use the data of Raveendiraraj (2009) on speswhite kaolin. 4.1 Triaxial tests with constant succion
The plastic modulus can then be calculated: Hb 1 e0 Al1( s ) (M )2( 1)[ ( 2 )] (s) g2 (25)
When the stress point lie inside the bounding surface, a second term has to be added to the plastic modulus: H Hb Hf We can take: Hf (26)
p 1 e0 h0 p M ( s ) max
(27)
Plastic strains are given by: dp p 1 ( n pdp nq dq ns ds ) n p H (28)
Russel & Khalili (Russel & Khalili 2006) have performed an experimental campaign on Kurnell sand. Triaxial drained tests were made on dense and loose specimen respectively at a confining pressure of 100 kPa and 50 kPa. Four suctions have been tested 50, 100, 200 and 400 kPa. These tests have been simulated with a program on Matlab using parameters of Table 1. Here we present the comparison of simulation and experimental data on triaxial tests at a constant suction of 50 kPa (Fig. 4). Figure 4a shows in the (q,1) plane the post peak softening which is well reproduced by the simulations. Overall, figure 4 shows good agreement between experimental data and simulations as well as on the transition between contractancy and dilatancy. These two figures prove the efficiency of our model to describe the behavior of fine sands, and to reproduce the basic features of unsaturated soils.
884
Table 1. Parameters used for simulation of Kurnell sand. Parameters M p h0 se (kPa) k1 k2 0 (kPa) Value 0.006 0.3 0.7 1.6 1 1.9 5 6 0.2 5 0.1 15.1
Two tests have been chosen: A9 and A11 wetting-drying path under a mean net stress of 10 kPa. In A9, the wetting-drying path starts after a supplementary isotropic loading and unloading at 200 kPa so the void ratio is lower than in A11 at the beginning of wetting. The results are presented in two planes: (e, s) in Fig. 5a and (vw , s) in Fig. 5b, where vw is defined as the specific water volume
Table 2. Parameters used for simulation of Speswhite. Parameters s M p h0 sex (kPa) sae(kPa) k1 k2 0 (kPa) K3 Value 0.03 0.2 0.01 2.2 0.72 1.17 0.73 5 0.011 0.03 8.5 2.5 0.07 9 0.025
Figure 4. Deviatoric stress (a) and volumetric strain (b) vs axial strain at a suction of 50 kPa. Comparison between experimental data (symbols) and model prediction (solid or dotted lines).
4.2
Wetting-drying paths
To test the performance of the model to simulate hydric hysteresis, we will next consider a particular loading path: the wetting-drying path. Data from (Raveendiraraj 2009) on speswhite kaolin will be used for model validation because of the stress path needed to obtained parameters and the large quantity of data available on this material. Parameters are given in parts in Raveendiraraj (2009), and in parts in other publications dealing with speswhite kaolin, such as Gallipoli, Wheeler & Karstunen (2003). The parameters adopted are summarized in Table 2.
Figure 5. Void ratio (a) and Specific water content (b) vs suction at mean net stress of 10 kPa. Comparison between experimental data (symbols) and model prediction (solid or dotted lines).
885
vw 1 Gsw. The simulations are close to experimental data both in (e, s) and (vw , s) planes. With only 15 parameters, the model proves its efficiency. 5 CONCLUSION
By accounting for the hysteretic water retention behaviour in our model, we obtain a 15-parameters model on saturated soils that is able to reproduce complex mechanical and hydric behaviours. Its precision and pertinence have been verified on mechanical and hydric loading-paths. The paths chosen were triaxial tests at constant suction and wetting-drying paths. The model can reproduce post peak softening, transition between contractancy and dilatancy, but also the irreversible strains dues to hysteresis, the variation of water content during suction variations. During the validation process, we notice a lack of extensive experimental campaign, in particular those coupling extensive hydrological campaign with mechanical campaign. To complete experimental data and the validation of this model an experimental campaign on unsaturated sand is in progress. The model could be applied to more complex hydro-mechanical paths. And also be incorporated into a finite element code in order to analyze complex geotechnical structures. REFERENCES
Bardet J.P. 1986. Bounding Surface Plasticity Model for Sand. Journal of engineering mechanics 112(11): 11981217. Bishop A.W. 1959. The principle of effective stress, Teknisk Ukeblad 106(39): 859863. Dafalias Y.F. 1986. Bounding Surface Plasticity I: Mathematical Foundation Hypoplasticity. Journal of engineering mechanics 112: 966987. Dafalias Y.F. & Herrmann, L.R. 1986. Bounding Surface Plasticity II: Application Isotropic Cohesive Soils. Journal of engineering mechanics 112: 12631991. Gallipoli D., Wheeler S.J. & Karstunen M. 2003. Modelling the variation of degree of saturation in a deformable unsaturated soil. Gotechnique 53(1): 105112. Huang S.Y., Barbour S.L. & Fredlund D.G. 1998. Development and verification of a coefficient of permeability function for a deformable unsaturated soil. Canadian Geotechnical Journal 35(3): 411425. Khalili N., Habte M.A. & Zargarbashi S. 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Computers and Geotechnics 35: 872889. Manzari M.T. & Dafalias, Y.F. 1997. A critical state twosurface plasticity model for sands. Gotechnique 47(2): 255272.
Morvan M., Wong H. & Branque D. 2009. An unsaturated soil model with minimal number of parameters based on bounding surface plasticity. International Journal for Numerical and Analytical Methods in Geomechanics. Article in press. Nuth M. & Laloui L. 2008. Advances in modelling hysteretic water retention curve in deformable soils. Computers and Geotechnics 35: 835844. Pereira J.M., et al. 2005. Adaptation of existing models to unsaturated states: application to CJS model. International Journal for Numerical and Analytical Methods in Geomechanics 29(11): 11271155. Raveendiraraj A. 2009 PhD thesis: Coupling of mechanical behavior and water retention behavior in unsaturated soils: University of Glasgow, Department of Civil Engneering. Russel A.R. & Khalili N. 2004. A bounding surface plasticity model for sands exhibiting particle crushing. Canadian Geotechnical Journal 41: 11791192. Russel A.R. & Khalili N. A 2006. Unified Bounding Surface Plasticity Model for Unsaturated Soils. International Journal for Numerical and Analytical Methods in Geomechanics 30: 181212. Sheng D., Sloan S. & Gens A. 2004. A constitutive model for unsaturated soils: thermomechanical and computational aspects. Computational Mechanics 33: 453465. Sugii T., Yamada K. & Kondou T. 2002. Relationship between soil-water characteristic curve and void ratio. UNSAT 2002, Proc. 3rd Unsaturated Soils Conference. Juca. Recife, de Campos and Marinho: 209214. Sun D., Sheng D., Cui H.B. & Sloan S.W. 2007. A density-dependent elastoplastic hydro-mechanical model for unsaturated compacted soils. International Journal for Numerical and Analytical Methods in Geomechanics 31: 12571279. Sun, D., Sheng D., Xiang L. & Sloan S.W. 2008. Elastoplastic prediction of hydro-mechanical behaviour of unsaturated soils under undrained conditions. Computers and Geotechnics 35: 845852. Sun, D., Sheng D. & Xu Y.F. 2007. Collapse behaviour of unsaturated compacted soil with different initial densities. Canadian Geotechnical Journal 44: 673686. Tamagnini R. 2004. An extended Cam Clay model for unsaturated soils with hydraulic hysteresis. Gotechnique 54(3): 223228. Wheeler S.J., Sharma R.S. & Buisson M.S.R. 2003. Coupling of hydraulic hysteresis and stress strain behaviour in unsaturated soils. Gotechnique 53(2): 4154. Wong H., Morvan M. & Branque D. 2009. A 13-parameters unsaturated soil model based on bounding surface plasticity. International symposium on unsaturated soil mechanics and deep geological nuclear waste disposal UNSAT-WATSTE 2009. W.M.Y. Y.J. Cui, B. Chen & Y.G. Chen (eds). Shanghai-China:216223. Yu H.S. 2006. Plasticity and Geotechnics: Springer.
886
Mapping soil-water profile utilizing non-linear neural network based model

Muhammad Mukhlisin
Department of Civil and Structural Engineering, Universiti Kebangsaan, Malaysia Department of Civil Engineering, Polytechnic Negeri Semarang, Indonesia
Ahmed El-Shafie & Mohd Raihan Taha
Department of Civil and Structural Engineering, Universiti Kebangsaan, Malaysia
ABSTRACT: Modeling unsaturated water flow in soil requires knowledge of the hydraulic properties of soil. However correlation among soil hydraulic properties such as the relationship between saturated soil water content S and saturated soil hydraulic conductivity Ks as function of soil depth is in stochastic pattern. On the other hand, soilwater profile process is also believed to be highly nonlinear, time-varying, spatially distributed, and not easily described by simple models. In this study, the potential of implementing Artificial Neural Networks ANN model was proposed and investigated to map the soil-water profile in terms of Ks and S with respect to the soil depth d. Site experimental data sets on the hydraulic properties of weathered granite soils were collected. These data sets include the observed values of saturated and unsaturated hydraulic conductivities, saturated water contents, and retention curves.The proposed ANN model was examined utilizing 49 records data collected from field experiments. The results showed that the ANN model has the ability to detect and extract the stochastic behaviour of the water in the soil and draw the water-soil profile with relatively high accuracy. 1 INTRODUCTION respectively. Many models for water retention (e.g., Brooks and Corey 1964; van Genuchten 1980; Russo 1998; Kosugi 1994; Fredlund and Xing 1994) have been developed, incorporating earlier models by Burdine (1953) and Mualem (1976), with the aim of deriving analytical expressions that can be used to predict the relative hydraulic conductivity of soil. The resultant water retention curve is considered one of the most fundamentally important hydraulic characteristics of soil (Assouline and Tessier 1998). The relationship between saturated soil water content S and saturated soil hydraulic conductivity Ks as function of soil depth is in stochastic pattern. However soilwater profile process is believed to be highly nonlinear, time-varying, spatially distributed, and not easily described by simple models. Most present model techniques are based upon conventional linear or nonlinear statistical models, such as regression analysis. Although these models have been utilized for many years to provide inflow prediction, they have several limitations to detect and mimic such highly stochastic pattern of watersoil profile. Recently, significant progress in the fields of nonlinear pattern recognition and system theory using Artificial Neural Networks (ANN) has been performed (e.g., Parasuraman et al., 2006;
The study of rainwater infiltration on forested hillslopes is a very important area of research in forest hydrology. Rainwater infiltration produces positive pore water pressures in soil layers and is closely related to the occurrence of slope failures (e.g., Mukhlisin et al., 2006, 2008). Soil data serve as an important initialization parameter for hydro-ecological and climatological modeling of water and chemical movement, heat transfer, or land-use change. Most soil hydraulic properties are difficult to measure and therefore have to be estimated in most cases. In addition, knowledge of the moisture content and water movement in a soil layer is fundamental to the analysis of biological reactions, plant activity, material transports in forest ecosystems, and stream water chemistry (Sklash and Farvolden 1979; Pinder and Jones 1969). To analyze rainwater infiltration into soil, it is important to have an understanding of the hydraulic properties of the soil, in particular the relationship between volumetric water content S and soil capillary pressure , and the relationship between unsaturated hydraulic conductivity K and . These relationships are known as the water retention curve and the hydraulic conductivity function,
887
Agyare et al., 2007; and Baker and Ellison, 2008). An ANN is a nonlinear mathematical structure, which is capable of representing arbitrarily complex nonlinear processes that relate the inputs and outputs of any system. Agyare et al., 2007 analyzed the measurement soil properties together with terrain attributes in Artificial Neural Networks (ANNs) to estimate saturated hydraulic conductivity (Ks), for two pilot sites in the Volta basin of Ghana. It was observed that good data distribution, range, and amounts are prerequisites for good ANN estimation and, therefore, data preprocessing is important for ANN. With adequate and sensitive data, ANN can be used to estimate Ks, using soil properties such as sand, silt, and clay content, bulk density, and organic carbon. Although the inclusion of terrain parameters can improve the estimation of Ks using ANN, they cannot be relied on as the sole input parameters as they yield poor results for the scale considered in this study. The source of training data was found to significantly influence the topsoil Ks, but the subsoil was not sensitive to training data source. Parasuraman et al., 2006 investigated the applicability of neural networks in estimating Ks at field scales and compared the performance of the fieldscale PTFs with the published neural networks program Rosetta, also compared the performance of two different ensemble methods, namely Bagging and Boosting in estimating Ks. Datasets from two distinct sites are considered in the study. The performances of the models were evaluated when only sand, Silt, And Clay Content (SSC) were used as inputs, and when SSC and bulk density b (SSCb) were used as inputs. The result showed that for both datasets, the field scale models performed better than Rosetta. The comparison of field-scale ANN models employing bagging and boosting algorithms indicates that the neural network model employing the boosting algorithm results in better generalization by reducing both the bias and variance of the neural network models. Although Artificial Neural Networks (ANNs) based PedoTransfer Functions (PTFs) have been successfully adopted in modeling soil hydraulic properties at larger scales (national, continental, and intercontinental), the utility of ANNs in modeling saturated soil water content S and saturated soil hydraulic conductivity Ks as function of soil depth has rarely been reported. 1.1 Problem statement
of weathered granite soils were collected from published (Ohta et al., 1985; Shinomiya et al., 1998; Hendrayanto et al., 1999) and unpublished studies (Shinomiya and Kosugi, personal communications). These data sets include the observed values of saturated and unsaturated hydraulic conductivities, saturated water contents, and retention curves. The proposed ANN model was examined utilizing 49 records data collected from field experiments. 1.2 Objective
The objective of this paper is to analyze the soilwater profile and to develop a robust prediction model of the soil moisture content at different depth utilizing ANN. While such a model is not intended to substitute physically based conceptual models, it can provide an accurate prediction for soil moisture content using only the hydraulic conductivity and the depth as available input and output time series data. The anticipated impact of this model is that it can predict the soil moisture content without the need to explicitly consider the internal geologic or hydraulic parameters.
2 2.1
METHODOLOGY Artificial Neural Network
In this study, the potential of implementing Artificial Neural Networks ANN model was proposed and investigated to map the soil-water profile in terms of Ks and s with respect to the soil depth d. A number of data sets on the hydraulic properties
Artificial Neural Networks (ANN) is densely interconnected processing units that utilize parallel computation algorithms. The basic advantage of ANN is that they can learn from representative examples without providing special programming modules to simulate special patterns in the data set (Gibson and Cowan 1990). This allows ANN to learn and adapt to a continuously changing environment. Therefore, ANN can be trained to perform a particular function by tuning the values of the weights (connections) between these elements. The training procedure of ANN is performed so that a particular input leads to a certain target output as shown in Fig. (1). Multi-layer ANN has been reported as a powerful modeling tool (Gibson and Cowan 1990, and Ooyen and Nienhuis 1992). The input and output layers of any network have numbers of neurons equal to the number of the inputs and outputs of the system respectively. The architecture of a multilayer feed-forward neural network can have many layers between the input and the output layers where a layer represents a set of parallel processing units (or nodes) namely, the hidden layer. The main function of the hidden layer is to allow the network to detect and capture the relevant patterns in the data and to perform complex non-linear mapping between the input and the output variables. The sole role of the input layer of
888
where netPJ is the weighted inputs into the j th hidden unit, n is the total number of input nodes, Wji is the weight from input unit i to the hidden unit j, xpi is a value of the i th input for pattern P, Wjo is the threshold (or bias) for neuron j, and g(netPJ) is the j th neurons activation function assuming that g is a logistic function. Note that the input units do not perform operation on the information, but simply pass it onto the hidden nodes. The output unit receives a net input of: net pk Wkj . g ( netPJ) Wko
Figure 1. Artificial neural network model diagram.
J 1 M
(4a) (4b)
ypk g(netpk)
nodes is to relate the external inputs to the neurons of the hidden layer. Hence the number of input nodes corresponds to the number of input variables. The outputs of the hidden layer are passed to the last (or output) layer, which provides the final output of the network. Finding a parsimonious model for accurate prediction is particularly critical since there is no formal method for determining the appropriate number of hidden nodes prior to training. Therefore, here we resort to a trial-anderror method commonly used for network design. In the prediction context, multi-layer feed forward neural network training consists of providing input-output examples to the network and minimizing the objective function (i.e. error function) using either a first order or a second order optimization method. Training can be formulated as one of minimizing a function of the weight, the sum of the non-linear least squares between the observed and the predicted outputs defined by: E 1 n (YO YP )2 2 P 1 (1)
where M is the number of hidden units, Wkj represents the weight connecting the hidden node j to the output k, Wko is the threshold value for neuron k, and ypk is the kth predicted output. Recalling that the ultimate goal of the network training is to find the set of weights Wji, connecting input units i to the hidden units j and Wkj, connecting the hidden units j to output k, that minimize the objective function (Eq. (1)). Since Eq. (1) is not an explicit function of the weight in the hidden layer, the first partial derivatives of E in Eq. (1) are evaluated with respect to the weights using the chain rule, and the weights are moved in the steepest-descent direction. This can be formulated mathematically as: Wkj E Wkj (5)
where n is the number of patterns (observations), Yo represents the observed response (target output) and Yp the model response (predicted output). In the back propagation training, minimization of the error function (see Eq. (1)) is attempted using the steepest descent method and computing the gradient of the error function by applying the chain rule on the hidden layers of the feed forward neural network. Consider a typical multi-layer feed forward neural network whose hidden layer contains M neurons. The network is based on the following equations: netPJ W ji x pi W jo
I 1 N
where is the learning rate, which scales the step size. The usual approach in back propagation training consists in choosing according to the relation 0 1 (Ooyen and Nienhuis 1992). The learning rule is a procedure for modifying the weights and biases of the network. This procedure may also be referred to as the training algorithm. The learning rule is provided with a known input/output set of data and an algorithm is then used to adjust the weights and biases of the network in order to move the network outputs closer to the targets. Therefore, modeling capabilities of networks trained using supervised learning algorithms are limited to the range of the input used in training the network. 2.2 ANN for modelling soil-water profile
(2) (3)
g ( netPJ )
1 e netPJ
ANNs model have been used in a broad range of applications including patterns classification, identification, prediction, optimization and control systems. ANNs learn by using some examples, namely patterns. In other words, to train and test a neural network, input data and corresponding target values are necessary, (Ripley 1996; Haykin, 1999).
889
In this study, 49 records for Hydraulic Conductivity (Ks) and Soil Water Content (SMC) with respect to soil Depth (D), obtained from (the experiment location), were used to train and test the model. Our pilot investigation showed that Soil Moisture Content (SMC) is based on Hydraulic Conductivity (KS) at certain depth D. Therefore, in this study, ANN with its nonlinear and stochastic modeling capabilities is utilized to develop a soilwater profile model. The Soil Moisture Content SMCdi predicted at certain depth di with respect to Hydraulic Conductivity HCdi can be expressed as: SMCd(i) f(HCd(i), d(i)) (6)
Figure 2. The exact neural network architecture utilized for soil water content SMCd(i). Table 1. The architecture of the neural network model. Layer Input layer Hidden layer I Hidden layer II Output layer Number neurons 2 4 2 1 Transfer function Tan sigmoid Pure line
The ANN model is established using the above equation. The architecture of the network consists of an input layer of two neurons (HCd (i), d(i)), an output layer of one neuron (corresponding to SMCd (i)) and a number of hidden layers of arbitrary number of neurons at each layer. In order to achieve the desirable prediction accuracy, ANN architectures were developed utilizing 40 records out of 49 experimental records in order to train the network. The performance and the reliability of the ANN model were examined using the rest 10 records. In order to accelerate the training procedure and to achieve minimum mean square estimation error, the records data was normalized (El-Shafie et al., 2008 and El-Shafie et al., 2009). All data in input and output layers were normalized in the (1, 1)(0, 1). Different MLP-ANN architectures (while keeping two neurons in the input layer and only one neuron in the output layer) were used to examine the best performance. The choice of the number of hidden layers and the number of neurons in each layer is based on two performance indices. The first index is the Root Mean Square (RMS) value of the prediction error and the second index is the value of the maximum error. The exact ANN architecture used for predicting Soil Moisture Content SMCd (i) is presented in Fig. (2). The number of hidden layers (R) and the number of neurons in each layer (N) for twelve networks are presented in Table 1. The transfer functions used in each layer of the networks are also listed in Table 1. All twelve networks utilize the back-propagation algorithm during the training procedure. Once the network weights and biases are initialized, during training process the weights and biases of the network are iteratively adjusted to minimize the network performance function mean square error MSEthe average squared error between the network outputs a and the target outputs t. There are several training algorithms for feedforward networks. All these algorithms use the
gradient of the performance function to determine how to adjust the weights to minimize performance. The gradient is determined using a technique called back-propagation, which involves performing computations backward through the network. The basic back-propagation algorithm adjusts the weights in the steepest descent direction (negative of the gradient), the direction in which the performance function decreases most rapidly. It turns out that, although the function decreases most rapidly along the negative of the gradient, this does not necessarily produce the fastest convergence. In the conjugate gradient algorithms a search is performed along conjugate directions, which produces generally faster convergence than steepest descent directions. Each of the conjugate gradient algorithms discussed so far requires a line search per iteration. This line search is computationally expensive, because it requires that the network response to all training inputs be computed several times for each search. The scaled Conjugate Gradient algorithm (SCG) was designed to avoid the time-consuming line search. This algorithm combines the model-trust region approach (used in the Levenberg-Marquardt algorithm), with the conjugate gradient approach (Bishop 1996). The scaled conjugate gradient criterion was used to update the ANN parameters while training, since it was reported that this method is the most suitable in case of high randomness on the input data, which is the case in this study (El-Shafie et al., 2008). This criterion is based on the conjugate gradient
890
Mean Square Error (MSE)
method, but with small modification that significantly saves computational time (Bishop 1996). Training of the network was performed by using LevenbergMarquardt (LM) feed-forward backpropagation algorithms. A computer program was performed under Matlab 6.5 software. Hyperbolic Tangent Sigmoid Function (HTSF) and Purelin Function (PF) were used as the transfer function in the hidden layer and output layer, respectively. These are shown by the Eqs. (7) and (8), respectively. y ex ex ex ex (7) (8)
10
10
-1
10
-2
10
-3
10
-4
10
20
30
y ax b
# Epochs
40
50
60
70
%Error-(Training Session)
Finally, the LevenbergMarquardt learning algorithm, which is a variant of back-propagation, was chosen as the learning algorithm of the model. The model was constructed, trained and tested using different number of neurons (from 2 to 30) in their hidden layer, using software developed by Matlab. For each network, RMSE value of the outputs was calculated. 3 RESULTS AND DISCUSSIONS
Figure 3. Training curve for the proposed ANN model utilizing 40 records of the data.
4 2 0 -2 -4 5 10 15 20 25 30 35 40
The ANN model architecture of Fig. (2) is employed in this study to provide prediction of soil moisture content SMCd(i). The measured soil moisture content SMCd(i) over the first 40 records was used to train the ANN model. The proposed ANN model successfully achieved the target MSE of 104. Figure 3 demonstrates the training curve and shows that the convergence to the target MSE is achieved after 73 iterations. In order to examine the performance of the proposed ANN, a simulation for the model output during the training is performed. Figure 4 illustrates the performance of the model over the first 40 records used for training. The statistical comparisons between predicted and measured soil moisture content SMCd(i) were performed by estimating the Prediction Error (PE) which measures the average squared error between the predicted SMCd(i) obtained from the model and the measured SMCd(i). The PE is described in Equation (9) PE 1 ( yti ypi )2 m (9)
Expr#
Figure 4. The error distribution for the ANN model during training session.
where ypi is the predicted value and yti is the experimentally measured value and m represents the number of samples in each testing group. Prediction errors for the two ANN networks is presented in Fig. (4). It is obvious from Fig. (4) that SMCd(i) prediction models using ANN have a maximum
error of 4% at only the experiments #1 and #8, while a maximum error 2% at the rest of the 40 records. In addition, it can be observed that almost 0.0% error for 10 experiments which is 25% of the whole examined records. As a result, the proposed ANN model successfully provides accurate predictions for SMCd(i) utilizing the Hydraulic Conductivity (Ks) at different Depth (D). To verify the performance of the proposed ANN-based soil moisture content, the experiments between #41 and #49 was used. Figure 5 shows the error distribution value of the soil moisture content error over these 9 experiments. Apparently, relatively higher levels of errors could be observed. However, the maximum error is within acceptable level of accuracy 15%. Furthermore, Figure 6 shows the neural network model output versus the actual saturated soil water content. It can be observed from figure 6 that the proposed neural network model output could mimic the dynamic pattern in the soil water content during training and testing.
891
%Error-(Test Session)
15 10 5 0 -5 -10 -15 2 3 4
CONCLUSION
Exp#
Figure 5. The error distribution for the ANN model during testing session.
The study analyzed the soil-water profile and developed a robust prediction model of the soil moisture content at different depth utilizing ANN. The model was successful to provide an accurate prediction for soil moisture content using only the hydraulic conductivity and the depth as available input and output time series data. The proposed ANN model was examined utilizing 49 records data collected from field experiments. The results showed that the ANN model has the ability to detect and extract the stochastic behavior of the water in the soil and draw the water-soil profile with relatively high accuracy. The performance accuracy of the model can be expressed from the training session that a maximum error 2% at the rest of the 40 records, while the maximum error for the testing session model is within acceptable level of accuracy 15%. REFERENCES
Agyare, W.A. Park, S.J. & Vlek, P.L.G. 2007. Artificial Neural Network Estimation of Saturated Hydraulic Conductivity. Vadose Zone J. 6: 423431. Assouline, S. & Tessier, D. 1998: A conceptual of the soil water retention curve. Water Resources Research, Vol. 34, No. 2, p.223231. Baker, L. & Ellison, D. 2008. Optimisation of pedotransfer functions using an artificial neural network ensemble method. Geoderma. 144: 212224. Bishop, C.M. 1996. Neural Networks for Pattern Recognition. 1st edition. Oxford: Oxford University Press. Brooks, R.H. & Corey, A.T. 1964: Hydraulic properties of porous media, Hydrol. Pap. 3, Civil Eng. Dept., Colo. State Univ., Fort Collins. Burdine, N.T. 1953: Relative permeability calculation from size distribution data. Trans. Am. Inst. Min. Metall. Pet. Eng. 198: 7178. El-Shafie, A. Abdin, A.E. Noureldin, A. & Taha, M.R. 2009. Enhancing Inflow Forecasting Model at Aswan High Dam Utilizing Radial Basis Neural Network and Upstream Monitoring Stations Measurements. Water Resources Management. 23: 22892315. El-Shafie, A. Noureldin, A.A. Taha, M.R. & Basri, H. 2008. Neural Network Model for Nile River Inflow Forecasting Based on Correlation Analysis of Historical Inflow Data. Journal of Applied Science. 8(24): 44874499. Fredlund, D.G. & Xing, A. 1994. Equations for the soilwater characteristic curve. Canadian Geotechnical Journal. 31(3): 521532. Gibson, G.J. & Cowan, C.F.N. 1990. On the Decision Regions of Multilayer Perceptrons. Proceedings of the IEEE. 78(10): 15901594. Haykin, S. 1999. Neural Networks: A Comprehensive Foundation. New York: Prentice Hall. Hendrayanto, Kosugi, K. Uchida, T. Matsuda, S. & Mizuyama, T. 1999. Spatial variability of soil hydraulic
Figure 6. Observed and predicted saturated soil water content for the ANN model during training and testing session.
In fact, it is common in ANN development to train several different networks with different architecture and to select the best one on the basis of performance of the networks with testing/validation sets. A major disadvantage of such an approach is that it assumes that performance of the networks for all other possible testing sets will usually be similar, which is statistically incorrect. Moreover, observing the performance of the developed ANN when tested, it is obvious that no single network has the optimal prediction for all the testing data sets. Therefore, a better accuracy than the best reported by any single network can be accomplished if an optimization algorithm that can utilize all these networks is developed. Another interesting observation is that, the effect of the transfer function is as important as the number of layers and neurons in each layer. This can be observed when comparing the performance of two networks with similar number of hidden layers and neurons but with different transfer functions. Further discussion on the effect of the optimal combination of different transfer function for specific applications is beyond the scope of this study.
892
properties in a forested hillslope. Journal of Forest Research. 4: 107114. Kosugi, K. 1994. Three-parameter lognormal distribution model for soil water retention. Water Resources Research. 32: 26972703. Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research. 12(3): 513523. Mukhlisin, M. Kosugi, K. Satofuka, Y. & Mizuyama, T. 2006. Effects of soil porosity on slope stability and debris flow runout at a weathered granitic hillslope. Vadose Zone Journal. 5: 283295. Mukhlisin, M. Taha, M.R. & Kosugi, K. 2008. Numerical analysis of effective soil porosity and soil thickness effects on slope stability at a hillslope of weathered granitic soil formation. Geosciences Journal. 12(4): 401410. Ohta, T. Tsukamoto, Y. & Hiruma, M. 1985. The behavior of rainwater on a forested hillslope, I, The properties of vertical infiltration and the influence of bedrock on it (in Japanese, with English summary.) J. Jpn. For. Soc. 67: 311321. Ooyen, V. & Nienhuis, B. 1992. Improving the Convergence of the Backpropagation Algorithm. Neural Networks. 5: 465471.
Parasuraman, K. Elshorbagy, A. & Cheng Si, B. 2006. Estimating Saturated Hydraulic Conductivity In Spatially Variable Fields Using Neural Network Ensembles. Soil Sci. Soc. Am. J. 70: 18511859. Pinder, G.F. & Jones, J.F. 1969. Determination of the ground-water component of peak discharge from the chemistry of total runoff. Water Resour. Res. 5: 438445. Ripley, B.D. 1996. Pattern Recognition and Neural Networks, New York. Cambridge University Press. Russo, D. 1988. Determining soil hydraulic properties by parameter estimation: on the selection of a model for the hydraulic properties. Water Resources Research. 24(3): 453459. Shinomiya, Y. Kobiyama, M. & Kubota, J. 1998. Influences of Soil Pore Connection Properties and Soil Pore Distribution Properties on the Vertical Variation of Unsaturated Hydraulic Properties of Forest Slopes. J. Jpn. For. Soc. 80: 105111. Sklash, M.G. & Farvolden, R.N. 1979. The role of groundwater in storm runoff. J. Hydrol. 43: 4565. van Genuchten, M.Th. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44: 615628.
893
A fundamental interpretation of the critical state strength of unsaturated soils

E.J. Murray
Murray Rix Geotechnical, Earl Shilton, Leicestershire, UK
V. Sivakumar
Queens University Belfast, Northern Ireland, UK
ABSTRACT: Any two of the three stress state variables (p ua), (p uw) and s (ua uw) may be used in describing the stress state in unsaturated soils. An equation for the deviator stress q at the critical state must reflect the duality of the stress regime implicit in this statement. In soils containing a significant percentage of fine particles, the duality arises from the bi-modal structure comprising aggregates of particles and water surrounded by air-filled macro-voids. A relationship is presented for q that links the dual stress regime to the specific volume and specific water volume w. Data on kaolin, a lateritic gravel, a residual soil and a rock powder are used to validate the relationship. Consistent trends are indicated for all materials examined indicating that the strength equation provides a good basis for understanding and predicting the critical state strength of a wide class of unsaturated soils. 1 INTRODUCTION packets surrounded by air voids. A relationship is presented for q that incorporates the coupling stress pc . The strength of unsaturated soils defined in terms of q is thus not independent of phase volumes as is usually considered to be the case in saturated soils. 2 ANALYSIS OF TEST DATA
The paper concentrates on the critical state strength of unsaturated fine-grained soils or soils containing a percentage of fines where aggregation of particles may be inferred to occur. A general equation for the deviator stress will be established and validated by comparison with published experimental data. Equation (1) defines the stress regime in unsaturated soils based on fundamental thermodynamic considerations (Murray, 2002; Murray et al., 2002 and Murray and Sivakumar, 2010). p c p ua s where, pc is the coupling stress linking the independent stress state variables p (p u a) is the net stress s (ua uw) is the matric suction ua and uw are the pore-air and pore-water pressures respectively w/ (Vw Vs)/V (ratio of specific water volume to specific volume) is the total volume of the aggregate packets per unit volume of soil An equation for the deviator stress q must reflect the duality of the stress regime represented by Eq. (1). The duality of the stress regime is a consequence of the three interacting phases which results in a bi-modal structure comprising aggregated vw v ps w v v (1)
Kaolin (Sivakumar, 1993; Wheeler and Sivakumar, 1995, 2000; Sivakumar, 2005 and Sivakumar et al., 2009a)The results of shearing tests in the triaxial cell on lightly and heavily statically compressed specimens and lightly dynamically compacted specimens of speswhite kaolin were reported by Sivakumar (1993) and Wheeler and Sivakumar (1995, 2000). The specimens were prepared at different water contents and were subsequently subjected to isotropic consolidation in the triaxial cell, prior to shearing to critical state along various stress paths. The shearing stages were carried out at constant suctions of 100, 200 or 300 kPa using the axis translation technique. These tests will be referred to as 1d-cs tests, i.e., one-dimensionally compressed, constant suction shearing tests. In Fig. 1, the results for q and pc for kaolin have been normalised with respect to s. The line may be represented by the following equation: q p M a c 1 s s (2)
895
3.0 2.5 2.0 Ma = 0.86
1.0 0.9 0.8 0.7 Ma = 0.86
Ma and Mb
q/s
1 1.5 1.0 0.5 0 1 1.5 2 2.5
0.6 0.5 0.4 0.3 0.2 1 Ma
Mb
= 0.6
= 0.6
3
' pc s
0.1
3.5
0 0.7
0.75
0.8
0.85
0.9
0.95
vw/v
Figure 2. Stress ratios Ma and Mb plotted against w / from 1d-cs tests results of Sivakumar (1993) and Wheeler and Sivakumar (1995, 2000) for unsaturated kaolin.
Figure 1. q/s against pc /s at critical state from 1d-cs test results of Sivakumar (1993) and Wheeler and Sivakumar (1995, 2000) for unsaturated kaolin.
where, Ma is the slope of the critical state line based on the normalized axes and is the intercept on the q/s axis at pc /s Substituting for pc from Eq. (1) gives, q Ma p Mbs v where, M b M a w 1 v (3) (4)
While Eq. (3) is in a form similar to that suggested by Toll (1990), the interpretation and values of the total stress ratio parameter Ma and suction stress ratio parameter Mb for critical state conditions is fundamentally different. The value of Ma from Fig. 1 is constant at 0.86. This is slightly greater than M 0.82 for a saturated soil (s 0). This difference, though small, is consistent with a difference in soil fabric in unsaturated and saturated soils. Of note in Fig. 1 is that the intercept on the q/s axis is given by 0.6. This is also the intercept value for other materials examined and tested to critical state. Taking 0.6, Fig. 2 shows values of the suction stress ratio Mb from Eq. (5) plotted against w/ based on Eq. (4). Mb decreases linearly with decreasing w/ at an inclination Ma. It decreases as the degree of saturation decreases, that is, with increasing values of suction. This is consistent with an increasing proportion of the shearing taking place between (not through) the aggregate packets as the soil becomes drier (Leroueil, 1997; Murray and Sivakumar, 2010). As saturation is approached, Fig. 2 shows Mb approaches 0.6. Most published data on unsaturated soils are on initially anisotropically compacted or compressed
soil specimens. However, data on truly isotropically prepared specimens of kaolin were presented by Sivakumar (2005) and Sivakumar et al., (2009a). The authors reported results of tests on specimens where the suction was maintained constant by the axis-translation technique during shearing (iso-cs tests), and the results of constant water mass tests where the suction was allowed to vary and recorded by thermocouple psychrometer (iso-cwm tests). The critical state results are presented in Fig. 3 as a normalized plot of q/s against pc /s, along with some additional results from constant water mass tests on specimens initially one-dimensionally compressed in layers then subject to isotropic compression prior to shearing (1d-cwm tests). The range of suctions in the suction controlled tests was up to 300 kPa and in the constant water mass tests was between 400 and 1000 kPa. The data for the iso-cs, iso-cwm and 1d-cwm tests in Fig. 3 follow the same general trend, within reasonable experimental differences, over a wide range of suctions. The value of Ma 1.00 is notably greater than the value of Ma 0.86 obtained for kaolin based on the 1d-cs test results of Fig. 1. The value of M for saturated kaolin (s 0) was close to that for the results of Sivakumar (1993) and Wheeler and Sivakumar (1995) (1d-cs tests) and there was no material characteristic difference to account for the greater value of Ma. For unsaturated specimens, the trend line of Fig. 3 has been drawn through a value of 0.6 as for Fig. 1, though points close to pc 1 appear to lie a little below this point. This is attributable to the development of open fissures influencing the data. Figure 4 presents the values of Mb, based on 0.6, but omitting the results considered
896
3.5 3.0 2.5
iso-cwm test 1d-cwm test iso-cs test Ma = 1.0
20
15
Ma =1.77 1
q/s
2.0 1.5 1.0 0.5 0 0 0.5 1
q/s
10
= 0.6
1.5 2
' pc s
= 0.6
0 1 3 5 7
' pc s
11
13
2.5
3.5
Figure 3. q/s against pc /s at critical state from iso-cs, isocwm and 1d-cwm test results of Sivakumar (2005) and Sivakumar et al., (2009a) for unsaturated kaolin.
Figure 5. q/s against pc /s at critical state from 1d-cwm test results of Toll (1990) for unsaturated Kiunyu gravel.
Suction stress ratios Ma and Mb
2.0
Ma =1.77
1.0
Ma= 1.0 iso-cs test
1.5
Stress ratios Ma and Mb
0.8
iso-cwm test 1d-cwm tests
1.0
0.6
Mb
0.5 Mb 0 0.6 0.7 0.8 0.9 1
= 0.6
Ma 1
0.4 1 0.2
Ma
= 0.6
0
0.7 0.75 0.8
vw/v
vw/v
0.85 0.9 0.95 1
Figure 4. Suction stress ratios Ma and Mb plotted against w/ from iso-cs, iso-cwm and 1d-cwm test results of Sivakumar (2005) and Sivakumar et al., (2009a) for unsaturated kaolin.
Figure 6. Stress ratios Ma and Mb plotted against vw/v from 1d-cwm test results of Toll (1990) for unsaturated Kiunyu gravel.
significantly influenced by fissuring. The data corresponds to that of Fig. 3, and again indicates a consistent trend with results close to a line drawn at an inclination of Ma in accordance with Eq. (4). Kiunyu gravel (Toll, 1990)The specimens were prepared by either static or drop hammer compaction in layers. The triaxial tests were carried out with the water contents and pore-air pressures kept constant. Shearing was continued until the specimens were at or reasonably close to critical state with recorded suctions in the range 0537 kPa. The original test data are re-evaluated and presented in Figs. 5 and 6. The only results not included are those where s was zero or very close to zero. The tests are designated 1d-cwm tests.
As shown in Fig. 5, Ma is again constant, and it is larger than the value of M for saturated material (s 0). For the lateritic gravel Ma 1.77 and M 1.62, and the intercept 0.6 is the same as for kaolin in Figs. 1 and 3. Taking 0.6, Eq. (4) can be used to determine values of Mb. These are plotted against vw/v in Fig. 6. Jurong soil (Toll and Ong, 2003)A series of constant water mass triaxial tests was carried out on unsaturated specimens with the matric suction controlled by the axis translation technique. The matric suction varied between 170 and 400 kPa. These tests are designated 1d-cwm tests. Figures 7 and 8 indicate a consistent interpretation of the data for unsaturated Jurong soil with Ma again constant and Mb decreasing with decreasing vw/v (increasing suction) at an inclination Ma. Ma is approximately 1.27, which is again greater than
897
12
8
10 8
6
1
q/s
6 4
q/s
Ma = 1.27
Ma = 1.63 1
2
2 0 1 2 3 4 5
' c
= 0.6
0
6 7 8 9 10 11
= 0.6
1 2 3
' pc s
p s
Figure 7. q/s against pc /s at critical state from 1d-cwm test results of Toll and Ong (2003) for unsatued Jurong.
Figure 9. q/s against pc /s at critical state from 1d-cs test results of Matouk et al., (1995) for unsaturated TroisRivires silt.
1.6
1.8
1.2 1.0 0.8 0.6 0.4 0.2 0 0.82 Mb 1
1.4
Ma =1.27
Ma = 1.63
1.6 1.4 1.2 1.0 0.8 0.6 0.4 0.2 0 0.6 0.7 0.8 0.9 1 Mb 1 Ma
0.6
Ma
0.86
0.9
0.94
0.98
vw/v
Figure 8. Stress ratios Ma and Mb plotted against vw/v from 1d-cwm test results of Toll and Ong (2003) for unsaturated Jurong soil.
vw/v
Figure 10. Suction stress ratios Ma and Mb against vw/v from 1d-cs test results of Matouk et al., (1995) for unsaturated Trois-Rivires silt.
M 1.23 for the saturated soil as quoted by the authors. Again the intercept on the q/s axis at pc /s appears close to 0.6. Trois-Rivires silt (Matouk et al., 1995) Specimens were formed by initially lightly tamping the material into a mould. Triaxial shearing tests were reported for saturated and unsaturated soil specimens that were compressed quasi-isotropically in the triaxial cell prior to shearing. Consolidated undrained shearing tests were performed on the saturated specimens and consolidated drained tests performed on unsaturated specimens. The tests on unsaturated specimens employed the axis translation technique with suctions held constant during shearing at values between 80 and 600 kPa. Failure was defined as the critical state. In is unclear whether to designate the tests on unsaturated Trois-Rivires silt as 1d-cs tests or iso-cs tests.
Normalising q and pc with respect to s, the critical state deviator stress results of Fig. 9 lie close to a unique line for all suctions. A line through the data at Ma 1.63 with an intercept of 0.6 on the q/s axis at pc /s reasonably follows the data points. The results indicate M Ma 1.63. This differs from the results for kaolin, Kiunyu gravel and Jurong soil where Ma M. However, the value of 0.6 complies with the other results. Using Eq. (4) and taking 0.6, Fig. 10 shows values of the suction stress ratio Mb plotted against vw/v. 3 INFLUENCE OF SOIL FABRIC AND PARTICLE AGGREGATION
All materials examined indicate a decrease in Mb with decreasing vw/v (decreasing water content and
898
= 0.6
increasing s) at a slope with a value close to Ma. The decrease of Mb is consistent with the water phase being drawn back into the finer pore spaces and the suction having less influence on the shear resistance. When Mb 0, the suction plays no part in the shear resistance other than maintaining the aggregate structure. The results for Ma and for kaolin, Kuinyu gravel and Jurong soil differ slightly from those for Trois-Rivires silt but it is possible to write that Ma M for all the materials. This is indicative of aggregated packets of particles acting as larger particles. Once created, experimental evidence indicates that the aggregates give rise to a relatively constant Ma that appears unaffected by change in suction. In an unsaturated fine-grained soil, it is not surprising that when aggregate packets form Ma can exceed M if the suction is sufficiently high to maintain the aggregations under the application of shear stresses (Toll, 1990). The equality M Ma for the Trois-Rivires silt suggests that the component of shearing resistance attributable to the net stress is unaffected by particle aggregation and the relatively low percentage of fines is considered to have resulted in the coarser soil fraction controlling the influence due to the net stress. However, for the Trois-Rivires silt as well as the other materials, 0.6 Ma indicates a reduction in shearing resistance due to the presence of aggregates attributable to preferential shearing between the aggregates (Murray and Sivakumar, 2010). Different preparation procedures produce different soil fabrics. Wheeler and Sivakumar (2000) concluded, from tests on kaolin, that differences in fabric caused by different one-dimensional compaction procedures could be erased by shearing to the critical state and any remaining influence of soil fabric was likely to be related only to the suction and the initial compaction water content. This does not tell the whole story. The constant suction critical state strength results 1d-cs of Fig. 1 for kaolin appear to confirm that the initial structure had no influence on the critical state strength. The results for kaolin in Fig. 3 also appear to lie on a unique critical state strength line, but this line indicates greater deviator stress than Fig. 1. As there was no measurable property change in the kaolin, it is necessary to consider carefully the differences in the methods of specimen preparation and the test procedures in order to appreciate the reasons for the difference. The shearing tests 1d-cs on one-dimensionally prepared specimens, and the shearing tests iso-cs on truly isotropically prepared specimens, were both carried out under constant suction conditions. It is argued by Murray and Sivakumar (2010) that this is attributable to the aggregation of the fine particles and their behaviour under
specimen preparation and shearing. The initial one-dimensional compression in the 1d-cs tests will have resulted in a heterogeneous soil fabric and distortion and possible breakage of aggregates arguably at odds with the specimens iso-cs prepared by isotropic compression. Subsequent shearing at constant suction entailed uptake of water, preferentially absorbed by the aggregates. The aggregates will have expanded and deformed in a manner influenced by the imposed stresses, with expansion of the individual aggregates into the inter-aggregate void spaces, reducing the overall potential expansion of a specimen (Thom et al., 2007 and Sivakumar et al., 2009b). It is these influences on aggregate shape and size that are considered the reason for the differences between the critical state results. Murray and Sivakumar (2010), have provided further evidence of the influence of aggregate structure on the shearing resistance and volumetric behaviour of unsaturated kaolin. They concluded that the critical state for unsaturated kaolin is nonunique. This contrasts with the evidence for saturated soils. The critical state in unsaturated soils appears to be influenced by the size and rigidity of the aggregates, which is influenced by the test specimen preparation (whether one-dimensional or truly isotropic) and the test procedure (whether constant suction or constant water mass). Other materials may exhibit further differences in shear behaviour and a wider range of critical state strength data. A further important influence of soil structure is the transition between unsaturated and saturated soils. Toll (1990) and Toll and Ong (2003) along with other researchers have assumed a smooth transition from unsaturated to saturated behaviour in analyzing experimental shear strength data. However, the foregoing analysis does not require this as an assumption, and direct comparison with the experimental data shows consistent trends with a discontinuity between unsaturated and saturated behaviour, Ma M, where aggregation of particles significantly influences the soil strength. Wheeler and Sivakumar (1995) and Wheeler and Sivakumar (2000) suggested the possibility of a discontinuity in shear strength for kaolin. No obvious breakdown of the aggregate structure during shearing is indicated for kaolin even for suctions below 100 kPa. It might be reasonably argued that if the suction in an unsaturated soil is sufficient to maintain an aggregate structure when the soil is at critical state, then it is likely to hold everywhere within a soil mass and at all lower stress levels. Only at relatively low suctions when the integrity of the aggregate packets can no longer be maintained under shearing, will they break down. In this case, all of the aggregates in the zones of shearing can be expected to break down. If this is so, it may be
899
unreasonable to expect a smooth transition from unsaturated to saturated behaviour. The lack of a smooth transition from unsaturated to saturated behaviour adds some complexity to strength analysis but as shown by Murray and Sivakumar (2004, 2010), the analysis which leads to this conclusion justifies a reduction in the number of variables and parameters necessary to describe the critical state of unsaturated soils. 4 CONCLUSIONS
Using pc and published experimental triaxial shearing data for the critical state strength of kaolin, a lateritic gravel (Kiunyu gravel), a residual soil (Jurong soil) and rock powder (Trois-Rivires silt), an equation for the critical state deviator stress q has been developed and analysed. This equation is in a similar form to that proposed by Toll (1990) from consideration of the independent stress state variables and includes the net stress ratio Ma and suction stress ratio Mb that give the components of deviator stress associated with the stress state variables p and s respectively. However, the interpretation of the stress ratios, which reflect the duality of the stress regime considered present in unsaturated soils, differs significantly from that of Toll (1990). The significance of Ma and Mb to the shear strength of unsaturated soils is described in terms of the soil fabric and preferential shearing within and between particles and aggregates. Consistent trends are indicated for all materials analysed indicating the equation for deviator stress as a reliable basis for the understanding and prediction of the critical state strength of a wide class of unsaturated soils. REFERENCES
Leroueil, S. (1997). Critical state soil mechanics and the behaviour of real soils. Recent Developments in Soil and Pavement Mechanics, Almeda (ed), Balkema, Rotterdam, 4178. Matouk, A., Leroueil, S. and La Rochelle, P. (1995). Yielding and critical state of a collapsible unsaturated silty soil. Geotechnique, 45(3), 465477.
Murray, E.J. (2002). An equation of state for unsaturated soils. Canadian Geotechnical Journal, 39(1), 125140. Murray E.J. and Sivakumar, V. (2004). Discussion on Critical state parameters for an unsaturated residual sandy soil, Toll, D.G. and Ong. Geotechnique, 54(1), 6971. Murray, E.J. and Sivakumar, V. (2010). Unsaturated Soils (A Fundamental Approach to the Interpretation of Soil Behaviour). To be published by Wiley-Blackwell. Murray, E.J., Sivakumar, V. and Tan, W.C. (2002). Use of a coupling stress for unsaturated soils. Proc 3rd Int. Conf. for Unsaturated Soils, UNSAT 2002, Recife (Brazil), Ed. Juca, J.F.T., deCampos, T.M.P. and Marinho, F.A.M., 1, 121124. Sivakumar, R. (2005). Effects of anisotropy on the behaviour of unsaturated compacted clay. Thesis submitted to the Queens University of Belfast for the degree of Doctor of Philosophy. Sivakumar, V. (1993). A critical state framework for unsaturated soils. Thesis submitted to the University of Sheffield, UK for the degree of Doctor of Philosophy. Sivakumar, V., Sivakumar, R., Brown, J.L. and MacKinnon, P. (2009a). Mechanical behaviour of unsaturated kaolin with isotropic and anisotropic stress history. Part 2: Performance under shear loading. Accepted for publication in Geotechnique. Sivakumar, V., Sivakumar, R., Murray, E.J., MacKinnon, P. and Brown J.L. (2009b). Mechanical behaviour of unsaturated kaolin with isotropic and anisotropic stress history. Part 1: Wetting and compression behaviour. Accepted for publication in Geotechnique. Toll, D.G. (1990). A framework for unsaturated soil behaviour. Geotechnique, 40(1), 3144. Toll, D.G. and Ong, B.H. (2003). Critical state parameters for an unsaturated residual sandy soil. Geotechnique, 53(1), 93103. Thom, R., Sivakumar, R., Sivakumar, V., Murray, E.J. and Mackinnon, P. (2007). Pore size distribution of unsaturated compacted kaolin: the initial states and final states following saturation. Geotechnique, 57(5), 469474. Wheeler, S.J. and Sivakumar, V. (1995). An elasto-plastic critical state framework for unsaturated soils. Geotechnique, 45(1), 3553. Wheeler, S.J. and Sivakumar, V. (2000). Influence of compaction procedure on the mechanical behaviour of an unsaturated compacted clay. Part 2: Shearing and constitutive modelling. Geotechnique, 50(4), 369376.
900
Determination of the effective stress parameter: A pore network study

E. Nikooee, G. Habibagahi & A. Ghahramani
Department of Civil and Environmental Engineering, Engineering School, Shiraz University, Shiraz, Iron
ABSTRACT: In Bishops equation for the effective stress, the contribution of suction to the effective stress is controlled through a parameter, , which is the so-called effective stress parameter. Unsaturated soils show complex hydro-mechanical behavior and the interaction of different phases controlling the behavior of unsaturated soils is a great concern in the determination of the effective stress parameter. This parameter is also affected by other factors such as stress history, soil placement, and soil- water hysteresis. Therefore, analytical determination of the effective stress parameter is a difficult task. In the present study, a pore network model is developed to predict the variation of the effective stress parameter at different saturation and suction levels. In the proposed method, the pore size distribution of the soil is estimated from Soil Water Retention Curve (SWRC) and the effective stress parameter, , is determined directly from pore network simulation. Moreover, model simulations are employed to shed light into the soil-water hysteresis phenomena and its subsequent effects on the effective stress parameter. 1 INTRODUCTION resorted to different approaches to overcome this complexity and to determine the effective stress parameter. Some researchers have chosen artificial intelligence and have employed artificial neural networks to obtain the effective stress parameter (Kayadelen 2007). In a number of studies the effective stress parameter has been simply considered to be equal to the degree of soil saturation. This has been challenged by others from different perspectives. For instance, Coussy et al. (2010) have recently pointed out that this equality is valid provided that phases are not disconnected and certain conditions for the deformation of pores must be satisfied. Some researchers have stated that this parameter is strongly affected by soil structure and it should be related to the soil structure and expressed the dependence through other parameters such as air entry value (Khalili & Khabbaz 1998). The effect of hydraulic hysteresis on effective stress parameter is another important issue. To tackle this issue, some researchers have implemented air entry value and air expulsion values associated, respectively, with main drying and wetting paths (Khalili et al. 2008). In fact, a more realistic simulation can be obtained by utilizing a pore network model to capture the influence of hysteresis phenomena on the effective stress parameter. To construct the pore network, the appropriate pore size distribution is obtained using an iterative technique. In this technique the pore size distribution was adjusted such that to minimize the error between the actual SWRC and the SWRC
There has been a long lasting debate throughout the literature of unsaturated soil mechanics on the appropriate set of stress state parameters for constitutive modeling of unsaturated soils. More recently, the notion of Bishops generalized effective stress has attracted a considerable amount of attention among researchers, (Nuth & Laloui 2008). Although there are still challenging questions on the role of suction on hydro-mechanical behavior of unsaturated soils (Laloui & Nuth 2009, Habibagahi & Ajdari 2010, Lu 2008), one, indeed, demanding task is to obtain clearly and objectively the effective stress parameter itself. Effective stress parameter specifies the contribution of the suction to the effective stress as defined by Equation 1, (Bishop 1959). Stress history, soil structure, placement conditions, hydraulic hysteresis are among factors affecting the effective stress parameter. ( ua) (ua uw) (1)
In this equation, is total stress, ua and uw are pore air and pore water pressures, respectively, and is the effective stress parameter. The effective stress parameter, , is equal to 1 for saturated soils and therefore, the aforementioned equation is reduced to the well-known effective stress equation for saturated soils and is equal to 0 for soils in their dry state. Due to the complexity inherent in the nature of the effective stress parameter, researchers have
901
predicted by the network. When this goal was achieved, the pore network was assumed to represent the real pore size distribution of the soil. Next, the pore network was explored to get information about configuration of different wetting and nonwetting phases (water and air). Such details are of remarkable help in finding the effective stress parameter when a conceptual model is at hand. In the subsequent sections of this paper, first a brief theoretical background on the pore network modeling is offered. Then, the proposed algorithm and conceptual model are described. Finally, the results are presented and shortcomings originated directly from the modeling assumptions are discussed in details. 2 2.1 PORE NETWORK MODELING Historical and theoretical backgrounds the soil suction, n, and the corresponding volumetric water content, n, the minimum radius of equivalent spherical pore space that would undergo dewatering subjected to the applied suction, n, was determined by Kelvins capillary model expressed by: 2T n (ua uw )n r s n (2)
Figure 1. Evolution of the pore network models (Berkowitz & Ewing).
In petroleum engineering, network models are used to explore the macroscopic properties of porous media on the basis of pore-scale descriptions of the multiphase flow phenomena. Pore network models in literature can be tracked back to the pioneering work of Fatt (1956). Recent pore networks overcome the limitations of early conceptualizations of pore space that considered bundles of capillary tubes of varying radii. Early pore network models were based on the assumption that the network intersections do not have volumes of their own, which was a mere generalization of the bundle of capillary tubes. However, since pore bodies govern the majority of the pore volume, most researchers nowadays use models that consist of pores and throats (e.g., Chatzis & Dullien 1985, Ferrand & Celia 1992, Bryant et al. 1993, Dillard & Blunt 2000). Illustration of different pore space models can be found in Figure 1. In this study a pore network model of the third type, Figure 1(c), was adopted. To simulate two phase flow and the equilibrium conditions as in pressure plate test, a pore network model can be used as follows. At the initial state, the constructed pore network is full of water (wetting phase), representing the saturated condition of the real soil. Then by applying successive suctions, pores become desaturated and the water content of the network for each level of suction is computed (Arvin et al. 2007). 2.2 Proposed algorithm and the conceptual model
In this equation, Ts is water surface tension, rn is radius of curvature which governs dewatering of pore space, n is the corresponding soil suction, ua and uw are pore air and pore water pressures, respectively, as described earlier in Equation 1. To obtain the soil pore size distribution, the number of spheres corresponding to each interval was calculated and then corrected to account for the porosity of the real soil. 2. Pores were randomly placed in a 3D pore network and between two adjacent pores a throat was defined, see Figure 2. An appropriate relation between diameter of the end pores and that of throat may be presented by: dth f(dp1, dp2) cdmin (3)
The following steps describe the proposed algorithm to generate the pore network model: 1. First the soil water retention curve was divided into n intervals. For a certain interval, knowing
In this equation, dth is the diameter of the throat which is a function of the diameters of the neighboring pores (dp1 and dp2). In other words, the throat diameter can be obtained by multiplying the minimum diameter of the adjacent pores by a coefficient. These coefficients were determined subsequently through an optimization procedure to minimize the error between predicted and simulated results of drying curve.
902
3. After arriving at the best pore network, the most suitable pore and throat size distributions, the wetting curve was simulated by decreasing the suction and detecting the pores and throats that were target of that suction. The so-called ink bottle effect governs the wetting phenomenon. It means that the wetting mechanism is controlled by the radius of the pores while drying is controlled by the radius of throats. 4. Finally, by exploring the network and finding the ratio of the total area of pores filled with water, Auw to the total pore area, Asw for each cross section of the pore network, the effective stress parameter was obtained for drying and wetting conditions (see Equation 4, and Figure 3). The average value of the effective stress parameters obtained from different sections of the pore network was considered as the equivalent effective stress parameter for the pore network.
Usually Asw is considered to be equal to the total cross sectional area, A, assuming the solid contacts to be negligible. 3 RESULTS AND DISCUSSION
3.1 Determination of the effective stress parameter from available data in the literature Few experimental investigations on the effect of hydraulic hysteresis on the effective stress parameter is available in literature. In the present study, data from CW and CD triaxial tests and drying and wetting curves reported by Thu et al. (2006) were employed to obtain the effective stress parameter for both drying and wetting cycles. Measured values of b were converted to the effective stress parameter using Equation 5, (Delage 2002 ):
Auw Asw
(4)
tan( b ) tan( )
(5)
As Thu et al. (2006) have pointed out; results of CD triaxial tests can be used to investigate the variation of unsaturated shear strength in a drying path while results obtained from a CW test can provide a situation similar to a natural wetting path. Their goal was to study the nonlinearity in b by expressing an equation for unsaturated shear strength in both drying and wetting conditions. More recently, Guan et al. (2009) have followed the same idea to obtain the shear strength equations under drying and wetting conditions. Here, the main focus is on determination of the effective stress parameter from pore network simulations and comparing the accuracy of results with the experimental data for both drying and wetting branches of SWRC. 3.2 Simulation results and discussion on results
Figure 2. A schematic illustration of a regular pore network model.
Figure 3. 2004).
Conceptual sketch for unsaturated soils, (Xu
Figure 4 shows the experimental and simulated drying and wetting paths obtained using the aforementioned procedure. Figure 5 depicts values of the effective stress parameter as determined by the proposed approach as well as those reported in experimental works. Even though, the network was constructed to simulate the drying branch of SWRC, the network performance to predict the wetting branch of SWRC, as shown in Figure 4, is satisfactory. Hence, the capability of the pore network model to capture the hysteresis phenomenon is justified. A glance at Figure 5 reveals that the prediction of the effective stress parameter for drying path and before residual saturation is of good accuracy. The deviation from experimental data at high suction values may be attributed to the nature of phase configurations at the residual saturations.
903
Moreover, Khalili et al. (2008) have considered a transition region during which the effective stress parameter decreases upon suction reversal process. They have pointed out that suction reversal is due to the change in contact angle between the airwater interface and the solid grains which changes from receding to that of advancing during the suction reversal process, (Khalili et al. 2008). In the proposed approach, based on the observed trend in the experimental data, the change of contact angle was assumed negligible. For this reason, results of the pore network model have not depicted such a transition region. Such a transition region can be considered through different contact angles and the wettability rules defined in the pore network. Therefore, convenient rules for defining required changes in contact angle should be considered based on the soil type and the experimental observations.
Figure 4. network. Modeling hydraulic hysteresis using pore
CONCLUSION
Figure 5. Simulation of the effective stress parameter in drying and wetting paths as compared to the experimental data.
Film effects, presence of pendular rings and local equilibrium of phase pressures can also add to the complexity of the problem for this range of suction values. The results of the simulations for the wetting cycle are far from experimental data. The main reason for this discrepancy in the wetting cycle may be attributed to the local trapping of the fluids and the authors believe that the large deviation of the simulation results from experimental data originates from neglecting this phenomenon in the modeling. Current research work is directed to account for the trapping phenomenon in the wetting cycle.
In this paper, application of the pore network model to investigate the hydraulic hysteresis and its effects on the variation of the effective stress parameter was examined. It was shown that by employing the available information on drying branch of SWRC, a pore network model can be generated to predict the wetting branch with a reasonable accuracy. However, it may depend on the soil type and configuration of the wetting (water) and/or non-wetting (air) phases inside the soil medium. Therefore, future studies are required to investigate the efficiency and applicability of pore network models in unsaturated soils mechanics as predictive models. The variation of the effective stress parameter in drying and wetting cycles was simulated using the conceptual pore network. Trapping and film effects were considered as the main shortcomings of the model. For the drying path and before residual saturation, assumptions made in the pore network were acceptable. For residual state and wetting path more caution should be taken into account where these assumptions are no longer valid. These deviations may also be attributed to the simplifying assumptions inherent in Equation 4. An averaging approach cannot always lead to true values for the effective stress parameter (Khalili & Khabbaz 1998, Coussy et al. 2010). These researchers have stated that this parameter should be obtained so that the effects of different issues such as those of the soil structure, configuration and interaction of different phases (non-wetting and wetting phases) and also the deformation of pores can be considered throughout the calculation technique. In this regard, Equation 4 seems to be very simplistic. Also, in the proposed pore
904
network modeling, elastic characteristics of pores and throats have not been considered. To achieve better results, pores and throats in the pore network should undergo mechanical deformation at different suction values. Such a network would also be capable of determining the effects of stress level on the calculated results. Further studies are required to differentiate the effects of different assumptions used in this modeling technique. REFERENCES
Arvin, M.R., Veis Karami, M., Ajdari, M. & Habibagahi, G. 2007. A Percolation Approach To Determination of Effective Stress Parameter in Unsaturated Soils. Proc. 3rd Asian Conf. on Unsaturated Soils, Nanjing, China: 163168. Beijing: Science Press. Berkowitz, B. & Ewing, R.P. 1998. Percolation theory and network modeling applications in soil physics. Surveys in Geophysics 19: 2372. Bishop, A.W. 1959. The principle of effective stress. Teknisk Ukeblad 106(39): 859863. Bryant, S.L., King, P.R. & Mellor, D.W. 1993. Network model evaluation of permeability and spatial correlation in a real random sphere packing. Transport in Porous Media 11: 5370. Chatzis, I. & Dullien, F.A.L. 1985. The modeling of mercury porosimetry and the relative permeability of mercury in sandstones using percolation theory. Int. Chem. Eng. 25: 4766. Coussy, O., Pereira, J.-M. & Vaunat, J. 2010. Revisiting the thermodynamics of hardening plasticity for unsaturated soils. Computers and geotecnics 37: 207215. Delage, P. 2002. Experimental unsaturated soil mechanics: State-of-art-report. Proc. 3rd International conference on unsaturated soils, vol 3, Recife, Brazil: 973996. Balkema. Dillard, L.A. & Blunt, M.J. 2000. Development of a pore network model to study non-aqueous phase liquid dissolution. Water Resources Research 36(2): 439454.
Fatt, I. 1956. The network model of porous media. Petroleum Transactions, AIME 207: 144177. Ferrand, L.A. & Celia, M.A. 1992. The effects of heterogeneity on the drainage capillary pressure: Saturation relations. Water Resources Research 28(3): 859870. Guan, G.S., Rahardjo, H. & Choon L.E. 2009. Shear Strength Equations for Unsaturated Soil under Drying and Wetting. J. of Geotech. Geoenv. Eng., ASCE, In press. Habibagahi, G. & Ajdari, M. 2010. Discussion on Is matric suction a stress variable? By: Lu N., J. Geotech Geoenviron Eng, ASCE, 134(7): 899905. J. of Geotech. Geoenv. Eng., ASCE 136(2): 407407. Kayadelen, C. 2007. Estimation of Effective Stress Parameter of Unsaturated Soils by Using Artificial Neural Network. Int. J. Num. Anal. Meth. Geomech. 32(9): 10871106. Khalili, N. & Khabbaz, M.H. 1998. A Unique Relationship for the Determination of the Shear Strength of Unsaturated Soils. Geotechnique 48: 17. Khalili, N., Habte, M.A. & Zargarbashi, S. 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hysteresis. Computers and Geotechnics 35: 872889. Laloui, L. & Nuth, M. 2009. On the use of the generalized effective stress in the constitutive modeling of unsaturated soils. Computers and Geotechnics 36: 2023. Lu, N. 2008. Is matric suction a stress variable? J. Geotech. Geoenviron. Eng. ASCE 134(7): 899905. Nuth, M. & Laloui, L. 2008. Effective stress concept in unsaturated soils: Clarification and validation of a unified framework. Int. J. Numer. Anal. Meth. Geomech. 32:771801. Thu, T.M., Rahardjo, H. & Leong, E.C. 2006. Effects of Hysteresis on Shear Strength Envelopes from Constant Water Content and Consolidated Drained Triaxial Tests. Proceeding of the Fourth International Conference of Unsaturated Soil, UNSAT 2006, Carefree, Arizona, April 26: 12121222. Xu, Y.F. 2004. Fractal approach to unsaturated shear strength. J. of Geotech. Geoenv. Eng. ASCE 3: 264273.
905
Assessment of volumetric collapse using probabilistic approach

I. Otlvaro Calle
Pontificia Universidad Javeriana, Cali, Colombia
M.P. Cordo-Neto G.M. Medero
University of Braslia, Braslia, DF, Brazil Heriot-Watt University, Edinburgh, Scotland, UK
ABSTRACT: In the last years, considerable advances have been done in modelling unsaturated soils. Several different approaches have been proposed which include conventional elastoplastic models (e.g. Barcelona Basic ModelBBM), and more complex ones, (e.g. coupling hydraulic and mechanical behavior). However few improvements were done in relation to the process of parameter calibration in these models. Moreover, there are some imprecision on the process of sampling, which also add an error in the calibration process. The result of this error (or imprecision) introduced during the process of achieving parameters can be evaluated through probabilistic methods which has been developed during the last years. In this work three probabilistic methods are applied to evaluate the importance of each parameter of Barcelona Basic Model on the volumetric collapse. Every probabilistic method has a specific function, where Monte Carlo Method (MCM) was used to define variability of parameters when the probabilistic distributions are not known. Another method that is presented is the First Order Second Moment (FOSM) which was used to calculate the contribution of each BBM parameter on the shape of yield curve. Finally, Point-Estimate Method was used to evaluate the mean and standard deviation of volumetric collapse. Keywords: Monte Carlo Method, FOSM, Point-estimate Method, unsaturated soils
INTRODUCTION
The common work of geotechnical engineering is made at the ground surface. Only a small portion exceeds the water table. The site engineers prefer to avoid operations below the water table to streamline their operations. This condition leads to the fact that most of the activities are developed in the unsaturated zone or vadose zone. In this zone you have the major damage to light structures such as houses and pavements. These damages are commonly caused by ground movements due to seasonal variations of environmental factors, as the water content, which do not depend on the change of the external loads. Expansion is called when the movements are due to wetting the soil by increasing the volume; otherwise it is called collapse (Dudley, 1970; Jimenez Salas et al., 1973; Maswose, 1985). In recent decades, this kind of behavior has been addressed with unsaturated soils mechanics. But still now the unsaturated soil mechanics seems to be restricted to the research and its applications in everyday geotechnical problems are quite limited, due to the complexity to determine the parameters and the high costs of equipment and tests involved.
The research of the unsaturated soils mechanics concentrated attempts to develop an effective stress expression. The effective stress was defined as excess stress applied in relation to the fluid pressure in the pores, using in the mathematics formulation an empirical factor to consider the suction contribution (Croney et al., 1958; Bishop, 1959; Aitchison, 1961; and others). Bishop (1959) was the first to suggest an expression for effective normal stress in terms of both pore air pressure and pore-water pressure. Matyas & Radhakrishna (1968) suggested to treat independently changes in the stress state and pore pressure. Fredlund (1979) proposeded an incremental elastic formulation, assuming the nonlinear character of the moduli, considering two independent sets of stress variables: the net stress defined as the excess of total stress over air pressure and the suction. Alonso et al., (1990) described the stressstrain behavior of partially saturated soils by extending the modified CamClay model using Fredlund (1979) independent stress variables mentioned above. A large number of constitutive models for partially saturated soils derivate of Barcelona Basic Model (BBM) introduced by Alonso et al., (1990).
907
According to Alonso et al., (1990), the stress strain behavior of partially saturated soils has the following fundamental characteristics: suction contributes to stiffening the soil against external loading; the soil can expand or collapse depending on the stress level; the reduction in suction (wetting) in soils can induce an irrecoverable volumetric compression (collapse) in open structured soils, when the soil is subjected to a high stress level; the volumetric response depends not only on the initial and final stress and suction values but on the particular path followed from the initial to the final state; when the soil is gradually loaded, it may have plastic deformation, if the yield stress is overcome. The value of yield stress increases with suction. Using four state variables: p mean net stress, q deviatoric stress, s suction, and e v specific volume is defined yield locus, which can be represented in the pqs diagram (Fig. 1). The yield locus is defined by the following equations: fLC q2 M2(p0 p)(p ps) 0 fSI s s0 0 (1) (2)
Figure 2. Relationship between yield stresses p0 and p0* in compression curves for saturated and unsaturated soil.
where M slope of the critical state line; ps increase in cohesion; p0 yield stress any suction; and s0 the maximum past suction. The LC surface can be written on the mean net stress and suction space as:
* P0 P0 c Pc P [ ( 0 ) x ] /[ ( s ) x ]
Figure 3. Collapse conceptual model as work to move the LC surface.
(3)
where p0* preconsolidation stress for saturated conditions; pc reference stress; elastic stiffness parameter for changes in net mean stress; (0) stiffness parameter for changes in net mean stress for virgin saturated states; and (s) stiffness parameter for changes in net mean stress for virgin
unsaturated states of the soil. The Figure 2 show physical interpretation of the variables. The Equation (4) provides an expression describing the increase in soil stiffness with suction: (s) (0)[(1 r) exp(s) r] (4) where r constant related to the maximum stiffness of the soil; and parameter which controls the rate of increase of soil stiffness with suction. Conceptually, the collapse defined in Figure 3 is the energy required to move the LC surface to its new position, which is the plastic deformation, and its calculated as: dvp p 2
(0 ) k dp0* v p0*
(5)
PROBABILISTIC METHODS
Figure 1.
BBM yield locus in pqs plane.
The uncertainty is always present in all geotechnical analysis. Due to natural variability of the soil or
908
by the poor knowledge of soil behavior, commonly reflected in the constitutive models for which they represent. According to Christian (2004), geotechnical engineers have developed several strategies for dealing with uncertainty, they include: ignoring it, being conservative, using the observational method and quantifying uncertainty. At present we have different probabilistic methods to evaluate the effects of variability in performance. Three of them have become popular among geotechnical engineers, the Monte Carlo Method (MCM), the method of First Order Second Moment (FOSM) and the Point Estimate Method (PEM). 2.1 Monte Carlo Method (MCM)
where E[] mean; and 2[] variance. Christian et al., (1992) propose to calculate the derivative of eq. (8) with the finite difference approximation as follows: g F 1 F 2 F xi xi1 xi2 xi (9)
where F1 its the function calculated for value xi1; and F2 its the function calculated for value xi2. The FOSM approximation only provides an estimative of the mean and variance. 2.3 Point Estimate Method (PEM) Rosenblueth (1975, 1981) proposed the PEM method that provides an approximation for the low-order moments for the dependent variable starting from the low-order moments of the independent variable x. In the PEM method, all possible combinations are taken into account for two-point estimate for each independent variable. If the pdf are considered symmetric, the points estimated are separated once standard deviation below and above average, the mean and variance can be calculated as: E[F ] 1 2n g[ x j ] 2 n j 1 1 2n ( g[ x j ] E [ F ])2 2 n j 1 (10) (11)
The simulation by Monte Carlo Methods (MCM) can solve problems by generating suitable random numbers (or pseudo-random numbers) and assessing the dependent variable for a large number of possibilities. The MCM involves the definition of the variables that generate uncertainty and Probabilistic Distribution Function (PDF); determination of the value of the function using variable values randomly obtained considering the pdf; and repeating this procedure until a sufficient number of outputs to build the pdf of the function. The number of required Monte Carlo trials is dependent on the desired level of confidence in the solution as well as the number of variables being considered (Harr, 1987), and can be estimated from: d2 NMCM 2 4(1 )
n
2 [F ]
(6)
where NMCM number of Monte Carlo trials, d the standard normal deviate corresponding to the level of confidence, the desired level of confidence (0 to 100%) expressed in decimal form; and n number of variables. If the problem has n variables, the number of trials increases geometrically, according to the power n. 2.2 First-Order Second-Moment (FOSM) analysis
where g[xj] obtained with the function we defined the dependence of g[xj ] and the independent variables, alternately substituting the values of these variables by E[xi] i, j 1, 2, ..., n, obtaining the 2n values of g[xj]. 3 PROBABILISTIC METHODOLOGY
The FOSM method is an approximate approach based on Taylors series expansion of function g(x1, x2, x3, , xn), in which the g function of a number of input variables xn that represents uncertainty parameters. For uncorrelated input variables, the mean and variance of the function are given by eqs. (7) and (8), respectively: E [F ] g(E [x1], E [x2], E [x3], ..., E [xn])
n g 2 [F ] 2 [ xi ] x i i 1 2
For the collapse evaluated were adopted as benchmarks values of BMM reported by Alonso et al. (1990), values and parameters subjected to variation are presented in Table 1. The values of the coefficient of variation COV were obtained from Duncan (2000), by analogy of
Table 1. Parameters and variables subject to variation. With uncertainty Yes Yes Yes Yes Yes No Mean 0.2 0.02 0.75 12.5 0.2 0.1 COV 24% 25% ? ? 20%
Parameter (0) r [1/MPa] p0*[MPa] pc [MPa]
(7) (8)
909
these parameters with other mechanical properties are already well studied. The general probabilistic analysis of the adopted methodology included: i. Selection of variables that induce uncertainty; ii. Estimation of the COVs for the variables that induce uncertainty. For typical soil properties estimated COV based on published values, assuming normal probability distribution function and for and r, which COV is not reported, it was done a previous study to determine accepted values; iii. Estimation variability for the increase in soil stiffness with suction (s), with MCM method used a spreadsheet developed in @Risk (Palisade Corporation 2008) software. iv. Definition of the COV for all parameters used FOSM for obtained the variability and the influence of parameters on the LC surface (Eq. 3). For comparison, the variability was obtained with MCM method, for the Eq. 3. v. Exclusion of variables that have minor influence on LC surface, discounting that directly control the collapse as (0). vi. Finally we obtained the influence of the stress path and wetting on the calculation of the collapse, including low order statistical moments of by PEM method. The mean and variance of the triangular distribution function was calculated with the following equations: E [] min mp max 3 (12)
4.1 MCM initial results For (s) variability using the MCM method, were made 5000 trials, since according to Eq. (6) 4575 trial are required for a confidence level of 90%, Figure 4 presents the results. The COV varies from 0% to 20% increasing non-linearly with suction. For the 0.2 MPa of suction obtained mean equal at 0.144 and 18% for COV. These results were recalculated using a Pert pdf, getting minor differences. 4.2 FOSM results for LC surface Having defined the statistical moments of all the parameters involved, an analysis was performed with the FOSM method. Initially it was verified the approximation of the derivative of Eq. (8) varying xi of the Eq. (9) from 0.001% to 10% of the mean. Figure 5 presents the results obtained. The approximation of the FOSM derivative of the variance is stable for xi values less than 1%. The results in terms of mean and variance obtained by FOSM method for a 0.2 MPa of the suction were 0.254 MPa and 0.007 MPa2, obtaining a 33% in COV. With the terms comprising the sum
2[] min2 mp2 max2 (min)(mp) (mp)(max) (max)(min)
(13)
where [ ] mathematical expectation or mean; 2[] variance; min a minimum; max maximum value; y mp most likely value. There are two main disadvantages of a Triangular distribution. First, all parameters used have the same weight in the mean and variance; see Eq. (12) and (13). Second, when the parameters result in a skewed distribution, then there may be an over-emphasis of the outcomes in the direction of the skew. 4 PROBABILISTIC ANALYSES RESULTS
Figure 4.
Simulation results with MCM for (s).
According to the physical analysis of the parameters which were of unknown variability, it was adopted 3 MPa1 for minimum of parameter , and 20.5 MPa1 maximum conceivable, and for parameter r the values adopted were 0.3 and 1.0 respectively.
Figure 5.
xi effect in the variance calculation.
910
of Eq. (8) it was calculated the contribution of each variable on the LC surface (see Table 2). 4.3 MCM results for LC surface
According to Eq. (6), taking into account 5 variables and a reliability of 90%, it was necessary 1.5 109 trials. Due to the high trials number, the authors assessed the variability in the statistical moments for different trials in LC surface for constant suction of the 0.2 MPa, see Figure 6. At least 5000 trials were required to stabilize the statistical moments. The LC surface variability by the MCM method for 5000 trials is presented in Figure 8. It was obtained that for the reference suction of 0.2 MPa the average 0.282 MPa in LC surface, 0.01 MPa2 for the variance, and 35% COV. Comparing the results obtained for the LC surface by the methods of FOSM and MCM, the authors observed that the means and variances of MCM were higher. This effect may be due to the skewed in the distribution of some of the independent variables, which is ignored by the method FOSM method. 4.4 Collapse assessment with PEM method
Figure 7. surface.
Simulation results with MCM for LC
We evaluated two paths, with the same initial stress, 0.15 MPa for mean net stress and 0.2 MPa suction.
Table 2. Contribution of the variability of LC surface, for suction 0.2 MPa. Varivel (0) p0* [MPa] [1/MPa] r xi 0.00020 0.00002 0.00020 0.01250 0.00075 F, Eq. (9) 0.00001 0.00001 0.00034 0.00002 0.00024 Eq. (8) 0.000005 0.000005 0.004663 0.000023 0.001894 Contrib. 0.1% 0.1% 70.6% 0.4% 28.9% 100.0% Figure 8. Typical specific volumestress path obtained with PEM method.
f2 0.00656
Figure 6. Number trial effect in the statistical moments for LC surface.
The first path in the soil was loaded at 0.35 MPa with constant suction and then wetted, reducing the suction to 0 MPa. The second path in the soil was loaded at 0.60 MPa with constant suction and then wetted, reducing the suction to 0 MPa. The variables considered for the calculation of the uncertainty in the stress path were p0*, (0) and r; including eight trials for each of the two stress paths. Table 3 shows the collapse index (IC e/1e0) calculated using PEM method with Eqs. (10) and (11) evaluated at eight points. For path 1, loaded to 0.35 MPa and followed by wetting, it was obtained 2.9% of mean and 1.7% of standard deviation for collapse index. In path 2, the values were 4.2% and 1.6% respectively. The paths analyzed showed a COV of 59% and 38%. The greater collapse index variability was presented in the path of lower load level, because some stress paths for estimative-points do not overlap the LC surface, generating only elastic deformations during loading, and the LC surface was mobilized only in the wetting, with a significant reduction in the
911
Table 3. Results from PEM method for collapse index IC with mean net stress at 0.6 MPa. Point p0* 1 2 3 4 5 6 7 8 0.240 0.240 0.240 0.160 0.240 0.160 0.160 0.160 (0) 0.248 0.248 0.152 0.248 0.152 0.248 0.152 0.152 r 0.828 0.538 0.828 0.828 0.538 0.538 0.828 0.538 g[xj] 0.06321 0.02900 0.03826 0.06316 0.01574 0.05305 0.03826 0.03205 (g[xj] E [F ])2 0.00047 0.00016 0.00001 0.00047 0.00067 0.00013 0.00001 0.00009 0.00200
ico CNPq of Brazil for providing the financial support for this research. REFERENCES
Aitchison G. 1961. Relationships of moisture and effective stress functions in unsaturated soils. Conf. on Pore Pressures and Suctions in Soils. Butterworths, London: 4752. Alonso E.E., Gens A. & Josa A. 1990. A constitutive model for partially saturated soils. Geotechniqu, 40: 405430. Bishop A. 1959. The principle of effective stress. Teknisk Ukeblad, 106: 859863. Christian J.T. 2004. Geotechnical engineering reliability: how well do we know what we are doing? Journal of Geotechnical and Geoenvironmental Engineering, 126(4): 307316. Christian J.T., Ladd C.C. & Baecher G.B. 1992. Reliability and probability in stability analysis. Proc., Stability and Perf. of Slopes and Embankments-H. ASCE, New York, N.Y., Vol. 2: 10711111. Croney D., Coleman J. & Black W. 1958. Studies of the movement and distribution of water in soil in relation to highway design and performance. Washington, D.C.: HRB Spec. Rep. No. 40. Dudley J.H. 1970. Review of collapsing soils. Journal Soil Mechanics and Foundation Engineering, ASCE SM3: 925947. Duncan J.M. 2000. Factors of safety and reliability in geo-technical engineering. Journal of Geotechnical and Geoen-vironmental Engineering, ASCE 130(10): 9851003. Fredlund D. 1979. Appropriate concepts and technology for unsaturated soils. Canadian Goetechnical Journal, 16(1): 121139. Harr M.E. 1987. Reliability-Based Design in Civil Engineering. McGraw-Hill Book Company. Jimenez Salas J.A., Justo J.L., Romana M. & Faraco C. 1973. Colapso de limos y arcillas yesiferas de baja plasticidad en climas ridos y semiridos. 8th International Conference of Soil Mechanics and Foundation Engineering. Moscow, 193199. Maswoswe J. 1985. Stress path for a compacted soil during collapse due to wetting. PhD thesis, Imperial College, London. Matyas E.L. & Radhakrishna H.S. 1968. Volume change characteristics of partially saturated soils. Geotechniqu 18(4): 432448. Palisade Corporation. 1996. @Risk: risk analysis and simulation add-in for Microsoft Excel V.5.0.1. Palisade Corporation, Newfield, N.Y. Rosenblueth E. 1975. Point estimates for probability moments. Proceedings, National Academy of Science 72(10): 38123814. Rosenblueth E. 1981. Two-point estimates in probabilities. Ap-plied Mathematical Modelling Vol. 5(2): 329335.
Sum 0.33274 E[F] 0.042 2[F] 0.002
collapse index. The Figure 8 shows the changes in volume in terms of net mean stress, it is observed as the LC surface modifies the trajectory and the final collapse by wetting for each estimate-point for Table 3. 5 CONCLUSIONS
In this paper three methods were used at different stages to evaluate the soil collapse, using the BBM model. It was observed an advantage of the FOSM method with respect to MCM in the evaluating the LC surface, due to low demand of resources for evaluation and it can obtain the influence of each variable in the variance evaluation. The FOSM and MCM methods presented differences in mean and variance in the LC surface, however produced measures of COV without significant differences. The triangular pdf is an acceptable alternative to evaluate the effect of variability when there is little information, since it only requires the maximum, minimum and most probable value of the variable. The assumption of maintaining constant , in the BBM model, does not affect of the collapse evaluation. The PEM method allowed the identification of the effect of stress path on the evaluation of collapse. The methodology, combining the three most popular probabilistic methods, showed to have a satisfactory performance. ACKNOWLEDGEMENTS The authors would like to thank of the Conselho Nacional de Desenvolvimento Cientifico e Tecnolg-
912
Is the degree of saturation a good candidate for Bishops parameter?

J.-M. Pereira & O. Coussy
Universit Paris-Est, UR Navier, cole des Ponts ParisTech, Marne-la-Valle, France
E.E. Alonso, J. Vaunat & S. Olivella
ABSTRACT: In unsaturated soil mechanics, the quest for an effective stress playing the same role as Terzaghis effective stress does for saturated soils has introduced a long standing debate, dating back to the 1960s. Several contributions have been proposed since the early work of Bishop. It is well recognized to date that a single constitutive stress is not sufficient by itself to catch the main features of the behaviour of unsaturated soils and it is often combined with matric suction. In this paper, focus is given to a largely used formulation for such a constitutive stress, based on the use of an averaged pore pressure. In particular, this paper discusses on thermodynamics bases the validity of the choice of the factor weighting the fluid pressures contribution to the constitutive stress. This factor is usually assumed to be equal to the degree of saturation of water. In this work it is shown that the choice of this natural candidate implies restrictive assumptions on the plastic flow rule. As shown from experimental data obtained from a literature review, this choice may not be pertinent for certain classes of materials, particularly high plasticity clays. 1 INTRODUCTION Later on, this model settled the bases of numerous models, addressing additional features of unsaturated soils behaviour, such as the effects of water content (Wheeler, 1996) or degree of saturation (Jommi and Di Prisco, 1994, Dangla et al., 1997, Lewis and Schrefler, 1998, Gallipoli et al., 2003, Sheng et al., 2004, Pereira et al., 2005) among others. Advances on the last point have reintroduced a strong debate dated back to the 1960s with Bishops proposal for an extended effective stress to unsaturated states (Bishop, 1959) who introduced an averaging parameter , this latter being a function of the water degree of saturation Sr: ua 1 (Sr) s1 (1)
In engineering applications involving mechanics of geomaterials, focus is generally put on the deformation of the solid skeleton so that a great interest is given to the definition of the stresses governing the skeleton deformation. The presence of one or several fluids within the porous space of these materials introduces a complexity in the choice of the stress state variables with respect to non-porous materials. (Terzaghi, 1936) stated that in the case of saturated soils (where a single fluid filling the porous space) a single stress (called effective stress) governs the deformation and strength. Roscoe and his co-workers (Roscoe et al., 1958) applied the mathematical theory of plasticity to saturated soil mechanics introducing the concept of critical state. This work resulted in the wellknown Cam-clay model for saturated soils involving Terzaghis effective stress. An extension of this model to unsaturated soils has been proposed by Alonso et al. (Alonso et al., 1990) within a simple elastoplastic framework. This extension points out the need of two independent state variables to capture the experimentally observed behaviour of unsaturated soils instead of a unique stress variable (effective stress) in the case of saturated states.
where is the total stress tensor, ua is the air pressure and s ua uw the suction. In recent years, the most commonly used assumption for this parameter is to assume Sr. In the first part of this paper, the thermodynamics of plasticity for unsaturated geomaterials is revisited with a particular attention paid to the constitutive stress governing the skeleton deformation. It is shown that Bishops proposal with the common assumption Sr rely on a restrictive assumption concerning the plastic flow rule. In a
913
second part of the paper, experimental evidences of an existing deviation to the Sr assumption are presented. These evidences are interpreted on the basis of microstructural considerations. 2 THERMODYNAMICS OF PLASTICITY
When considering an unsaturated soil, the analysis of the contribution of each fluid phase to the strain work of the solid skeleton (composed of the solid matrix together with the interfaces) is not as straightforward as it may be when the soil is saturated. In this case, changes of the Lagrangian partial porosities filled by each fluid phase da and dw are due not only to changes of the porous volume occupied by each phase but also to the invasion of the volume previously containing one phase by the other phase. To overcome this difficulty, the partial porosities are split into two parts, thus separating invasion and deformational processes: w sr Sr 0 w a (1 sr) (1 Sr) 0 a 0 a w (2) (3) (4)
the water retention curve. This relation does not account for hysteretic effects. Accounting for them is not contradictory to the approach presented here (see e.g. (Dangla et al., 1997)), but requires the consideration of appropriate energy couplings that would weight down the text. Since the considered system is a closed system (fluids have been removed), Clausius-Duhem inequality which contains the first and the second laws of thermodynamics expresses that, for isothermal evolutions, the strain work input dw to the system has to be greater or equal to the infinitesimal free energy dF that the system can store, the difference dD being spontaneously dissipated into heat. Assuming the solid grains incompressibility and using Eqs. (5)(7), Clausius-Duhem inequality writes as follows: dD (p ua) da (p uw) dw q dq 0 (ua uw) dSr dF 0 (9)
Any further analysis requires the assumptions on the dependency of the free energy F on the state variables. Denoting the elastic energy and Z the locked energy (Collins, 2005), the following dependencies are assumed: F (a ap, w wp, q qp, Sr) Z(Sr, ) U(Sr) (10)
where sr, Sr, , 0, w and a respectively stand for the Eulerian and Lagrangian degrees of saturation, (Coussy, 2007, Coussy et al., 2009), the porosities in the actual configuration and in the reference configuration and the changes due to deformation only of the Lagrangian porosity for the part of the porous network occupied by water and air. It should be noted that this partition can not be obtained when Eulerian variables are used. In this latter case, changes in sr may be indifferently due to both drying/wetting and deformation processes. Use of Eqs. (2)(4) allows expressing the work input as: dw d d d pdv ua da uw dw q d d 0 (ua uw) dSr (5) (6) (7)
The state equations are obtained considering elastic evolutions thus leading to null plastic strains and an equality in (9). They read: p ua ; q ; w q Z dU 0 ua uw Sr dSr p uw ; a (11)
According to Eqs. (5)(7), the work input is split into the contribution needed to deform the skeleton, d, and that required for the invasion process to occur, d. By noting U the fluid-solid interface energy per unit of initial volume, Eq. (7) allows for stating: dU 0 (ua uw) dSr (8)
The first three sub-equations capture to the elastic behaviour of the solid matrix. The last one corresponds to the expression of the water retention curve. It includes the effects of the deformation of the porous volume, except those leading to hysteretic phenomena as stated before. Using state equations (11) and assumptions on dependencies of F given by equation (10), Clausius-Duhem inequality is expressed as: dD (p ua) dap (p uw) dwp q dqp d 0 (12) (13)
Z Sr ,
This relation implies that U and also suction ua uw must be functions of Sr only so that ua uw r(Sr), which is the classical expression of
is the energy conjugate of the hardening variable , depending on Sr. It will be called hardening
914
0.8 0.6
at m ion )
Barcelona Basic Model

0.4 0.2
=
( Sr
iso
-d
r fo
Barcelona Basic Model
20
40 Sr (%)
60
80
100
Figure 1. Family of functions (Sr) suitable for entering in the definition of an extended Bishops effective stress. Linear function (Sr) Sr define the classical Bishops stress, step function (Sr 1) 0, (1) 1 the pair net stress/Terzaghis effective stress.
force and subsequently associated to the limit of elasticity. Introducing the plastic incompressibility of the solid grains which expresses as:
p p p p d a (1 ) d v ; d w d v
(14)
it comes from (12):

p p dD pBd v qd q d
(15)
stress where factor is assumed to be equal to the degree of saturation Sr). The remaining part of the paper is devoted to the description of a proposal with the aim of providing a better description of the suction contribution to the constitutive stress. Starting from common observations on pore size distributions of soils, it may be argued that two classes of pores are generally distinguishable. The first class corresponds to the largest pores (macropores). The fraction of the water filling these pores will be denoted SrM . The second class corresponds to the smallest pores (micropores). The fraction of the water filling these pores will be denoted Srm. It is obvious that the availability of the water filling one or the other class of pores will be different. For the first class, water exchanges are mainly governed by capillary effects. For the second class, water is quite more attached to the solids by physico-chemical interactions. This part of the water is not so freely available. As an illustration, some authors have reported experimental observations showing that this water does not participate to the darcean transport of water, thus leading to a reduced apparent permeability of the soil (Romero, 1999). Based on these microstructural considerations, it is proposed to assume that factor is no more equal to the total degree of saturation but to an effective degree of saturation defined as follows: Sre Sr Srm 1 Srm (18)
where pB is Bishops stress defined as: pB p [1 (Sr)] ua (Sr)uw (16)
Figure 1 shows Family of functions (Sr) suitable for entering in the definition of an extended Bishops effective stress. Interestingly, assuming isodeformation of the porous volumes occupied by water and air, that is:
p d w
where x 1/2(x |x|) represents Macaulay brackets. Such an assumption on factor results in the following expression for the constitutive stress: pg 1 Sre s 1 (19)
0Sr
p d a
0 1 Sr
(17)
and comparing (14) and (17) leads to (Sr) Sr, which is actually the commonly used assumption. According to (15), Bishops stress plays the same role in unsaturated states as does Terzaghis proposal (p p u) in saturated conditions. Actually, both definitions arise from solid grains incompressibility assumptions. 3 MICROSTRUCTURAL INTERPRETATION
Of course, the rough separation of the pore sizes presented above may not be so sharp in reality. Furthermore, it may be interesting to avoid the second order discontinuity at Sr Srm . This is particularly true when dealing with numerical analyses. Smoothing techniques for the corner of the piece-wise proposal have been examined. The following expression for Sre: S re = Sr Srm Sr Srm 1 + ln 1+ exp n n 1 Srm 1 Srm (20)
The previous study has lead to the conclusion that some care must be taken when using the so-called average constitutive stress (equivalent to Bishops
provides a smoothing of the corner, which is controlled by parameter n. As n increases, Eq. (20) becomes closer to the piecewise approximation (see Figure 2). In order to avoid the introduction of a new parameter and since no direct physical meaning may be attributed to n, it is advised to fix
915
Figure 2. Effective degree of saturation as a function of the total degree of saturation based on the exponential m smoothing function; influence of n for Sr 0.5.
stuff from Hong Kong (Fredlund et al., 1996) were studied. Experimental data were fitted using Eq. (22) and adjusting parameter Srm . As an illustration and for the sake of conciseness, only the case of the glacial till from Canada is presented here. Figure 3 shows the water retention curve of this soil (experimental data and bestfitted curve using MVG model). Figure 4 presents the evolution of the shear strength with suction. In addition to the values obtained using the proposal made in this study, shear strength data have been simulated using the assumption Sr. The values are reported in Fig. 5. It is obvious that these values overestimate the experimental data. Simulations using the proposal show that a good fit has been obtained.
the value of this smoothing parameter n. A high value, for instance n 20 may be used in practice. (Alonso et al., 200) proposed another smoothing function. 4 VALIDATION AND DISCUSSION
The validation of this proposal is checked on the basis of elastic stiffness and shear strength data. A set of soils have been chosen, covering a large range of soil types from mostly granular materials to high plasticity clays. In what follows, all the water retention curves have been modelled using experimental data available in the literature to fit the modified van Genuchten model MVG proposed by (Romero and Vaunat, 2000):
Figure 3. Water retention curve of a glacial till from Canada: experimental data after (Vanapalli et al., 1996) and simulated curve using modified van Genuchten model.
Sr ( s ) 1 l s
n m
ln 1 1 ln 1
s sres a sres
(21)
4.1
From shear strength data
Substitution of equation (19) into the classical Mohr-Coulomb criterion in terms of effective stress gives:
e c Sr s tan pg tan
(22)
where the first term between brackets in the right hand side corresponds to the apparent cohesion due to suction. Shear strength data of a sandy-silt from Switzerland (Geiser et al., 2006), a glacial till from Canada (Vanapalli et al., 1996) and a decomposed
Figure 4. Shear strength of a glacial till from Canada: experimental data after (Vanapalli et al., 1996) and simulated curve using the constitutive stress proposed in this paper (Eq. 19).
916
Figure 5. Water retention curve of Boom clay at a dry unit weigth of 16.7 kN/m3: experimental data after (Romero, 1999) and simulated curve using modified van Genuchten model.
Figure 6. Elastic parameter for Boom clay at a dry unit weigth of 16.7 kN/m3: experimental data after (Romero, 1999) and simulated values using the constitutive stress proposed in this paper (Eq. 19).
4.2
From elastic properties
Assuming the validity of the constitutive stress for unsaturated states and using the usual elastic equation, it comes: dp e dv p (23)
Table 1. Parameters of the water retention curve model for the different soils. n Soil Decomposed tuff Vallforns dam core Sion silt Jossigny silt Glacial till Boom clay (d 13.7 kN/m3) Boom clay (d 16.7 kN/m3) 3.63 1.11 3.25 4.56 0.59 1.14 m 0.14 0.67 0.24 0.026 0.67 0.196 sr A MPa 1000 1000 1000 1000 1000 274 274
MPa1 MPa 36.14 0.38 19.08 35.54 0.72 21.29 1.55 1000 1000 1000 1000 1000 274 274
which should be compared to the relation used in formulations using net stress as the mechanical stress variable such as in the Barcelona Basic Model (Alonso et al., 1990) and which reads as follows:
e dv
0.75 0.354
dp p
(24) shear moduli for compacted specimens of the core of Vallforns dam. The shear moduli were obtained using resonant column tests (Alonso, 1998). As an illustration, the case of the densest Boom clay is presented. The water retention curve and the elastic parameter evolution as a function of suction for this soil are showed in Figures 5 and 6 respectively. Again, comparison with simulations using Sr assumption overestimate the overall contribution of suction to the effective stress thus predicting too high values of the soil stiffness (low values of ). 4.3 Discussion Table 1 summarizes the material parameters fitted from experimental water retention properties of the different soils analyzed in this paper. In all cases, the theoretical model is a variation of van Genuchten equation (MVG model). Table 2
Various authors have reported that values were dependent on applied suction (see (Romero, 1999) for instance). It is obvious when looking at Eq. (23) that a dependence of the apparent stiffness upon suction values is directly accounted for via the presence of the suction in the definition of the constitutive stress. The analysis of values of the elastic parameter reported in the literature has been performed. This data has been reinterpreted in terms of effective stress using Eq. (23) with constant . A good fit of this data particularly of its evolution with suction may be obtained if a proper definition of the constitutive stress is considered. The data analyzed concern an aeolian deposited silt from Jossigny (France) (Cui and Delage, 1996) and a high plasticity clay from Boom (Belgium) at two different dry unit weights (respectively 13.7 and 16.7 kN/m3) (Romero, 1999). A similar analysis has also been performed on the elastic
917
Table 2. Granulometric properties and microscopic degree of saturation of the different soils. Sand/Silt/ Clay fractions Soil Decomposed tuff Vallforns dam core Sion silt Jossigny silt Glacial till Boom clay (d 13.7 kN/m3) Boom clay (d 16.7 kN/m3) % 60/35/5 54/40/6 20/72/8 4/62/34 28/42/30 18/30/52 18/30/52
Srm
0.02 0.25 0.40 0.56 0.64 0.42 0.63 Figure 7. Quasi-immobile intra-aggregate water fraction (% of total porous volume): experimental data after (Romero, 1999).
presents the material parameter Srm used to define the effective degree of saturation appearing in the constitutive stress instead of the usual assumption consisting in letting Sr. In this table, the soils are sorted in the order of increasing Srm values. It is interesting to note that the values of the material parameter Srm involved in the proposed definition of the effective stress are well correlated with the microstructure of the different soils accounted for in this study. Granulometric properties of these soils are summarized in Table 2. They show that increasing values of Srm also correspond to increasing values of the content of finer solid particles. This correlation corroborates the microstructural interpretation given earlier. Indeed, three soil classes may be tentatively identified from the fitted parameters. The first class corresponds to fairly granular soils for which the amount of microscopically trapped water is negligible thus leading to the validity of the assumtion Sr. The second group gathers silty soils for which the amount of microstructural water represents intermediate values. The finest soils correspond to the third group and thus to the highest values of Srm . An influence of the dry unit weigth is also observed: denser materials seem to be characterized by higher values of Srm which is coherent with our interpretation (denser material correspond to lower volumes of macropores and thus to higher fractions of micropore relatively to the total volume of pores). Another important point is illustrated in Figure 7. The amount of quasi-immobile water as introduced by (Romero, 1999) is plotted as a function of the dry unit weight of Boom clay. This notion is equivalent to the microstructural water Srm used in this study. The values reported by Romero correspond to estimations obtained from completely independent techniques (Mercury Intrusion Porosimetry (MIP) or permeability tests). The Srm values obtained here from elastic stiffness have been plotted and nicely fit in the original plot.
CONCLUSIONS
Thermodynamics of unsaturated soils plasticity has been revisited and a thermodynamically consistent framework for hardening plasticity has been presented. It has also been shown that the choice Sr rely on a restrictive assumption on the plastic flow rule. The proposed framework uses a generalized effective stress which includes both the commonly used effective stress with Sr and the net stress as limiting cases. In between these limiting cases, the general evolution of factor with the degree of saturation may not be identified without supplementary peace of information. To this end, a microstructural interpretation of the repartition of the water phase into the porous space has been used. This proposal has been validated on the basis of experimental data from a given set of soils ranging from low to high plasticity soils. It has been concluded that the commonly used choice for factor that is Sr, must be taken with care since important deviation from this assumption appear for relatively high plasticity soils. REFERENCES
Alonso, E.E. (1998). Suction and moisture regimes in roadway bases and subgrades. In International Symposium on subdrainage in roadway pavements and subgrades, Granada, pp. 57104. World Road Association. Alonso, E.E., A. Gens, and A. Josa (1990). A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Alonso, E.E., J.M. Pereira, J. Vaunat, and S. Olivella (2009). A microstructurally-based effective stress for unsaturated soils. Gotechnique (in press). Bishop, A.W. (1959). The principle of effective stress. Teknisk Ukeblad 106(39): 859863.
918
Collins, I.F. (2005). The concept of stored plastic work or frozen elastic energy in soil mechanics. Gotechnique 55(5): 373382. Coussy, O. (2007). Revisiting the constitutive equations of unsaturated porous solids using a lagrangian saturation concept. International Journal for Numerical and Analytical Methods in Geomechanics 31(15): 16751694. Coussy, O., J.M. Pereira, and J. Vaunat (2009). Revisiting the thermodynamics of hardening plasticity for unsaturated soils. Computers and Geotechnics, doi: 10.1016/j.compgeo. 2009.09.003. Cui, Y.J. and P. Delage (1996). Yielding and plastic behaviour of an unsaturated compacted silt. Gotechnique 46(2): 291311. Dangla, P., L. Malinsky, and O. Coussy (1997). Plasticity and imbibition-drainage curves for unsaturated soils: a unified approach. In Proc. 6th Int. Symp. Numer. Models Geomech. (NUMOG VI), Montreal, pp. 141146. Balkema, Rotterdam. Fredlund, D., A. Xing, M. Fredlund, and S. Barbour (1996). The relationship of the unsaturated soil shear to the soil-water characteristic curve. Canadian Geotechnical Journal 33(3): 440448. Gallipoli, D., A. Gens, R. Sharma, and J. Vaunat (2003). An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Gotechnique 53(1): 123135. Geiser, F., L. Laloui, and L. Vulliet (2006). Elastoplasticity of unsaturated soils: laboratory test results on a remoulded silt. Soils and Foundations 46(5): 545556. Jommi, C. and C. Di Prisco (1994). A simple theoretical approach for modelling the mechanical behaviour of unsaturated soils (in italian). In Il ruolo dei fluidi nei problemi di Ingegneria geotecnica; Proc. Italian Conference, Volume 1, Mondovi, Italy, pp. 167188.
Lewis, R. and B.A. Schrefler (1998). The Finite Element Method in Static and Dynamic Deformation and Consolidation of Porous Media (second ed.). Chichester: John Wiley & sons. Pereira, J.M., H. Wong, P. Dubujet, and P. Dangla (2005). Adaptation of existing behaviour models to unsaturated states: Application to cjs model. International Journal for Numerical and Analytical Methods in Geomechanics 29(11): 11271155. Romero, E. (1999). Characterisation and ThermoHydro-Mechanical behaviour of unsaturated Boom clay: an experimental study. PhD thesis, Universitat Polytcnica de Catalunya. Romero, E. and J. Vaunat (2000). Retention curves of deformable clays. In Tarantina A. and Mancuso C. (Eds.), Experimental Evidence and Theoretical Approaches in Unsaturated Soils, Proc. Int. Workshop on Unsaturated Soils, Trento, Italy, pp. 91106. Balkema: Rotterdam. Roscoe, K.H., A.N. Schofield, and C.P. Wroth (1958). On the yielding of soils. Gotechnique 8(1): 2253. Sheng, D., S. Sloan, and A. Gens (2004). A constitutive model for unsaturated soils: thermomechanical and computational aspects. Computational Mechanics 33: 453465. Terzaghi, K. (1936). The shearing resistance of saturated soils and the angle between the planes of shear. In Proc. 1st Int. Conf. Soil Mech. Found. Engng, Volume 1, Cambridge, MA, pp. 5456. Vanapalli, S., D. Fredlund, D. Pufahl, and A. Clifton (1996). Model for the prediction of shear strength with respect to soil suction. Canadian Geotechnical Journal 33(3): 379392. Wheeler, S.J. (1996). Inclusion of specific water volume within an elasto-plastic model for unsaturated soil. Canadian Geotechnical Journal 33(1): 4257.
919
On the time-dependent behaviour of unsaturated geomaterials

J.-M. Pereira & V. De Gennaro*
Universit Paris-Est, UR Navier, cole des Ponts ParisTech, Marne-la-Valle, France
ABSTRACT: In this paper a rate dependent model for unsaturated geomaterials is presented. The constitutive laws for the description of the time-dependent behaviour of saturated and partially saturated geomaterials generalize the isotach approach to the elastoplastic strain hardening based constitutive laws for partially saturated soils. The formulation encompasses rate and creep effects together with suction dependency on creep. It constitutes an extension of the work presented in (De Gennaro et al. 2009; Pereira and De Gennaro 2009). Some perspectives about the constitutive modelling of time-dependent behaviour of geomaterials saturated by a single or two fluids, including soft rocks are proposed. Practical applications include modelling of (i) shallow quarries submitted to humidity changes, (ii) oil reservoir formations or (iii) geological storage of CO2. In this study, the formulation of the constitutive model is presented. Numerical predictions including explicit influence of suction on time dependency of materials behaviour are compared to available experimental data on oil reservoir and quarry chalks and discussed. 1 INTRODUCTION clearly show the effects of partial saturation states on the time dependent behaviour (particularly on creep phenomena) of geomaterials. In this paper, a brief overview of the observed influence of partial saturation or fluid nature on the time dependency of the mechanical behaviour is depicted. An elasto-visco-plastic constitutive model is then proposed and validated on the basis of experimental data on two different chalks (from oil reservoir and a shallow quarry). 2 MODEL FORMULATION
Mechanical behaviour of geomaterials usually exhibit a time dependency. This dependency may be distinguished on the basis of the deformational processes which are involved. It is generally observed a small dependency on applied strain rate of the material stiffness at small strains thus leading to mainly non-viscous elastic strains (see e.g. Sorensen et al. 2007 and Tatsuoka 2009). Regarding material strength, a significant dependence on the rate of applied load or strain may be experimentally observed. Irreversible deformation due to creep under constant applied load are also generally observed. Experimental creep curves classically show a proportionality of the strains to the logarithm of time, the proportionality coefficient being the secondary compression index C or t, often used in engineering practice. These experimental observations concern a large variety of materials, including sands, clays (Bjerrum 1967), rockfill (Oldecop and Alonso 2007), rocks and soft rocks (Suklje 1957; Pasachalk 2004; Priol et al. 2007). Most of the preceding studies on this topic are related to saturated materials. Actually, despite a significant practical interest, only few studies have focused on the couplings between partial saturation states and time effects on the behaviour of geomaterials. Recently, researches on partially saturated rockfill material (Oldecop and Alonso 2007) and reser voir chalk (Pasachalk 2004; Priol et al. 2007) have been published. These works
* Presently at Schlumberger, EPRC-DCS, Pau, France.
2.1 Assumptions It is generally accepted that the elastic limit is related to the strain rate at which the material is loaded (e.g. Leroueil and Marques 1996): y A log (1)
where A and are material parameters. As suggested by Figure 1 in the case of a reservoir chalk, parameters A and depend on the nature of the fluid (in particular its wettability) saturating the porous space and on the suction if the material is saturated by two immiscible fluids. In the remaining part of the paper, it will be assumed that Eq. (1) is valid in the range of strain rates of practical interest in engineering (typically comprised between 108 to 104 s1) and that the two parameters A and are suction-dependent. Relation (1) together with this latter assumption constitute the key point of the time-dependent
921
d se
1 dq 3G
(3)
Figure 1. Evolution of yield stress of a reservoir chalk during constant rate of strain oedometric tests, 0.2 MPa refers to oil-water suction (after Priol et al. 2007).
constitutive modelling of unsaturated geomaterials proposed here. It is worth noting that the elastic regime of the mechanical behaviour will be assumed to be inviscid as mentioned in the introductory part of this paper. 2.2 Constitutive model formulation
e where ve and q are the elastic part of the volumetric and deviatoric strains respectively; and s are the elastic stiffness parameters for changes in mean net stress and suction, respectively; G is the shear modulus; patm is the atmospheric pressure taken as a reference datum and e is the void ratio. Concerning the plastic behaviour of unsaturated soils, the key feature of BBM is the introduction of an apparent suction hardening within the definition of the preconsolidation pressure, which defines the elastic limit under isotropic conditions, p0 p0 (vp s ). The proposition made in this paper extends this model to include rate effects by assuming that the preconsolidation stress also depends v , p0 p0 (vp on the strain rate v s ). Considering expression (1), extension of BBM to include rate effects is introduced in the following way:
( s ) p ( p ) p0 (vp v s ) pc 0 v v pc ( 0 )
(4)
The elastoplastic model presented here, called RASTRA, is a rate dependent model developed adopting the isotach approach proposed by uklje (1957) and Bjerrum (1967) within the framework of hardening plasticity. It is an extension of the works presented in (De Gennaro et al. 2009; Pereira and De Gennaro 2009). Besides the time dependent modelling, it also includes effects of partial saturation both on the time dependency itself and on the mechanical behaviour. These features have been developed on the basis of the well-known Cam-clay model for saturated soils and particularly its extension to unsaturated states proposed by Alonso et al. (1990): the Barcelona Basic Model (BBM). For the sake of conciseness, BBM will not be presented in detail here. As in (Alonso et al. 1990), net stress ( ua ) and suction (s ua uw) have been considered as stress state variables ( is the total stress tensor, ua and uw are the air and water pressures). It is worth noting that as far as stress state variables are concerned other choices are possible (see e.g. Pereira et al. 2005). Only triaxial stress states are considered in the following. The mean net stress p and deviatoric stress q are thus classically introduced. Since it is assumed that time effects do not influence the elastic behaviour of the material, the incremental elastic law of BBM is adopted without modification: d ve
where the saturated preconsolidation pressure reads:

1 e v ref p0 vp (vp exp v ) p0 ref v
(5)
As suggested by experimental observations reported in Fig. 1 and the schematic interpretation depicted in Figure 2a, the parameter is assumed to vary with suction. In a first approach, a simple linear relation is introduced: (s) 0 bs (6) Only three parameters related to the modelling of rate effects and their couplings with partial saturation have been introduced in the proposed vref, 0 and b. model with respect to BBM, namely ref p0 corresponds to the preconsolidation pressure of the undeformed material under saturated conditions if the material is loaded at the reference strain ref vref. In practice, p0 rate will be deduced from Eq. (5) if p0 that is the saturated preconsolidation corresponding to a given strain rate at a reference (undeformed) state is given (see Figure 2). The elastic domain is then delimited by defining the yield surface in triaxial stress space as follows: f ( p0 ) q 2 M 2 ( p ps )( p0 p ) 0 (7)
dp s ds 1 e p 1 e s patm
(2)
where ps is the tensile strength for the material at a given suction s. As in the original BBM, ps is assumed to depend linearly upon suction:
922
3.1 Isotach framework Preliminary tests are performed to illustrate the capabilities of the model to reproduce the main features related to the isotach framework and which corresponds to RASTRA added value with respect to the original BBM. The parameters used in the virtual tests presented in Figures 3 and 4 correspond to those of Lixhe chalk in Table 1. Figure 3 illustrates the isotach framework used in RASTRA model. It shows the simulation of constant strain-rate isotropic compression tests under various volumetric strain rates and a constant suction value for all tests (equal to 2 MPa). The change from one curve to the following corresponds to an increase by a factor 10 of the volumetric strain rate at which the compression is performed. It can be observed that an increase of the volumetric strain rate results as expected in the increase of the yield point. The figure also shows
(a) Schematic representation of the evolution of the preconsolidation pressure with strain rate and coupled inuence of suction.
Table 1. Model parameters used in the simulations of Estreux and Lixhe chalks.
(b) Evolution of the yield locus in der saturated conditions). plane (un-
Parameter s (0) r pc k ref p0red 0 b
Unit MPa1 MPa s1 MPa MPa1
Estreux 5 104 0.005 0.12 0.94 0.5 0.002 0.185 0.5 15.5 0.108 0.106
Lixhe 5 104 0.004 0.15 0.9 0.25 0.01 0.185 0.5 15.5 0.108 0.106
Figure 2. Influence of the applied volumetric strain rate on the yield locus.
ps ks
(8)
where k is a material parameter. It is worth noting that the yield locus associated to increases in suction (called SI in the original work by Alonso et al. (1990) is not introduced in this work. The model is completed by giving the plastic flow rule. As in original BBM, it will be nonassociated and chosen as (g being a function of parameters M, (0) and , see Alonso et al. 1990 for more details): d vp d f f d qp g d p q (9)
0.64 0.62 0.6 Void ratio, e () 0.58 0.56 0.54 0.52 0.5 increasing strain rate
where d 0 is the plastic multiplier which can be obtained using the consistency condition df 0. 3 PERFORMANCE OF RASTRA
0.48 0.46 1 10 10 Mean net stress, p (MPa)

0
Numerical simulations using RASTRA model are now presented in order to show the capabilities of the model in terms of coupled modelling of time and partial saturation effects. The model parameters used in the simulations presented in the following part of the paper are summarized in Table 1.
10
Figure 3. Simulation of constant strain-rate isotropic compression tests under various volumetric strain rates showing isotach behaviour (two following curves concern a volumetric strain rate increase by a factor 10).
923
0 0.02 0.04 Acceleration 0.06 0.08 Creep stage

Volumetric strain, v ()
0 0.02 10 0.3 Creep
Volumetric strain, v ()
0.04 0.06 0.08 0.1 0.12 0.14 0.16
10
Creep
0.1 0
10 15 20 25 Vertical stress, v (MPa)
30
35
0.18 2 10
10 10 Mean presure, p (kPa)
10
Figure 4. Simulations of Constant Rate of Strain (CRS) oedometer tests at various strain rates (all simulations are performed at a constant suction of 2 MPa).
Figure 5. Isotropic compression at various rates, including creep stages: comparison between numerical simulations and experimental data on chalk from Lixhe under constant oil-water suction of 0.2 MPa.
that the elastic and elastoplastic slopes of the soil response are not affected by the inclusion of time dependency into the constitutive model. Figure 4 shows the effects of changes of the strain rate during an oedometric compression at Constant Rate of Strain (CRS tests). The numerical test consists in three compression stages at two distinct strain rates (a fast strain rate corresponding to 106 s1 and a slow strain rate, 10 times slower). After the first compression, a creep phase is simulated. The latter is numerically performed by simply decreasing the strain rate from the fast rate to the slow rate. It may be observed that the creep phase effectively induces irreversible strains at constant vertical stress and that the acceleration of the strain rate allows the material to undergo higher stresses (strain rate hardening) before plastic strains are generated again. Two simple compressions at fast and slow constant rates are also simulated for comparison purposes (dotted curves). 3.2 Validation from experimental data
A first comparison to experimental data is now proposed. Figure 5 presents experimental isotropic compressions on Lixhe chalk at various loading rates, including creep stages. It appears that RASTRA is able to reproduce not only qualitatively but also quantitatively the overall behaviour of reservoir chalk under partial saturation. Changes in strain rate and creep phases are satisfactorily reproduced. It may be noted that the experimental data seems to show a slight dependence of the elastic modulus upon strain rate (compare the initial slopes of the first parts of both compressions which are performed at two distinct applied strain rates). This observation is however not accounted for in RASTRA and the model response thus
shows a unique response within the elastic domain (see also Figures 3 and 4). Finally, an oedometer compression test on dry Estreux chalk (Nguyen et al. 2008) has been considered. The initial suction was estimated to 30 MPa. The standard multiple stage loading procedure was simulated imposing an equivalent loading rate derived from the known elapsed time of the test. During the test the mechanical loading was stopped at a total vertical stress of 20 MPa and water was then injected within the sample under constant vertical stress. The material parameters corresponding to the elastoplastic part of RASTRA (that is those related to BBM) have been published by Nguyen et al. (2008) and have been considered for these simulations (Table 1). The missing viscous parameters for Estreux chalk have been assumed as a first estimate equal to the parameters obtained for Lixhe chalk. It should be noted that due to the particularly high value of suction at which the oedometer compression test is performed (30 MPa) a lower bound has been chosen for (see Equation (6)) assuming min 0.035. The results of the simulations are presented in Figure 6. As it can be observed, the instantaneous compaction due to water injection at the end of the compression is clearly underestimated by the model if viscous behaviour is neglected (Figure 6(a)). In other words, the mechanism of suction hardening proposed by BBM via the LC curve, frequently used in unsaturated soil mechanics to simulate instantaneous compaction under decreasing suction (wetting) is not sufficient to predict such a large compaction. Actually, if one adds to the simulation a creep phase following water injection, RASTRA model predicts a total amount of collapse which compares well to the experimental data (see Figure 6(b)).
924
0.7 0.65 0.6 Void ratio, e () 0.55 0.5 0.45 0.4 0.35 102 103 104 Vertical stress, v (kPa) 105
REFERENCES
Alonso, E.E., A. Gens, and A. Josa (1990). A constitutive model for partially saturated soils. Geotechnique 40(3), 405430. Bjerrum, L. (1967). Engineering geology of norwegian normally-consolidated marine clays as related to settlement of buildings. Geotechnique 17, 81118. De Gennaro, V., J.M. Pereira, M. Gutierrez, and R.J. Hickman (2009). On the viscoplastic modelling of porous chalks. Italian Geotechnical Journal 1/2009, 4464. Leroueil, S. and M.E.S. Marques (1996, Nov.). Importance of strain rate and temperature effects in geotechnical engineering. In Measuring and modeling time dependent soil behaviour, Proc. of the ASCE Convention, Number 61, Washington, DC, USA, pp. 160. ASCE. Nguyen, H.D., V. De Gennaro, P. Delage, and C. Sorgi (2008). Retention and compressibility properties of a partially saturated mine chalk. In D.G. Toll, C. Augarde, D. Gallipoli, and S.J. Wheeler (Eds.), Unsaturated Soils: Advances in Geo-Engineering, Durham, UK, pp. 283289. Taylor & Francis Group. Oldecop, L.A. and E.E. Alonso (2007). Theoretical investigation of the time-dependent behaviour of rockfill. Geotechnique 57(3), 289301. Pasachalk (2004). Mechanical behaviour of partially and saturated chalks fluid-skeleton interaction: main factor of chalk oil reservoirs compaction and related subsidencepart 2, final report. Technical report, EC Contract no. ENK6-2000-00089. Pereira, J.M. and V. De Gennaro (2009). Time dependent behaviour of fluids filled geomaterials: application to reservoir formations. In H.I. Ling, A. Smyth, and B. Raimondo (Eds.), Poromechanics IVProceedings of the 4th Biot Conference on Porome-chanics, Columbia University, New-York, pp. 983988. DEStech Publications, Inc. Pereira, J.M., H. Wong, P. Dubujet, and P. Dangla (2005). Adaptation of existing behaviour models to unsaturated states: Application to cjs model. International Journal for Numerical and Analytical Methods in Geomechanics 29(11), 11271155. Priol, G., V. De Gennaro, P. Delage, and T. Servant (2007). Experimental investigation on the time dependent behaviour of a multiphase chalk. In T. Schanz (Ed.), Springer Proc. Physics 112, Experimental Unsaturated Soil Mechanics, pp. 161167. Sorensen, K., B. Baudet, and B. Simpson (2007). Influence of structure on the time-dependent behaviour of a stiff sedimentary clay. Geotechnique 57(1), 113124. Suklje, L. (1957). The analysis of the consolidation process by the isotache method. In Proc. 4th Int. Conf. on Soil Mech. and Found. Engng., Volume 1, London, pp. 200206. Tatsuoka, F. (2009). Rate effects on elastic and inelastic stress-strain behaviours of geomaterials observed in experiments. In H.I. Ling, A. Smyth, and R. Betti (Eds.), Poromechanics IVProceedings of the 4th Biot Conference on Poromechanics, Columbia University, New-York, pp. 4356. DEStech Publications, Inc.
(a) Elastoplastic simulation (BBM).

0.7 0.65 0.6 Void ratio, e () 0.55 0.5 0.45 0.4 0.35 102 103 104 Vertical stress, v (kPa) 105
(b) Elasto-visco-plastic simulation (RASTRA).
Figure 6. Simulation of wetting induced compaction during one-dimensional compression test on Estreux chalk.
CONCLUSIONS
An elasto-visco-plastic model for partially saturated geomaterials has been presented. Based on the isotach framework, it extends the Barcelona Basic Model proposed by Alonso et al. (1990) for partially saturated materials to take into account time effects in terms of susceptibility to applied strain rate and creep phenomena. These time effects have been restricted to the plastic part of the model, assuming an inviscid elastic behaviour. Numerical simulations have demonstrated the capabilities of the model to qualitatively and quantitatively address fundamental aspects of the rate dependent behaviour of partially saturated geomaterials. Foreseen improvements to RASTRA include a refined dependency upon suction of the parameter of Equation (6). This refinement could also include the wettability of the saturating fluid to catch the behaviour illustrated in Figure 1.
925
A porous model based on porosimetry to simulate retention curves

E. Rojas, G. Gallegos & J. Leal
Universidad Autnoma de Quertaro, Quertaro, Qro., Mxico
ABSTRACT: In this paper a probabilistic porous model is used to simulate the retention curves for different soils. The model is built on the basis of porous networks and uses the same rules for the empting and filling of pores. However it shows important advantages over network models. For example, the computer code is fairly small and the equations are solved within seconds while network models usually require large computational codes that take several hours to run when large networks are analyzed. This model considers three different elements: the macropores, the sites and the bonds. It is able to simulate the retention curves of a material when its pore size distribution is known. It can also be used to interpret in a more realistic manner, the results of mercury intrusion porosimetry tests. In this paper the model is used to obtain the pore size distribution of different soils using the retention curve as data. The comparison between theoretical and experimental results shows that the model reproduces approximately the pore size distribution of the soil obtained from mercury intrusion porosimetry tests. 1 INRODUCTION Mbah, 1998). The degree of confidence of these models depends strongly on the quality and quantity of the data used for the statistical correlation. Other methods use the Pore Size Distribution (PSD) which in some cases is estimated from the GSD (Arya and Paris, 1981; Arya and Dierof, 1992; Basile and DUrso, 1997), and still in other cases is obtained directly from porosimetry tests (Simms and Yanful, 2002; Simms and Yanful, 2005). Because there is a direct relationship between PSD and RC then, the precision on the measurements of the PSD affects the results obtained by these models. Androutsopoulos and Man (1979) proposed a bidimensional square network model made of cylinders of different sizes randomly distributed. In this model, the Laplace equation is used to determine which pores drain or saturate at a given suction. The process of saturation or drainage of the network starts at the borders and continues through a saturated or drainage path in a quasistatic flow where suction monotonically increases or decreases. Considering a log-normal distribution for the diameter of the pores, these researchers were able to reproduce the RC of a cobalt/molybdenum porous media for both the adsorption and the desorption path. However, the experimental PSD of this material was not reported and its comparison with the theoretical PSD used for the simulation remains unknown. Simms and Yanful (2005) used a similar model but it included the phenomenon of pore shrinkage by suction increase. These researchers used this model to reproduce the
The Retention Curve (RC) is the relationship between suction and the water content or the degree of saturation of the soil. The relevance of the RC in the study of the behavior of unsaturated soils has increased since the hydro-mechanical coupling phenomenon in these materials has been acknowledged. Evidence of the hydro-mechanical coupling is the influence of the degree of saturation on the strength of unsaturated soils when identical samples are subjected to the same test conditions. This implies that a soil sample following a wetting path shows different strength as compared to one following a drying path even if suction is the same for both samples. Other evidence of this coupling is the influence of volumetric deformations on the RC induced by mechanical actions on the soil sample. Recent constitutive models for unsaturated soils involve the modeling of the RC to introduce the hydro-mechanical coupling (Wheeler S., 2003; Tamagnini R., 2004; Sheng et al., 2004). Different models and empirical relationships have been proposed for the RC. Some of these equations use parameters related to the air entry value, the residual water content and the main slope of the curve (Fredlund and Xing, 1994; van Genuchten, 1980). Others are based on the Grain Size Distribution (GSD) and other soil properties, and also use a statistical correlation between soil data and water content (Fredlund et al., 1997; Aubertin et al., 1998, Mbagwu and
927
experimental RC and PSD obtained from Mercury Intrusion Porosimetry (MIP) tests. However, it is known that MIP tests do not produce the real pore size distribution of porous media (Abell et al., 1999). In addition, one of the main problems with computational models is the size of the network because it influences the results and therefore very large networks are required to overcome this problem. However, the capacity of Personal Computers (PC) is limited in that sense. On the other hand, when the size of the network increases, the computing time multiplies up to the point that it becomes impractical. In the next sections, a more realistic model is described which is able to interpret MIP tests and reproduce the RC using the PSD of the material. 2 THE POROUS MODEL
The description of the porous model can be found elsewhere (Rojas and Rojas, 2005). In this paper only the main aspects of the model are emphasized. As aforementioned, unsaturated soils show a hydro-mechanical coupling phenomenon in which the hysteresis of the RC and its dependence on porosimetry play an important role. In that sense, it is of major importance for the proposed model to reproduce the porosimetry of the soil, and to be able to simulate the phenomenon of hysteresis as well as the volumetric deformation of the material. It is well known that hysteresis is basically generated by two factors: the diversity of pore sizes and the interconnectivity between pores (Everet, 1967). Furthermore, in a series of porosimetry tests made by mercury intrusion on different soils, Simms and Yangful (2001) observe the two types of pores; the large pores or macropores and the small pores. The first type reduces its size upon loading or suction increases while the second remains unaffected for the most part. Accordingly, a porous network formed by the following three elements is proposed: the macropores, the sites and the bonds. The macropores are the largest pores and are responsible for most of the volumetric deformation of the material. The sites are small pores that do not change their size upon loading. The conjunction of macropores and sites are called cavities. Finally, the bonds are the elements that link all cavities together. In this model it is considered that the filling or draining of pores is regulated by the Laplace equation which is written as: ua uw 2T cos /RC (1)
interphase and RC is the critical radius i.e., the maximum size of pores that remain saturated at a particular suction. This relationship is used by the vast majority of porous networks to model the process of imbibitions or draining of the pores. In general the contact angle between water and soil minerals is considered to be zero ( 0). In that case, the above equation simplifies to ua uw 2T/ RC. With this relationship it is possible to establish the conditions for a pore to drain or saturate. For example, if R represents the size of pores, then those pores satisfying the condition ua uw 2T/R are able to saturate while those satisfying the condition ua uw 2T/R are able to dry. The simplest porous networks are bi-dimensional. In such a case, the cavities are represented by circles and the bonds by rectangles. In a three-dimensional configuration, the cavities are represented by spheres and the bonds by cylinders. The connectivity of the network represents the number of bonds concurrent to a site. It can take on a constant value throughout the network or may vary from site to site. In this last case it is sufficient to consider a certain number of bonds with null radius. Figure 1 shows a portion of a bi-dimensional network with its different elements. Because the network is built with elements of different sizes, it must comply with the following geometrical restriction or constructive principle which ensures that the model can be physically built: if rb1 and rb2 represent the size of two concurrent bonds meeting at 90 angle to a site size of rst they should satisfy the following condition rb12 rb 2 2 rst (2)
This condition ensures that concurrent bonds do not overlap one another. This principle also
where represents the contact angle between water and solids, T is the tension in the air-water
Figure 1.
Portion of a porous network.
928
implies that any site should be larger than its concurrent bonds. Once the porous network has been defined, it is possible to establish the general conditions for a pore to drain or saturate. For example, when the network is subject to a drying process, only those pores complying with the following two conditions will be able to dry: a) their size is larger than the critical size RC and b) at least one of the elements connected to this pore has already been drained, or in other words, it is connected to the bulk of gas. On the other hand, when the network is subject to a wetting process, only those pores complying with the following conditions may saturate: a) its size is smaller or equal to the critical size RC and b) at least one of the elements connected to this pore is already saturated, meaning it is connected to the bulk of water. Different computational porous models have been developed to build up virtual networks which include several millions of nodes and simulate certain hydraulic phenomena. In general, these models show the influence of the size of the network on results attained (Androutsopoulos and Mann, 1979; Simms and Yanful, 2005). This occurs because the pore size distribution of soils involves several orders of magnitude. Therefore, not all the sizes of the pores are well represented in small networks simply because a single gram of fine soil has billions or trillions of pores. Also the ratio between the number of pores at the borders and the number of pores inside the borders is substantially affected by the size of the network. For that reason, computational porous networks intended to realistically simulate a hydraulic process require a memory size thousands or millions of times larger than that available in an ordinary PC. In addition, the computing time is a factor that must be considered. For these reasons, these types of models become impractical, at least for the case of fine soils. 3 THE PROBABILISTIC MODEL
probability represents the proportion of pores of certain size that are dried or saturated at a given suction. With this data, the degree of saturation of the material can be obtained. Finally, both the wetting and drying RCs can be plotted. This process requires the knowledge of the distributions of the relative volumes of cavities (VRS) and bonds (VRB) as a function of their size. The relative volume is defined as the volume of elements of certain size divided by the total volume of those elements. These distributions can be obtained from the results of porosimetry tests. The analysis to define the probability equations of the model can be found elsewhere (Rojas and Rojas, 2005). Due to the limited space, this paper only presents the application of these equations to interpret the results of MIP tests and to simulate the RC of different soils. By simulating the RC of a soil, its porosimetry is produced and it can be compared with the experimental measurements. Another advantage of the probabilistic model is that all pore sizes are well represented even if their size distributions involve several orders of magnitude, which is not the case for the computational models. This has important consequences on the results. While the probabilistic model does not require defining a network size, it is a basic parameter for computational models. Probably the largest possible network size in an ordinary PC may not be sufficient to nullify the size effect on the results. To simulate the RC, the model requires the PSD of macropores, sites and bonds. However, a MIP test only provides the PSD of macropores and sites but not that of bonds, as these pores are too tiny and their volume is so small that these cannot be identified using this procedure (Dullien, 1992). To overcome this problem, the inverse approach can be used. In such a case, the RC is used as the data and the PSD is obtained as the result. In that case, an initially proposed PSD is successively modified until the best fit for the RC is obtained. This procedure is used in this paper to test the proposed model. 4 MIP TESTS
Based on the framework of the computational network, it is possible to develop a probabilistic porous model. The procedure to develop the model is as follows: First of all, an infinite bi or three-dimensional network made of macropores, sites and bonds is considered. Thereafter, the conditions to drain or saturate a pore (macropore, site or bond) during a drying or a wetting process are established. Then, based on the size distribution of each element, it is possible to write the above conditions as probability equations. These equations can then be simultaneously solved. Then, the probability of a pore of certain size to drain or saturate during a drying or wetting process can be determined. This
The pore size distribution from MIP tests is obtained by assuming that the soil can be represented as a bundle of capillary tubes with no connection between them. However, this hypothesis to interpret MIP test is clearly unrealistic. It is equivalent to assuming that only those pores with the same size are interconnected while there is no interconnection between pores of different sizes. In fact, it has been recognized that the results of MIP tests exaggerates the frequency of small pores while under-estimating
929
that of large pores (Abell et al., 1999). This occurs because bonds control the entrance of fluid in sites and macropores. Therefore, sites and macropores fill at the same time as the first of their concurrent bonds does. The porous model described in the previous section can be used to interpret the results of MIP tests. The invasion of mercury (which is a non wetting fluid) is similar to a drying process where pores are invaded by air (which is also a non wetting fluid) forcing the water to drain out of the sample. In both cases, the largest pores are the first to fill in with fluid. When pores of all sizes are interconnected and there is an increase of mercury pressure, only a fraction of those pores the size of the critical radius will saturate while the rest remain blocked by smaller bonds. In addition, some larger pores saturate during this pressure increase because at least one of their interconnected bonds belongs to those that saturate with this increment. This mechanism is sketched in Figure 2. Consider that the critical radius reduces from RC1 to RC2 due to the increase in mercury pressure. The blank zone below the curve represents the volume of pores that have not yet been invaded by mercury. The single shadow zone represents the volume of pores invaded by mercury before the new pressure increment. Finally, the double shadow zone represents the volume of pores invaded during the new pressure increment. Furthermore, the appropriate values for the contact angle and superficial tension have to be taken into consideration for each fluid. Consequently, the ratio of the suctions of a pore filled with water (sw) and mercury (sm) is given by the relation sw Tsw cos w sm Tsm cos m (3)
Therefore, the porous model can simulate a MIP test if it is considered that the sample is subject to a drying process where the volume of air represents the intruded mercury and the critical radius is computed using Equation 3. 5 RESULTS COMPARISON
where Tsw and Tsm represent the superficial tension for water and mercury, respectively and w and m are the contact angles for water and mercury with the soil minerals, respectively.
Drying
Volume cm3/gr
RC2 RC1
Radius
Figure 2. Pores invaded by mercury (shaded zone) during a MIP test.
In order to test the capabilities of the probabilistic model and verify the interpretation of MIP tests as described in the previous section, the experimental results reported by Li and Zhang (2007) are used. These researches produced five different types of soil from mixtures of completely decomposed granite with sizes varying from gravel to silty clay. The five types of fabricated soil were: poorly graded gravel (GP), well graded gravel (GW), poorly graded sand (SP), well graded sand (SW) and a silty sand (SM). The basic characteristic of these materials can be found in Li and Zhang (2007). Different tests were performed on these materials which included the GSD, the PSD and the RC at drying. The PSD was obtained from MIP tests for both conditions saturated and oven dried. The RC for each material was obtained using a pressure plate apparatus up to 500 kPa, and in some cases this technique was complemented with the psychrometer technique reaching values of suction up to 50 000 kPa. The usual values for water and mercury parameters are Tsw 0.072 N/m, w 0, Tsm 0.472 N/m and m 140, respectively. The size of the samples for MIP tests was limited to 15 cm3. For that reason and in order to maintain the homogeneity of the material, soil particles larger than 10 mm were taken out of the soil while preparing the samples. This procedure could affect the porosimetry of the different soils to some extent, especially those with a large content of gravel (GP and GW). In this section, the comparison between experimental and theoretical results for the RC and the PSD for three of these soils are presented. For all samples, the theoretical RC has been fitted with the experimental results by consecutively modifying an initially proposed PSD as explained earlier. Once the best fit of the curves has been obtained, the resulting PSD of the soil was used to simulate a MIP test according to the principles which are outlined in section 4. Finally, the PSD obtained from this simulation of the MIP test was compared with the experimental data. These results are presented in Figures 3 to 5 for soils SM, GP and SW, respectively. All tested soils (Figure 3 to 5) show good results when comparing the theoretical PSD obtained from the simulation of the MIP test with the experimental results. It can also be observed that the theoretical results obtained from the simulation
930
SM Degree of saturation
1 0.8 0.6 0.4 0.2 0 0.01
Theoretical
Experimental
SW
Theoretical
Experimental
1 0.8 0.6 0.4 0.2 0 0.001
0.1
10
100
0.01
0.1
10
100
Suction (MPa)
Suction (MPa)
(a)
SM
(a)
MIP(exp) PSD
SW 0.25 MIP(theo) MIP(exp) PSD
MIP(theo)
0.2
Relative volume
0.25
Relative volume
0.01 0.1 1 10 100 1000
0.2 0.15 0.1 0.05 0 0.001 0.01 0.1 1 10 100 1000
0.15 0.1 0.05 0 0.001
Radius (m)
Radius (m)
(b)
(b)
Figure 3. Theoretical and experimental results for (a) the drying retention curve and (b) the pore size distribution for soil SM.
GP Degree of saturation
1 0.8 0.6 0.4 0.2 0 0.0001 0.001 0.01 0.1 1 10 Theoretical Experimental
Figure 5. Theoretical and experimental results for (a) the drying curve and (b) the pore size distribution for soil SW.
Suction (MPa)
(a)
GP MIP(theo) MIP(exp) PSD 0.15 0.1 0.05 0 0.001
0.01
0.1
10
100
1000
Radius ( m)
(b)
Figure 4. Theoretical and experimental results for (a) the drying curve and (b) the pore size distribution for soil GP.
of a MIP test are consistently displaced to lower pore sizes than the experimental values. This disagreement between theoretical and experimental PSD could be caused by the difference in the value of the contact angle between soil minerals and water or mercury when the fluid intrudes or retracts from a pore (Bear, 1979). According to
Dullien (1992) this phenomenon is caused by the surface immobility in a macromolecular scale. In other words, there could be some difference in the values of the contact angle measured in static conditions in the laboratory and those effectively occurring in the pores of soil during the invasion or retraction of fluid. It is noteworthy that the difference between the theoretical PSD obtained from the best fir of the RCs and that obtained from the simulation of a MIP test is about one order of magnitude. This result agrees with the observations made by Abell et al. (1999) in the sense that MIP tests exaggerate the frequency of small pores while they tend to under-estimate that of large pores. This result also explains why the efforts to directly reproduce the RC from the PSD obtained from MIP test, is usually disappointing (Prapaharan et al., 1985). On the other hand, a better evaluation of the model could be done if both retention curves, the drying and the wetting branch, were available for each sample. Either way, according to these comparisons, it can be concluded that the probabilistic model is able to approximately reproduce the PSD of the soil based on the RC or vice versa. Finally, Figure 6 shows a plot of the experimental GSD for soil SM and the theoretical PSD obtained for the same soil. This result agrees with the observations made by Alonso et al. (2008) in the sense that both PSD and GSD are closely related. This result also provides an explanation about why those models based on the GSD may simulate the RC of soils fairly well.
Relative volume
931
SM
0.2
Pores
Solids
Relative volume
0.15 0.1 0.05 0 0.01 0.1 1 10 100 1000 10000 100000
Radius ( m)
Figure 6. Theoretical pore size distribution and experimental grain size distribution.
CONCLUSIONS
The probabilistic porous model presented in this paper can be used to make a more realistic interpretation of MIP tests. The theoretical and experimental results comparisons show that this model can approximately reproduce the PSD obtained from MIP tests when the retention curve is used as data. For a better evaluation of the model both the wetting and drying RC would be necessary. It is possible that some differences between theoretical and experimental results can be attributed to the variation of the contact angle between soil minerals and water (or mercury) when the fluid is invading or retracting from the pores. ACKNOWLEDGEMENT The authors would like to acknowledge Silvia C. Stroet of the Engineering Faculty at UAQ University for her efforts in editing the English of this document. REFERENCES
Abell A.B., Willis K.L. and Lange D.A. 1999. Mercury intrusion porosimetry and image analysis of cementbased materials. Journal of Colloid and Interface Science 211: 3944. Alonso E.E., Rojas E. and Pinyol N.M. 2008. Unsaturated soil mechanics, Reunin Nacional de Mecnica de Suelos. Aguascalientes, Mxico, Especial Volume:117205. Androutsopoulos G.P. and Mann R.1979. Evaluation of mercury porosimeter experiments using a network pore structure model. Chemical Engineering Science 34: 12031212. Arya L.M. and Dierolf T.S. 1992. Predicting soil moisture characteristics from particle size distributions: an improved method to calculate pore radii from particle radii. Proc. Int. Workshop Indirect Meth. Estim. Hydraulic Prop. Unsat. Soils. Univerity of California, Riverside: 115124.
Arya L.M. and Paris J.F. 1981. Physicoempirical model to predict the soil moisture characteristic from particlesize distribution and bulk density. Soil Science Society of America Journal 45(6): 10231030. Aubertin M., Richard J.F. and Chapuis R.P. 1998. A predictive model for the water retention curve: Applications to tailings from hard.rock mines. Can. Geotech. J. 35(1): 5569. Basile A. and DUrso G. 1997. Experimental corrections of simplified methods for predicting water retention curves in clay-loamy soils from particle size determination. Soil Technology 10(3): 261272. Bear J. 1979. Hydraulic of Groundwater. McGraw-Hill, Series of Water Resources and Environmental Eng. Dullien F.A.L. 1992. Porous media, fluid transport and pore structure. Academic press, USA. Everet D.H. 1967. The solidgas interface. Edisson Flood ed., vol II, p 10051010, Dekker, New York. Fredlund M.D., Fredlund D.G. and Wilson G. 1997. Prediction of the soil-water characteristic curves from grain-size distribution and volume-mass properties. Proc. Third Brazilian Symp. Unsat. Soils, Rio de Janeiro, Brazil. Fredlund D.G. and Xing A., 1994. Equations for the soil-water characteristic curve, Can. Geotech. J. 31: 521532. Li X. and Zhang L. 2007, Prediction of SWCC for Coarse soils considering pore size changes. Experimental Unsaturated Soil Mechanics. Ed. Elsevier, Part VI: 401412. Mbagwu J.S.C. and Mbah C.N. 1998. Estimating water retention and availability in Nigerian soils from their saturation percentage. Communications in Soil Science and Plant Analysis 29(7/8): 913922. Prapaharan S., Altschaeffl A.G. and Dempsey B.J. 1985. Moisture curve of compacted clay: mercury intrusion method. J. Geotech. Engng. 111(9): 11391143. Rojas E. y Rojas F. 2005. Modelling hysteresis of the soilwater characteristic curve. Soil & Foundations 45(3): 135146. Sheng D., Sloan S.W. and Gens A. 2004. A constitutive model for unsaturated soils: Thermomechanical and computational aspects. Computational Mechanics 33(6): 453465. Simms P.H. and Yanful E.K. 2001. Measurement and estimation of pore shrinkage and pore distribution in a clayey till during soil-water characteristic curve tests. Can. J. Geomech. 38(4): 741754. Simms P.H. and Yanful E.K. 2002. Predicting soil-water characteristic curves of compacted plastic soils from measured pore-size distributions. Gotechnique 52(4): 269278. Simms P.H. and Yanful E.K. 2005. A pore network model for hydromechanical coupling in unsaturated compacted clayey soils. Can Geotech. J. 42: 499514. Tamagnini R. 2004. An extended Cam-clay model for unsaturated soils with hydraulic hysteresis. Gotechnique 54(3): 223228. van Genuchten 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Am. J. 44: 892898. Wheeler S.J., Sharma R.S. and Buison M.S.R. 2003. Coupling hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Gotechnique 53(1): 4154.
932
Modeling the tensile strength of soils

E. Rojas, D. Hurtado, M.L. Prez-Rea & A. Zepeda
Universidad Autnoma de Quertaro, Qro., Mexico
ABSTRACT: In this paper, a solid-porous model is used to simulate the tensile strength of unsaturated soils tested at different water contents. According to the independent stress state variables principle, the strength of unsaturated soils can be split into two parts: one related to the net stress and the other to suction. The strength generated by suction originates from the additional contact stress induced by water meniscus to solid particles. This additional contact stress is called suction stress. In that sense, the tensile strength of soils represents the suction stress of the material at that particular water content. The theoretical and experimental result comparisons of the tensile strength of an unsaturated soil tested at different water contents show that the solid-porous model presented in this paper can define the value of the suction stress with sufficient accuracy. 1 INTRODUCTION where represents Bishops effective stress parameter, f s Vs/V is the saturated fraction of the soil and Vs and V the saturated and the total volume of the material, respectively, s is the sucu u Vw Vvu is the degree of saturation of the tion, Sw unsaturated fraction, and Vwu and Vvu are the water and voids volumes of the unsaturated fraction, respectively. The above equations takes into consideration that the soil is a homogeneous isotropic material and in that sense the suction stress represents an isotropic stress. Furthermore, all additional electrochemical forces between solid particles are not considered herein. During a direct tensile test the maximum strength reached by the soil represents the linking stress between solid particles and therefore it also represents the suction stress of the material at that particular water content (Vesga, 2008). Thus, tensile tests can be used to determine the suction stress of a soil subjected to different water contents and these values can be compared with the theoretical results given by Equation (2). However, one of the main disadvantages of u Equation (2) is that Parameters f s and Sw cannot be obtained directly in the laboratory. For example, Klubertanz et al. (2002) have been using the neutron tomography procedure and the image processing to study the flow of water and the strains of unsaturated materials. These images discriminate the solid, the water and the air phases in different cuts and the solid structures of the material can be reproduced with the assistence of a computer. This procedure could be used to determine the value of u paramters f s and Sw required to establish the equivalent stress of an unsaturated soils. Nevertheless,
Based on the static analysis performed on an elementary volume of soil showing a bimodal structure, the following equation for the shear strength of unsaturated soils can be obtained (Rojas, 2008)
n tan ( nt ) n s tan nt n tan c , where n
(1)
(nt)n and s represents the normal equivalent, the net normal and the suction stress, respectively, is the internal friction angle of the material and c is the cohesion. Bimodal structure means that the soil possesses a microstructure and a macrostructure (Sridharan et al., 1971). The former is related to the smallest pores of the soil generally present in the form of aggregates or pellets of fine material. On the other hand, the macrostructure is related to those pores resulting from the assemblage of pellets and coarse materials. In this way, when a small suction is applied to the material it can be assumed that the whole microstructure and that most part of the macrostructure remain saturated and only in some parts of the macrostructure, the meniscus of water appear. The total volume of the saturated zones including the solids, represent the saturated volume of the soil whereas its compliance represents the unsaturated volume. The suction stress (Lu, 2008) represents an additional contact stress induced to solid particles by water meniscus and is given by the following relationship (Rojas, 2008)
u 1 f s s s s f s Sw
(2)
933
the resolution of the most sophisticated equipment is around 0.125 mm which means that this method is not applicable for silty or clay soils and, only the images of coarse and medium sands have been generated so far. An alternative to determining parameters f s u and Sw is the use of a solid-porous model. In such a case, the model should be able to simulate the structure of soils and the distribution of water inside the pores at any value of suction. A model of this type is described in the next section. 2 THE SOLID-POROUS MODEL
A porous network attempting to simulate the structure of soils requires different elements. For example, when a set of spheres of different sizes is randomly placed in a transparent box, three different types of elements can be discriminated. First, a small quantity of very large pores or macropores that are formed from a particular arrangement of solids whose size is smaller than the pore itself. This type of pores is highly unstable and is responsible for most of the volumetric strains of the material upon loading (Simms and Yangful, 2001). Second, a large quantity of medium pores, called sites, that are smaller than their surrounding solids meaning that they are more stable and practically do not participate in the volumetric deformation of the material. The collection of macropores and sites are called cavities. Finally, it is observed that all these cavities are linked together by throats or bonds. These elements are very flat as they represent the windows or passages through which all pores communicate. Therefore, their volume is very small and very difficult to discriminate from porosimetry tests. These observations are confirmed from the porosimetry tests reported by different researchers, see for example Simms and Yangful (2001), Penumadu and Dean (2000), Delage et al. (1996). Then, in order to develop a model capable of describing the porous structure of soils, at least these three elements need to be considered. Each one of them has its own size distribution which can be determined from porosimetry tests. On the other hand, the porous structure of the soil can be reproduced by means of the phenomenon of segregation. This phenomenon consists of pores possessing a similar size that gather together in different zones of the network. In this way, the microstructure of the material is represented by the small pores that gather together in different zones, simulating the packets of fine soil. In the same way, the macrostructure of the material is represented by the large pores that gather in other zones of the network. The phenomenon of segregation appears naturally in porous networks when
the size distribution of bonds and sites overlap. The larger the overlap in the size distributions of these elements, the more the macrostructure and the microstructure of the soil differentiate. In addition to the porous structure, the solid phase of the soil must be included into the model. To that purpose, the grain size distribution is used as data and the solids are placed in between the pores following a solid-pore size strategy. This strategy consists of selecting the size of the solid to be placed in a certain node, according to the size of its surrounding pores. In this way, the smallest solids are accommodated in the zones representing the microstructure of the material. On the contrary, while the largest solids are placed in those zones representing the macrostructure of the soil where the largest pores appear. Using the information related to the distribution of water in the pores of the network at each increment of suction, the model is capable of reproducing the wetting and drying retention curves of the soil. With the same information it is also possible u to establish the value of paramters f s and Sw which are required to compute the suction stress and the strength of the material as a function of its water content. The procedure to build up a solid-porous network is as follows: first, a bi or tridimensional network is generated where each node is identified. The number of nodes of the network represents the total number of cavities (sites and macropores) and the connectors represent the bonds. The pore size distribution of each element namely macropores, sites and bonds is divided in sections, each section represents the proportion of elements of each size. This proportion is then multiplied by the corresponding total number of elements of each type which results in the number of elements of each size that has to be accommodated within the network. Once the number of macropores, sites and bonds of each size is known, they are randomly placed in the network. At this stage, it is necessary to verify that the network is physically possible. For that purpose a construction principle is established. This principle avoids the overlap between different elements of the network and can be enunciated as follows: two bonds with sizes rb1 and rb2 converging to the same site of size rst at an angle of 90, should satisfy the condition rb12 rb 2 2 rst (3)
This condition avoids the overlap between adjacent bonds. It also implies that the size of a bond is always smaller than the sites it connects. The construction principle is tested everywhere in the network. When a violation to this principle occurs,
934
the bonds involved are permuted with the bonds of another site untill no violation to the construction principle can be found. Once the porous network has been built, the solids are also randomly placed according to the aforementioned size strategy. Once the model has been completed, it is possible to establish the conditions for the imbibitions or drainage of the pores. The model considers that the filling or draining of pores occurs according to the Laplace equation ua uw 2T cos/RC (4)
successively modifying an initially proposed pore size distribution. Once the best fit is achieved, the resulting pore size distribution can be used to determine the cohesive stress at any value of the degree of saturation as it is explained in the next section. 3 APPLICATION TO TENSION TESTS
where represents the contact angle between water and soil minerals. For the case of soils it is considered that 0. T is the tension in the airwater interphase and RC is the critical radius which represents the maximum radius of pores that are able to saturate at a particular suction. Using the Laplace equation it is possible to establish the conditions for a pore to dry or saturate. For example, consider a drying process. Initially, all pores are saturated and suction equals zero. At some point, suction is increased to a particular value and some pores will drain. These pores are able to dry because they comply with the following conditions: a) its size is larger than the critical size given by Equation (3) and b) at least one of its connecting elements is already dry, meaning that it is connected to the bulk of gas. Now consider a wetting process. Initially all pores are dry and suction is high. When suction reduces to a certain value, only those pores complying with the following conditions are able to saturate: a) its size is smaller or equal to the critical size and b) at least one of its connecting elements is already saturated meaning that it is connected to the bulk of water. Keeping all these considerations in mind, it is possible to create a computational code capable of generating the porous network, determining the number of macropores, sites and bonds of each size, distributing all these elements randomly and verifying the construction principle throughout the network. Subsequently, the solids can be distributed according to the aforementioned size strategy. Then, the wetting and drying process can be simulated and the number of pores filled with water can be determined. With this information, the retention curves at wetting and drying can be u plotted. Also, parameters f s and Sw can be obtained at each increment of suction. Finally, the suction stress can be determined from Equation 2. When the pore size distribution of the material is not available, it is possible to invert the procedure. In such a case, the theoretical retention curves are fitted to the experimental curves by
Vesga and Vallejo (2006) performed a series of direct tension tests on kaolin samples with different degrees of saturation following a drying path. At the same time these researchers reported the retention curve of the material. The tension tests were performed on flat bowtieshaped samples. In this manner, the samples could be fixed at their extremes and the failure always occurred at the centers. The samples were 7 cm long, 2.2 cm thick with a central neck 2.5 cm wide. These samples were casted in a flat mould where the material was placed at a water content close to the liquid limit (40%). Then a vertical load of 30 kPa was applied for 24 hours. Subsequently, the sample was subjected to a drying process in controlled humidity conditions until the sample reached the required water content. Finally the sample was placed in a membrane during 48 hours to allow the homogenization of the humidity before the test was performed. Unfortunately, all these tests were performed following a drying path and no information on the wetting path was reported. In addition, as the porosimetry of the material was missing, the porous-solid model was used to simulate the retention curve of the material by successively adjusting an initially proposed pore size distribution as previously explained. Figure 1 shows the experimental retention curve obtained by the researchers using the filter paper method. This figure also shows the fitted theoretical retention curve obtained with the porous model. In this case a single logarithmic function was considered for both macropores and sites whereas a double logarithmic distribu-
Theoretical Degree of saturation

1 0.8 0.6 0.4 0.2 0 0.01 0.1 1
Experimental
10
100
Suction (MPa)
Figure 1. Simulation of the retention curve.
935
1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1 Degree of saturation
Figure 2.
Theoretical pore size distribution.
Figure 4.
SWU
u Values of parameter Sw .
1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1
1 0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 Degree of saturation 0.8 1
fS
Figure 3.
Values of parameters f s.
Figure 5.
Values of parameter .
tion was used for bonds in order to achieve the best fit for the retention curve. The required data for each element consist of the mean radius and the standard deviation. These parameters establish the frequency of the pores of each size in the porous network. The best fit of the retention curve was obtained with the following parameters for macropores (subindex M) sites (subindex S) and bonds (subindexes B1 and B2): RM 0.0014 m, RS 0.075 m, RB1 0.0009 m, RB2 0.03 m, S1 5.1, S2 1.5, B1 7 and B2 1.5. Using this data and the size of the pores, it was possible to determine the theoretical relative volume of each size and for each element as is shown in Figure 2 where the size distribution of cavities (macropores and sites) and that of bonds have been plotted. Once the relative volume for each size of pore is known, it is possible to determine u and from the expressions preparameters f s, Sw sented in section 1. The values of these parameters are shown in Figures 3, 4 and 5, respectively. Once the value of parameter has been obtained, it is possible to compute the suction stress from Equation 2. Finally, Figure 6 shows the comparison between the experimental tensile stress and the suction stress. This comparison shows that the model predicts a maximum tensile stress slightly greater than the experimental value. Furthermore, the theoretical maximum stress is
Theoretical
Experimental
Cohesive stress (Mpa)
0.8 0.6 0.4 0.2 0 0 0.2 0.4 0.6 0.8 1
Figure 6. Theoretical and experimental results comparison for direct tensile tests.
displaced to the left with respect to the maximum experimental value. However, the shape and values of the theoretical curve correspond approximately to the experimental points. According to these results, it can be asserted that the proposed solid-porous model approximately reproduces the porous structure of a soil and can be used to interpret the results of direct tension tests performed at different water contents. One point that requires further investigation is the fact that during the direct tensile tests suction was not controlled and therefore it could modify its value during the deformation of the sample.
936
CONCLUSIONS
The solid-porous model described in this paper u can determine the values of parameters f s and Sw which are required to compute the suction stress of soils. A direct measurement of the suction stress is obtained from tension tests. The comparison between theoretical and experimental results shows that the model is capable of simulating the tensile strength of soils tested at different water contents with sufficient accuracy. REFERENCES
Delage P., Audiguier M., Cui Y-J. y Howat M.D. 1996. Microstructure of a compacted silt. Can. Geotech. J. 33: 150158. Kluberlanz G., Laloui L., Vulliet, L. y Gachet Ph., 2002. Experimental validation of the hydromechanical modeling of unsaturtade soils, Workshop Chemo-Mechanical Coupling in Clays: 223230.
Lu N. 2008. Is matric suction a stress variable? Journal of Geotechnical and Geoenvironmental Engineering 134(7): 899905. Penumadu D. and Dean J. 2000. Compressibility effect in evaluating the pore-size distribution of kaolin clay using mercury intrusion porosimetry, Can. Geotech. J. 37: 393405. Rojas E. 2008. Equivalent stress for unsaturated soils. Part I: The equivalent stress. International Journal of Geomechanics 8(5): 285290. Simms P.H. and Yangful E.K. 2001. Measurement and estimation of pore shrinkage and pore distribution in a clayey till during soil-water characteristic curve tests. Can. J. Geomech. 38: 741754. Sridharan A., Altschaeffl A.G. and Diamond S. 1971. Pore size distributions studies. J. Soil Mech. Found. Div. 97(5): 771787. Vesga L.F. and Vallejo L.E. 2006. Direct and indirect tensile tests for measuring the equivalent effective stress in a kaolinite clay. Fourth Int. Conf. Unsat. Soils, Arizona, USA, 1: 129013. Vesga L.F. 2008. Equivalent effective stress and compressibility of unsaturated kaolinite clay subjected to drying. J. Geotech. Geoenv. Engrng. 134(3): 366378.
937
Modelling the unsaturated behaviour of structured soils

M. Rouainia
School of Civil Engineering, University of Newcastle, Newcastle, UK
J.-M. Pereira
Ecole Nationale des Ponts et Chaussees, Marne la Vallee Cedex, France
ABSTRACT: The mechanical behaviour of natural clays is significantly affected by their in situ or initial structure in the form of cementation or interparticle bonding. This behaviour can differ substantially from the behaviour of reconstituted clays. In this paper, a constitutive model that combines unsaturated and structured behaviour is proposed. The model is capable of accounting for the phenomenon of volumetric collapse upon wetting and the stiffening effect that suction has on the soil skeletal response. Illustrative numerical results for isotropic consolidated tests on both on saturated and unsaturated compressed samples of So Paolo soil demonstrate the potential of the proposed model. 1 INTRODUCTION approach (Loret and Khalili 2002; Gallipoli, Gens, Sharma, and Vaunat 2003; Pereira, Wong, Dubujet, and Dangla 2005; Russell and Khalili 2006). The choice of stress variables has nevertheless been the subject of much debate (Khalili, Geiser, and Blight 2004; Gens, Sanchez, and sheng 2006; Nuth and Laloui 2008). Constitutive models for unsaturated soils require an additional equation relating changes in suction to is related changes in degree of saturation, just as to in saturated models. This equation describes the water retention properties of the material which are in general strongly coupled to the mechanical behaviour. For monotonic (wetting or drying) processes, a single water retention curve is sufficient for modelling purposes. In general, however, the suction-saturation relationship is hysteretic, and cannot be expressed simply as a function of state. The significance of hydraulic hysteresis in geotechnical problems is a subject of increasing research interest. Techniques have recently been established for the inclusion of its effects in constitutive models (see for instance (Gallipoli, Wheeler, and Karstunen 2003; Li 2005). Wheeler and co-workers (Wheeler, Sharma, and Buison 2003) have developed a conceptual model to explain the phenomenon of hysteresis in terms of the structure of the pore space. The combined effects of fabric and bonding will be referred to using the term structure in this paper. There have been considerable advances in the constitutive modelling of natural soils to account for structure, and damage to structure, which a single yield locus elasto-plastic framework (Gens and Nova 1993; Liu and Carter 2000), and more recently, using the framework of kinematic hardening and bounding surface plasticity (Rouainia and
The majority of research and analysis in soil mechanics to date has assumed full saturated conditions, not least because this simplifies the development of analytical, physical and numerical models. Similarly, the mechanics of structured soils is relatively well understood, despite a growing appreciation of the importance of structure in natural soils. In recent years, researchers have begun to look at the modelling of both of these difficult features of real soils, namely partial saturation and structure. Saturated models, for example, cannot predict the sudden collapse of a slope due to rainfall infiltration; and standard elasto-plastic models cannot predict the irrecoverable degradation of strength and stiffness that is observed when natural soils are deformed. The first full constitutive model for unsaturated soils (Alonso, Gens, and Josa 1990), known as the Barcelona Basic Model (BBM), was based on the Modified Cam-Clay (MCC) model, whose isotropic hardening variable is the effective consolidation pressure. The BBM was formulated in the stress space and it included suction hardening. Wheeler and Sivakumar (Wheeler and Sivakumar 1995) suggested that the specific water content is a more appropriate kinematic state variable than the degree of saturation. Houlsby (Houlsby 1997), on the other hand, derived an alternative set of work-conjugate stress variables retaining the more natural kinematic variables. (Bolzon, Schrefler, and Zienkiewicz 1996) developed an unsaturated model, based on a generalised plasticity model for saturated soils, that was formulated in terms of (, s). The advantages of using (rather than net) as a constitutive stress are now widely accepted (Jommi 2000), and several recent studies have pursued this
939
Wood 2000; Kavvadas and Amorosi 2000; Baudet and Stallebrass 2004). These models converge on the MCC model in the limit of total destructuration. They employ an additional isotropic hardening variable, namely the degree of structure to model the enhanced strength of a structured soil, relative to a reconstituted soil at the same void ratio (Leroueil and Vaughan 1990). The combined effects of bonding and partial saturation were considered by (Alonso and Gens 1994). It was predicted that the combined effects of partial saturation and bonding would result in an increase in yield stress that is larger than the sum of the individual contributions. This was confirmed by (Leroueil and Barbosa 2000), who conducted experiments on an unsaturated saprolite from Brazilian gneiss, artificially bonded with cement. The question of how to model this suction-structure interaction by generalizing the definition of the effective pc , has only recently been considered by (Rouainia and Pereira 2008; Koliji, Vulliet, and Laloui 2008). 2 2.1 CONSTITUTIVE MODELLING Choice of stress variables
Interface energy
se 1
Figure 1.
Sr
Water retention curve and interface energy.
The coefficient 2/3 appearing in Equation (3) arises from the account for the effects of skeleton deformation (via porosity changes) on the water retention properties of the soil . 2.2 Constitutive modelling of partial saturation
In a saturated condition, Terzaghi (Terzaghi 1943) suggested that the mechanical response of a soil depends only on the effective stress tensor u (1) where is the total continuum stress tensor, u is the (liquid) pore pressure, and is the identity tensor. Attention is restricted to the small deformation regime, isothermal conditions and the rate-independent behaviour. Thus, the basic elasto-plastic assumption is the additive decomposition of the strain rate, p, into an elastic and a plastic part, e p and , respectively: e p (2)
According to the framework described in (Pereira, Wong, Dubujet, and Dangla 2005), the stress defined by Equation (1) together with the equivalent pore pressure defined by Equation (3) can be used instead of Terzaghis effective stress to extend elastoplastic models to unsaturated states. Starting from Modified the Cam-Clay model, the constitutive stress-strain relations of the model presented in this paper are summarized hereafter. For simplicity, the model formulation is expressed in triaxial stress space. The elastic law is represented by the incremental elastic-strain/stress relations:
e d v e d q
dp v p
dq 2G
(5) (6)
The effective mean pressure is p p u, where p is the total pressure and u is the equivalent pore pressure. Following an energetic approach, (Coussy and Dangla 2002) proposed the following relation, accounting for the energy of the interfaces that separate liquid and gas phases: 2 u uG Sr s U Sr , 3 (3)
e e where v and q are the isotropic and deviatoric components of the elastic strain tensor, is the specific volume, and and G are material constants. The yield function is given by:
f (p, q, p0) q2 M2 p (p0 p) 0
(7)
where is the porosity and U(Sr,) corresponds to the interfaces energy (see Figure 1) and is given by: U (Sr ) s(Sr )dSr
Sr 1
(4)
where M is a material constant and p 0 represents the (generalized) hardening parameter of the model. It accounts not only for the (classical) mechanical hardening arising from irreversible volumetric strains but also for the suction-induced (apparent) hardening.
940
The remaining ingredient to complete the extension of the elastoplastic framework is the definition of the hardening due to suction. This feature, which is a key point in the modelling of unsaturated soil behaviour, accounts for an additional role of suction on the strength of the soil (apart from its contribution to static equilibrium). This function corresponds to the well-known Loading-Collapse (LC) curve of the Barcelona Basic Model (BBM, (Alonso, Gens, and Josa 1990)). The identification of this relation may be achieved by using a direct comparison to the LC curve of the BBM using the following relation: p 0 p 0 uG u (8)
as is the case in the BBM. The choice of the l(s) function should be related to available experimental data. Here, one form considered: k l (Sr ) 1 p ( s Sr se ) atm (13)
This expression is obtained by equating the total preconsolidation pressures expressed in terms of effective stress Equation (1) and net stress pg . p0 is the net preconsolidation pressure as defined in the BBM: p ( s ) 0 p 0 pc c p
( 0 )
where patm is a reference pressure, se is the air entry suction and k is material constant for a given soil. It should be noted that this function may be expressed in terms of the degree of saturation Sr instead of suction s may induce some advantages in terms of numerical modelling (concerning transitions between saturated and unsaturated conditions). In terms of consistency, this approach (already used by several authors, see for instance (Jommi 2000) presents the advantage to include in the arguments of the function l two variables of p the same thermodynamical nature, namely v and Sr which are both generalized strains. The hardening rule is classically given by: d p 0 p 0 v dvp (14)
(9)
where pc is a reference pressure and (s) is given by: (s) (0) [r (1 r) exp( s)] (10)
2.3
Constitutive modelling of structure
where (0) is the slope of the virgin compression line at saturation in the net stress-specific volume plane, r is the fraction of (0) corresponding to the slope of the virgin compression line of the dry material in the net stress-specific volume plane and is a material parameter linked to the rate of change of the soil compressibility with suction. Another way to derive the apparent hardening due to suction is to give an explicit functional definition: p 0
p 0 l (s)
Following (Rouainia and Wood 2000), a scalar parameter R 1 is introduced as a measure of the degree of structure of the material. This parameter is supposed to degrade from its initial value to unity as the material deforms and loses its initial structure. Starting from the saturated preconsolidation pressure of the equivalent remolded soil, the apparent preconsolidation pressure of the structured soil is R p 0 at saturation, and for a given value of suction it is given: R p ( s ) p 0 pc c 0 p
( 0 )
(15)
(11)
where p 0 is the preconsolidation pressure under saturated conditions at the same specific volume. In this case, the identification of the l(s) function should be based on experimental data on the evolution of the apparent yield locus with respect to changes in the degree of saturation (or suction). It is worth noting that other choices are possible for the expression of the preconsolidation pressure as a function of suction. However, it is easy to show that the one given in Equation (11) ensures that d p 0 d p 0 p0 p0 (12)
Other possibilities exist. For instance, if the alternative choices to the LC curve of BBM previously presented in Equation (13)) is used, the following expressions for the effective preconsolidation pressure are obtained (see Figure 2): k ( s Sr se ) p 0 ( s ) R p 0 1 p atm (16)
The rate of degradation of the soil structure is also assumed to be saturation dependent and given by dR k (Sr ) R 1 d d (0) (17)
941
1 0.9 0.8 R0 =5
which shows the three possible (and coupled) mechanisms involved in the generation of plastic strains.
R0 =1 R0 =3
Suction s (MPa)
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0 0 0.5
2.4
Modelling the water retention properties
...... BBM type p0*l(s) tanh
Stress p (MPa)
1.5
2.5
In a first step, the water retention properties of the soil will be modelled using Brooks & Coreys model. This original formulation is modified to account for coupled effect of the deformation of the material on its water retention properties. This is simply performed by introducing the influence of porosity changes upon the water retention properties. A further modification consists in introducing an hysteresis loop. The main drying and imbibition branches are described by the following relations:
D se SrD s D I se SrI s I
Figure 2. Simulated LC loci: coupled influence of suction and structure effects and different choices of the LC curves.
(22)
where dd is a measure of the incremental irreversible deformation (including both volumetric and shear strains): d d
2 (1 A)d vp
(23)
p2 Ad q
(18)
with A a material constant. Without further indication on the effective couplings between structure and partial saturation effects, the structure degradation function, k(Sr), is assumed to have the following form: k (Sr ) k0 p 0
p0
X where se and X (X {D, I}) are dependent on porosity changes. For the sake of simplicity, it is supposed that drying and imbibition branches of the hysteresis loop are similarly affected by the deformation of the skeleton. It thus follows that: X X se ( ) se 0 exp ( 0 )
(24) (25)
X ( ) X 0 exp ( 0 )
(19)
where k0 is a material parameter representative of structure degradation under saturated conditions and is a material parameter that determines the effect of partial saturation on the rate of degradation via the difference between the saturated and unsaturated preconsolidation pressures (power law). The modified preconsolidation stress as defined in Equations (8) and (15) (or alternatively in any of Equations (16)) is then included in the yield function (Equation (7)). Invoking the consistency condition gives: f f f dp dq d p0 0 p q p0 (20)
X X se are material parameters corre0 and 0 sponding to a reference (undeformed) state of the soil (that is to a reference porosity 0), is another X material constant. It should be noted that se 0 is linked to the air entry value (drying branch) or to the trapped air value. The scanning curves (located inside the hysteresis loop) are assumed to follow constant degree of saturation lines.
MODEL PERFORMANCE
Accounting for the hardening function (12), the volumetric part of the plastic deformation increment may formally be written as: dp 1d 2ds 3dR (21)
The ability of the proposed constitutive model to describe important features of the response of structured and unsaturated soils is demonstrated by simulating the compression behaviour of a residual soil extracted from a depth of 8 m from So Paolo. The material parameters of the soil water characteristic curve for hydrological modelling were estimated by back-fitting the published data by (Machado and Vilar 2003) for samples from 8 m depth. A good fit was achieved for se 0.55 kPa and 8.57 (see Figure 3). It should be noted that
942
Figure 3. Water retention curve of So Paolo soil samples from 8 m: experimental and numerical results. Table 1. Model parameters. Parameter (0) M rLC pc se A k(0) p* 0 e0 R0 () () () () (kPa1) () (kPa) (kPa) () () () () (kPa) () () So Paolo soil 0.015 0.055 0.3 N.A. 0.01 0.98 0.001 0.55 8.57 0.0 1.5 0.1 80.0 0.7 2.0
Figure 4. Isotropic compression on So Paolo soil samples from 8 m at two different suctions.
Figure 5. Evolution of the degree of structure during isotropic compression of So Paolo soil samples from 8 m at different suction values.
the hysteresis was not considered here due to the lack of data. A summary of the model parameters, which have been obtained by optimizing the model, is given in Table 1. Isotropic consolidation tests performed both on saturated and unsaturated compressed samples of So Paolo soil (Machado and Vilar 2003) are used to illustrate the capabilities of the proposed model. The unsaturated samples were subjected to matric suctions of 100 and 200 kPa. From Figure 4, it can be seen that the proposed constitutive model is able to predict the increase in the preconsolidation stress with suction. From the experimental results, it is observed that increasing the value of suction leads to an increase in the slope of virgin compression curve, which is not reproduced by the present model.
Figure 5 depicts the evolution of structure during the Isotropic saturated and unsaturated consolidation tests. It can be seen that the smaller the suction is, the quicker the initial structure decays, which is a direct consequence of the volumetric plastic strain development in the samples. 4 CONCLUSIONS
In this paper, a constitutive model is proposed that combines unsaturated and structured behaviour, and is capable of accounting for the phenomenon of volumetric collapse upon wetting and the stiffening effect that suction has on the soil skeletal response. In this model, the effective stress is defined in terms of an equivalent pore pressure which generalise Terzaghis effective stress. Illustrative numerical results for isotropic consolidated
943
unsaturated tests have demonstrated the potential of the proposed model. It should be noted that the formulation of the presented model can easily be extended to the multiaxial stress space. That will be shown in forthcoming paper. REFERENCES
Alonso, E. and A. Gens (1994). Keynote lecture: On the mechanical behaviour of arid soils. In Proc. First Int. Symp. Eng. Char. of Arid Soils, London, UK, pp. 173205. Alonso, E., A. Gens and A. Josa (1990). A constitutive model for partially saturated soils. Geotechnique 40(3), 405430. Baudet, B. and S. Stallebrass (2004). A constitutive model for structured clays. 54(4), 269278. Bolzon, G., B. Schrefler and O. Zienkiewicz (1996). Elastoplastic soil constitutive laws generalized to partially saturated states. Geotechnique 46(2), 279. Coussy, O. and P. Dangla (2002). Approche energtiqu du comportement des sols non saturs. In Maniqu des sols non saturs. Lavoisier: Paris. Gallipoli, D., A. Gens, R. Sharma and J. Vaunat (2003). An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Geotechnique 53, 123135. Gallipoli, D., S. Wheeler, and M. Karstunen (2003). Modelling the variation of degree of saturation in a deformable unsaturated soil. Geotechnique 53(1), 105112. Gens, A. and R. Nova (1993). Conceptual bases for a constitutive model for bonded soils and weak rocks. In Geomechanical Eng. of Hards Soils-Soft Rocks, Athens, Greece, pp. 485494. Gens, A., M. Sanchez and D. Sheng (2006). On constitutive modelling of unsaturated soils. Acta Geotech. 1(3), 137147. Houlsby, G. (1997). The work input to an unsaturated granular material. 47(1), 193196. Jommi, C. (2000). Remarks on the constitutive modelling of unsaturated soils. experimental evidence and theoretical approaches. In Proc. Int. Workshop on Unsaturated Soils, Trento, Italy, pp. 139154. Kavvadas, K. and A. Amorosi (2000). Geotechnique 50(3), 263273. Khalili, N., F. Geiser and G. Blight (2004). Effective stress in unsaturated soils: Review with new evidence. Int. J. of Geomech. 4(2), 115126.
Koliji, A., L. Vulliet and L. Laloui (2008). New basis for constitutive modelling of unsaturated aggregated soil with structure degradation. In Unsaturated Soils: Advances in Geoengineering, pp. 641646. Leroueil, S. and A. Barbosa (2000). Combined effect of fabric, bonding and partial saturation on yielding of soils. In Proc. Asian Conf. Unsaturated Soils, Singapore, pp. 527532. Leroueil, S. and P. Vaughan (1990). The general and congruent effects of structure in natural soils and weak rocks. Geotechnique 40, 467488. Li, X. (2005). Modelling of hysteresis response for arbitrary wetting/drying paths. Computers and Geotechnics 32, 133137. Liu, D. and J. Carter (2000). Modelling the de-structuring of soils during virgin compression. Geotechnique 50(4), 479483. Loret, B. and N. Khalili (2002). An effective stress elasticplastic model for unsaturated porous media. Mech. Mater. 34, 97116. Machado, S. and O. Vilar (2003). Geotechnical characteristics of an unsaturated soil deposit at sao carlos, brazil. In Proc. Int. Workshop on Characterisation and Engineering Properties of Natural Soils, Volume 2, Singapore, pp. 13051321. Nuth, M. and L. Laloui (2008). Effective stress concept in unsaturated soils: Clarification and validation of a unifiedframework. Int. J. Numer. Anal. Meth. Geomech. 32(7), 771801. Pereira, J., H. Wong, P. Dubujet, and P. Dangla (2005). Adaptation of existing behaviour models to unsaturated states: Application to cjs model. Int. J. for Num. and Anal. Meth. in Geomech. 29(11), 11271155. Rouainia, M. and D.M. Wood (2000). A kinematic hardening constitutive model for natural clays with loss of structure. Geotechnique 50(2), 153164. Rouainia, M. and J. Pereira (2008). A constitutive model for unsaturated soils with structure degradation. In 8th World Congress of Computational Mechanics, WCCM8, Venise, Italy. Russell, A. and N. Khalili (2006). A unified bounding surface plasticity model for unsaturated soils. Int. J. Numer. Anal. Meth. Geomech. 30(3), 181212. Terzaghi, K. (1943). Theoretical Soil Mechanics. Wiley: New York. Wheeler, S., R. Sharma, and M. Buisson (2003). Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Geotechnique 53(1), 4154. Wheeler, S. and V. Sivakumar (1995). An elasto-plastic critical state framework for unsaturated soil. Geotechnique 45(1), 3553.
944
Fractals, double porosity and the soil-water characteristic curve

A.R. Russell
Centre for Infrastructure Engineering and Safety, School of Civil and Environmental Engineering, The University of New South Wales, Sydney, Australia
ABSTRACT: A theoretical expression for the soil-water characteristic curve of a double porosity soil is derived. A double porosity soil is modeled as a collection of agglomerated particles, in which a single agglomerate is made up of discrete particles bonded together. Separate fractal distributions for pore sizes around and within agglomerates are defined. The particle size distribution of the double porosity soil is also modeled using a fractal distribution, which may have a fractal dimension very different to those defining the pore sizes. Equating the surface areas of the particles and the pores within the agglomerates enables an expression linking two fractal dimensions to be defined. The curves fit experimental data well. 1 INTRODUCTION 2 FRACTAL DESCRIPTIONS OF POROSITY
Many natural and engineered soils exhibit two distinct pore sets, due to the aggregated nature of the soil, or due to the presence of fissures, root holes or worm holes. This characteristic is referred to as double porosity. Correctly characterising porosity is important as it has a major influence on the soil-water retention properties and is therefore of great interest in understanding water storage in the ground as well as strength variations in engineered infrastructure. There have been numerous attempts to present theoretical expressions for soil-water characteristic curves for single porosity soils in terms of easily measured (or estimated) pore and particle size characteristics which obey fractal theory (Hunt & Gee, 2002; Perfect, 2005; Yu et al., 2009). There have been very few studies which have attempted to provide a theoretically derived expression for the soil-water characteristics of a soil with double porosity, even though Gimnez et al. (1997) concluded that different fractal properties appropriate to different scales are required for more accurate matches between theory and experiment. In this paper a new fractal description of pore size distributions for soils with double porosity (Russell, 2010) is studied. Key parameters describing the pore sizes are linked to particle sizes, which are said to also exhibit fractal characteristics. A theoretical expression for the soilwater characteristic curve is then presented for double porosity soil. The expression and the assumptions used to derive it are checked using experimental data.
There is a mathematical order to particle and pore size distributions in many naturally occurring and engineered soils (McDowell et al., 1996; Coop et al., 2004; Muir Wood, 2007). A fractal geometry implies self-similarity. A volume of soil can be split into smaller volumes, each of which has physical properties which are a reduced-size copy of the original. 2.1 Single porosity
In a soil suppose the probability P of a pore being of size dp is proportional to dp raised to a power: P (d p ) d p
p
(1)
A subscripted p is used to indicate association with pores. Equation 1 is a characteristic of a fractal distribution of pore sizes. The number of pores being of any size L larger than dp is then: N p (L d p ) d p
Dp
(2)
where Dp p 1 is the fractal dimension characterizing the distribution of pore sizes. Now consider a solid material of volume V, in which pores of different sizes are progressively added to create a fractal distribution of sizes, without changing the overall volume (in other words, the volume of solid plus volume of pores is always V). Each pore size is denoted by an order k, numbered upwards from an order 0 for the smallest pore size. The first order
945
pore size added has a size dpk, which also denotes the largest pore size. The total volume of pores of order k is defined to be V. is the ratio between the volume of pores of order k and the total volume V and must be smaller than 1. The total number of pores of order k and size dpk is therefore equal to V/(dpk3) (in which is a geometric shape factor). Now it is supposed that the second order (k 1) of smaller pores added have a size dpk /n1/3, where n represents the ratio between individual pore volumes of successive orders and is a material constant. The total volume of pores of order k 1 is defined to be pV, where p is a material constant representing the ratio between total pore volumes of successive orders. The total number of pores of order k 1 is npV/(dpk3). Continuing this sequence, the total volume of pores from order k to 0 is: Vp V(1 p p2 ) (3)
leading to simple relationships between n, p and Dp: ln p ( D 3 ) / 3 Dp 3 1 or p n p ln n (9)
which, by realising that p must be less than 1, gives: Vp
V 1 p
(4)
The porosity is then equal to:
1 p
(5)
The Menger sponge (Fig. 1) can be described in this way. Consider the initial volume to be a cube of nonporous material with side length 1, or an arrangement of 27 equally sized smaller cubes of side length 1/3. The first order of pores are created by removing seven cubes of side length 1/3 (six from the centres of the faces along with the central cube), so n 27 and 7/27. The remaining 20 cubes are then split into 27 smaller cubes of side length 1/27, and seven are removed using the same pattern, and so on. In the limit p 20/27, Dp 2.73, 1. It is difficult to establish Dp directly by measurement, as this would involve highly sophisticated imagery and analysis of the soil or mercury intrusion tests. However, it is possible to link Dp to particle size characteristics, as particle sizes too obey a fractal distribution in many soils, and the surface area of particles must be equal to the surface area of pores (assuming the sum of the contact areas between particles is negligible compared to the overall surface area). To aid with this, the total surface area of pores of size L larger than dpi is expressed as: S p ( L d pi )
and must be less than 1 p for porosity to be less than 1. The total number of pores of size L larger than dpi (where dpi denotes the size of pores of order i) is: N p ( L d pi ) V ( np )k i 1 1 1 d pk 3 np ( np )2 (6)
V np d pk n2 / 3 1 1 1 2/3 np/n np/n2 / 3
( k i )

(10)
in which is a geometric shape factor.
For np 1 this converges to the solution: N p ( L d pi )
V ( np )( k i ) np np 1 d pk 3
(7)
From Equation 7, when a large number of orders exist, the ratio of the number of pores of order i 1 or higher to the number of pores of order i or higher is np. Recall also that the ratio of the pore size of order i 1 to pore size of order i is n1/3. If the distribution of pore sizes obeys Equation 2 then: N p ( L d pi 1 ) d pi 1 N p ( L d pi ) d pi
Dp
n 1 / 3
Dp
np
(8)
Figure 1. A Menger sponge after the first and second order pores have been added.
946
For np/n2/3 1 Equation 10 converges to: S p ( L d pi )
V np d pk n2 / 3
( k i )
np n2 / 3 np n2 / 3 1
(11)
Again when a large number of orders exists: S p (L d p ) d p

2 Dp
(12)
For a fractal distribution of particle sizes in a soil, we can say that the number of particles being of any size L larger than ds obeys: Ns ( L d s ) d s Ds (13)
where Ds s 1 is the fractal dimension for particle sizes. Ds can be measured directly from the particle size distribution curve. For most soils formed by fragmentation or crushing processes, Ds is a constant between 2.3 and 2.7 (Perfect, 1997; Coop et al., 2004). However, soils formed by other process may have Ds values outside this range. If a fractal distribution exists then the surface area of particles being of size L larger than ds is: Ss ( L d s ) d s 2 Ds (14)
The agglomerates have a size distribution that obeys a fractal distribution of dimension Ds1: and this dimension could be determined from sieving and hydrometer testing as long as the agglomerates are not broken down prior to testing. The particles which make up each agglomerate also have a distribution of sizes that obeys a fractal distribution of dimension Ds2: the dimension that would be measured after sieving and hydrometer testing as long as the agglomerates are broken down prior to testing. Two sets of pores are now defined. One set is the pores around agglomerates, sometimes referred to in other studies as macropores, the other set is the pores within agglomerates, sometimes referred to as micropores. The size distributions of each pore set are assumed to obey fractal theory. The total volume of pores surrounding the agglomerates is: Vp1
1V 1 p1
(17)
Now it is possible to link particle and pore surface areas. For a given soil, in which the minimum pore size is denoted as dpmin and minimum particle size as dsmin, expressions for the total particle and pore surface areas may be equated after substituting the minima d into Equations 12 and 14: d p min
2 Dp
in which a subscript 1 is used to indicate an association with the pores around the agglomerates. The porosity is calculated by assuming pore volume is made up of only the pores around the agglomerates is then 1 1/(1 p1). The total number of pores around the agglomerates of size L larger than dp1i is: N p1( L d p1i )
V ( n1 p1 )( k i ) n1 p1 n1 p1 1 1d p1k 3
(18)
in which n1p1 1. Focusing now on the pores within the agglomerates, their total volume is defined as: Vp 2
d s min2 Ds
(15)
It is reasonable to assume that Equation 15 applies to all soils as long as Equations 2 and 13 are approximately obeyed. This assumption is reasonable and necessary. As long as the ratio between maximum and minimum sizes is large, Equations 2 and 13 will be obeyed very closely. From Equation 15 we can write: d p min d p max 2.2
2 Dp
2V 1 p2
(19)
in which a subscript 2 is used to indicate association with pores within agglomerates. The porosity calculated by assuming pore volume is made up of only the pores within the agglomerates is then 2 2/(1 p2). The total number of pores within the agglomerates of size L larger than dp2i is: N p 2 ( L d p 2i )
d s min d
s max
2 Ds
(16)
2V ( n2 p2 )( k i ) n2 p2 n2 p2 1 2d p2 k 3
(20)
Double porosity
in which n2p2 1. The surface area of the pores within agglomerates obeys the relation: S p2 (L d p2 ) d p2
2 Dp 2
Imagine that a double porosity soil is made up of agglomerated particles. A single agglomerate is a collection of particles that are bonded together, and it is assumed here that the contact areas between particles within agglomerates are very small compared to the total surface area of the particles.
(21)
and, from Equation 16, leads to: Dp 2 2 ( Ds 2 2 ) ln d s 2 min /d s 2 max ln d p 2 min /d p 2 max
(22)
947
in which ds2 indicates true particle sizes (as compared to agglomerate sizes) and dp2 indicates pores sizes within agglomerates. 3 SOIL-WATER CHARACTERISTIC CURVES
in which sa1 is the suction associated with air entry/ expulsion in the pores around the agglomerates and s1max is the suction associated with draining/ filling the smallest pores around the agglomerates. Following a similar procedure, the water retention in the pores within the agglomerates is: 1 D 3 D p 2 3 s p 2 s D 3 2 max D 3 S 2 sa 2 p 2 s2 max p 2 0 for s sa 2 for sa 2 s s2 max for s2 max s (29) in which sa2 is the suction associated with air entry/ expulsion in the pores within agglomerates and s2max is the suction associated with draining/filling the smallest pores within agglomerates. An expression for overall water retention is then easily obtained by scaling Equations 28 and 29 by the volumetric fractions of the pores around the agglomerates and pores within the agglomerates: 1 S S1 S 2 (30)
A procedure similar to that of Tyler & Wheatcraft (1990) is now followed. The incremental number of pores around agglomerates is given by:
N p1 C1Dp1d p1
Dp1 1
d p1
(23)
in which C1 is a positive material constant of proportionality. The incremental pore volume is then:
Vp1 1d p13 N p1
(24)
The water content within the pores around the agglomerates per unit volume of soil, expressed as a function of the capillary diameter within the pores around the agglomerates dc1 and the minimum capillary diameter around the agglomerates dc1min, is: 1 1(dc1 ) V
dc1min d c1
Vp1
(25)
Substituting Equations 23 and 24 into 25 leads to: 1(dc1 ) 1C1Dp1 V (3 Dp1 )
d
3 D p1 c1
dc1min
3 D p1
(26)
The water content when the pores around the agglomerates are saturated, per unit volume of soil, is: S1 1C1Dp1 V (3 Dp1 )
d
3 Dp1 a1
dc1min
3 Dp1
(27)
in which da1 is the capillary diameter in the pores around agglomerates associated with the air entry/ expulsion pressure. It is assumed here, for soil drying, that da1 is equal to dp1max (ie. dp1max represents the characteristic maximum pore size of the continuous network of flow paths made up from smaller pore sizes). It is also assumed that dc1min is equal to dp1min. As capillary diameter is inversely proportional to capillary pressure (suction), an expression for water retention is then easily obtained: 1 D 3 D p1 3 s p1 s D 3 1max D 3 S1 sa1 p1 s1max p1 0 for s sa1
in which 1/(1 2) is a ratio of porosities, where 1 2 is the true overall porosity of the soil. For a single porosity soil it can be assumed that 1. For a single porosity soil, and the case when the minimum capillary diameter is zero so s1max , Equation 30 becomes the fractal description of the classic Brooks & Corey (1964) expression if the residual water content is zero. Figure 2 presents theoretical soil-water characteristic curves for soils which have double porosity. In presenting the curves it has been assumed that s1max and s2max . Kinks in the soil-water characteristic curve are always present when s sa1 and s sa2. The kink at s sa2 is very pronounced when the overall pore space is dominated by the pores within the agglomerates, that is when is slightly larger than 0. It is possible, although not reflected in Figure 2, for a horizontal segment to exist in the soil-water characteristic curve at /s 1 when s1max s sa2.
EXPERIMENTAL VALIDATION
for sa1 s s1max for s1max s (28)
The theoretical expression for the soil-water characteristic curve (Eq. 30), and the fractal assumptions on which it is based, will now be compared with experimental measurements. Pore size distributions must be known to compare experimental results with theoretical curves, and in the cases considered they were obtained using mercury
948
1 0.9 0.8 0.7 0.6
0 .1
0.5 S
0.4 0.3 0.2 0.1 0 0.1 1 10 100 1000
D p 1 2 .1
D p 2 2.9
0 .5
sa 2 100 sa1
0 .9
A compacted bentonite from the Cortijo de Archidona deposit of Almera, Spain has been tested by Villar (2002) and Lloret et al. (2003). The cumulative pore volumes as a function of equivalent pore diameters for dry densities 1.55 Mg/m3 and 1.73 Mg/m3 are presented in Figure 3. Also presented are soil-water characteristic data for soils at dry densities 1.67 Mg/m3, 1.70 Mg/m3 and 1.75 Mg/m3, as is the particle size distribution curve. The data are overlaid by theoretical idealizations of the particle size distribution curve, cumulative pore volume curves and soil-water characteristic curves.
1 0.9 0.8 0.7 0.6
s sa1
0.1
0.5 S
0.4 0.3 0.2 0.1 0 0.1 1
D p1 2.5
D p 2 2 .5
0 .5
sa 2 100sa1
10
0 .9
s sa1
100
1000
Figure 2. Theoretical soil-water characteristic curves for double porosity soils.
intrusion tests. From fractal theory, for a single set of pores, the cumulative volume of pores being of any size L larger than dp obeys:
3 Dp 3 D d dp p p max Vp L d p 3 Dp 3 D d p min p d p max
(31)
in which is the maximum pore volume that may be occupied by the mercury. For a double porosity soil the total cumulative pore volume is separated into parts representing the pore volume around the aggregates (with total volume 1) and pore volume within the aggregates (with total volume 2). The total cumulative volume is then found by simply summing the two, noting the limited range of pore sizes for which each part applies. Soil-water characteristic curves are also needed to conduct comparisons in this study. For each soil it is also necessary to have particle size distributions.
Figure 3. Experimental and theoretical particle size distribution curve, cumulative pore volume curves and soil-water characteristic curves for compacted bentonite exhibiting double porosity (experimental data from Villar, 2002; Lloret et al., 2003).
949
The particle size distribution curve can be matched reasonably well, at least for particles larger than 0.002 mm. The cumulative pore volume curves are also matched well. Continuous lines represent a dry density of 1.73 Mg/m3 and dashed lines represent a dry density of 1.55 Mg/m3. The particle size parameters used include dsmax 0.5 mm, dsmin 1 1019 mm, Ds 2.93. For the soil compacted at 1.55 Mg/m3: 1 0.27, dp1max 30 m, dp1min 0.1 m, Dp1 2.4, sa1 9.7 kPa, s1max 2910 kPa, 2 0.47, dp2max 0.03 m, dp2min 1.5 1012 m, Dp2 2.75, sa2 9700 kPa, s2max 3.24 1015 kPa, and 0.37. For the soil compacted at 1.73 Mg/m3, 1 0.11, dp1max 25 m, dp1min 0.1 m, Dp1 2.4, sa1 11.6 kPa, s1max 2910 kPa, 2 0.46, dp2max 0.05 m, dp2min 1.4 1011 m, Dp2 2.89, sa2 5800 kPa, s2max 2.04 1013 kPa, and 0.19. Although difficult to detect, the theoretical cumulative pore volume curves and soil water characteristic curves contain horizontal segments for both soils at dp1min dp dp2max and s1max s sa2. Using Dp1, sa1, s1 max, Dp2, sa2 and s2 max for dry densities 1.55 Mg/m3 and 1.73 Mg/m3 produces two soil-water characteristic curves that bound (almost) all experimental data for dry densities 1.67 Mg/m3, 1.70 Mg/m3 and 1.75 Mg/m3. Although there is significant scatter in the experimental data, and despite it being for wetting rather than drying, it can be generalized that a soil-water characteristic curve which would best fit the data for 1.67 Mg/m3 would have a larger than that for 1.75 Mg/m3. The observation that denser soils tend to have lower values of is consistent with the theoretically idealized soil-water characteristic curves also presented. As explained by Lloret et al. (2003) and others, an increase in dry density causes a significant reduction in pore volume around agglomerates without significantly changing the pore volume within agglomerates, a characteristic of soils having double porosity. 5 CONCLUSIONS
the agglomerates enables the two fractal distributions to be linked. A theoretical soil-water characteristic curve for a double porosity soil may then be derived. The theoretical soil-water characteristic curve, and the underlying assumptions regarding the distributions of pore and particle sizes, showed promising agreement with experimental data for compacted bentonite. During drying, first air entry occurs when suction is sufficiently large to drain the larger pores around the agglomerates. A kink is observed in the curve at a second air entry value thought to be related to the maximum pore size within agglomerates. REFERENCES
Brooks, R.H. & Corey, A.T. 1964. Hydraulic properties of porous media. Hydrology Paper 3, Colorado State University, Fort Collins, Colorado. pp. 2227. Coop, M.R., Sorensen, K.K., Bodas Freitas, K.K. & Georgoutsos, G. 2004. Particle breakage during shearing of a carbonate sand. Gotechnique. 54:157163. Gimnez, D., Perfect, E., Rawls, W.J. & Pachepsky, Y. 1997. Fractal models for predicting soil hydraulic properties: a review. Engineering Geology. 48:161183. Hunt, A.G. & Gee, G.W. 2002. Application of critical path analysis to fractal porous media: comparison with examples from the Hanford site. Advances in Water Resources. 25:129146. Lloret, A., Villar, M.V., Snchez, M., Gens, A., Pintado, X. & Alonso, E.E. 2003. Mechanical behaviour of heavily compacted bentonite under high suction changes. Gotechnique. 53:2740. McDowell, G.R., Bolton, M.D. & Robertson, D. 1996. The fractal crushing of granular materials. Journal of Mechanics and Physics of Solids. 44:20792102. Muir Wood, D. 2007. The magic of sands: 20th Bjerrum Lecture presented in Oslo 25 November 2005. Canadian Geotechnical Journal. 44:13291350. Perfect, E. 1997. Fractal models for the fragmentation of rocks and soils: a review. Engineering Geology. 48:185198. Perfect, E. 2005. Modelling the primary drainage curve of prefractal porous media. Vadose Zone Journal. 4:959966. Russell A.R. 2010. Water retention characteristics of soils with double porosity. European Journal of Soil Science. doi: 10.1111/j.13652389.2010.01237.x. Tyler, S.W. & Wheatcraft, S.W. 1990. Fractal processes in soil water retention. Water Resources Research. 26:10471054. Villar, M.V. 2002. Thermo-hydro-mechanical characterization of a bentonite from Carbo de Granta. PhD thesis, Universidad Complutense, ENRESA Technical Report 04/2002, Madrid. Yu, B., Cai, J. & Zou, M. 2009. On the physical properties of apparent two-phase fractal porous media. Vadose Zone Journal. 8:177186.
A soil with double porosity may be modeled as a collection of agglomerated particles, in which a single agglomerate is made up of discrete particles bonded together. Separate fractal distributions for pore sizes around and within agglomerates may be defined so the two sets of pores occupy separate fractions of the overall soil volume. The particle size distribution of the double porosity soil may also be modeled as a single fractal distribution, with properties very different to those defining the pore volumes. Equating the surface areas of the particles to the surface areas of the pores within
950
Constitutive model developments for compacted unsaturated fine grained soils based on porosimetry
R. Salgueiro, S. Olivella & J. Suriol
Department of Geotechnical Engineering and Geosciences, Universidad Politcnica de Catalunya, Barcelona, Spain
ABSTRACT: Modeling the mechanical behavior of compacted unsaturated soils against stress and suction changes is complex. One reason is that the structure mostly depends on the initial moisture content and compaction energy applied. Herein, it is proposed an alternative method to model this kind of soils based on its Pore Size Distribution (PSD), as a possible characterization of its structure. This latter can be obtained by means of the Mercury Intrusion Porosimetry test (MIP). The proposed model should permit the prediction of the volumetric deformations produced as a consequence of stress and suction changes. In order to do this, relationships have been developed to be applied to every pore size considered. The integration over all the sizes leads to the calculation of macroscopic values. Those relationships use fundamental physical principles. In order to determinate the stress-strain behavior, it has been considered that volumetric deformation moduli, both for changes in mean stress and suction, are function of pore size. Finally, the preliminary model equations have been applied to simulate suction controlled oedometer tests, including the subsequent samples collapse observed for different values of vertical stress. The investigated soil is a compacted silty clay for which the initial pore size distribution is known. 1 INTRODUCTION from PSD to estimate water and air permeability (Kw and Ka) and the retention curve (SWCC ) for a given stress state [Garca-Bengoechea et al., 1979]. On the other hand, PSD can be very useful to understand the way volumetric deformations take place under stress and suction changes. In this regard, oedometric tests with controlled suction have been simulated. After compression, the sample was saturated and the collapse of the structure took place. The initial PSD and the stress-suction path have been considered. In addition, the retention curve can be obtained, including changes in pore structure as the wetting or drying take place. An example of PSD evolution during compression and wetting is shown in Figure 1 (Thorn et al, 2007). The present paper contains three sections. In the first one, the methodology to determine hydraulic properties is described. Secondly the approach that is used to relate deformations and PSD is presented. Finally, a simulation devoted to simulate collapse phenomena has been done as a first attempt to use the model under these circumstances. 2
Figure 1. Observation of PSD during saturation of a sample compacted at different stress level (Thorn et al, 2007).
Knowledge about description of voids in porous media may be very useful to understand and to model mechanical and hydraulic processes. Porosity Distribution (PSD) can be obtained by Mercury Intrusion Porosimetry (MIP) for a given state of the soil. It is possible to separate between processes related to the underformed microstructure and the processes related to deformation. In the first case, it is possible to use information
PERMEABILITY AND PSD
The estimation of Kw and Ka from soil PSD is based on the assumption that the soil is composed by
951
parallel tubes and the use of the Hagen-Poiseuille and Darcy equations: Q gsr 4 8 (1) (2)
effective stress is considered following the classical assumption that a fraction of the surface of the particle is in contact with water while the remaining surface is in contact with the air [Bishop, 1959]: p i p sSri (5)
q Ki
where Q is the flow rate, r is the radius of the cylinders, i is the hydraulic gradient, is the viscosity, q is the flux, K is the permeability, and g is the specific weight. Combination of these equations and using a distribution of radius, [Garca-Bengoechea et al., 1979] has obtained:
where pi is the effective stress acting on the soil part that is characterized by ri and p is the macroscopic mean stress, which is assumed to be constant. In order to relate pore size and suction, the Laplace equation is considered: seq.i 2Tsw cos ri (6)
K
f (ri )ri 2 g n
8
i 1 N
f (ri )
i 1
(3)
where seq.i is the suction in equilibrium, and ri is the characteristic radius, Tsw is the surface tension, and is the contact angle. It is assumed that two states (fully saturated or fully dry pores) are only possible. Deformations are then calculated as: vi p i Ki (7)
where n is porosity and f(ri) is frequency. This equation can be extended to unsaturated soils using a function to calculate the fraction of void volume that is assigned to each tube depending on the capillary pressure. The degree of saturation of the pore Sri is calculated from capillary pressure or suction s. The simplest case is the two state model (Sri 0 1) but other functions such as van Genuchten [van Genuchten, 1980] could be used assuming that the pore is not perfectly cylindrical. On the other hand, the calculation of Sri and the volume of voids associated to each size permits to calculate a global degree of saturation simply because the amount of water is additive.
where Ki is the stiffness of pores at each characteristic size. If the pores have different sizes and are assumed cylindrical, it is possible to assume the following dependency: r Ki o K o ri
n
(8)
Sr
Sri ei
i 1 N
(4)
ei
i 1
where ei is the void ratio associated to each pore size. The model that is under development should take into account deformations, and this can be done via variations of void ratio in equation (4). 3 VOLUMETRIC STRAIN AND PSD
Deformation by volume changes can be related easily to void ratio variations. Actually, PSD gives information of the volume and its distribution in sizes. In what follows the properties are assumed homogeneous, isotropic and elastic. The distribution of pores is discretized using characteristic sizes ri and its associated volume Vvi. An
where ro is the reference size, Ko the stiffness at that size and n is a shape parameter. The idea is to apply to a PSD at a given stress state a prescribed path in the space p and s which corresponds to a test. Afterwards, calculations for each size can be integrated to obtain a global response of the material. Calculations for each size can be done for deformations, volumetric water content, and permeability. The idea is to try to reproduce results such the ones shown by [Griffiths y Joshi, 1989]. For instance, when suction is changed some classes in the PSD undergo swelling while other parts undergo contraction. We considered first that degree of saturation cut be only zero or one. This assumption leads to no response to suction changes of the class with big sizes, because it is unsaturated (fully dry) and the effective stress remains constant. However, experimental evidences by [Simms y Yanful, 2001; Koliji et al., 2006] indicate that the increase of suction at constant stress leads to macropore reduction and increase of micropores. As compression of pores take place, the volume of each pore reduces and this implies radius reduction.
952
It the radius of the pore reduces, the associated volume changes from one class to another. To take this into account, a transference process is considered. The following assumptions are considered in order to carry out the transference process. The pores are cylindrical with a radius of ri and a length li. This implies that the void volume of a pore is: Vvi ri 2li And the void ratio can be obtained as: ei Vvi ri 2li Vs Vs (10) (9)
where m is a shape parameter and ro and o define reference values in the equation. 4 MODEL APPLICATION
Derivation of these equations leads to the variation of void ratio as: ei 2 rl r 2 i i ri i li Vs Vs (11)
An oedometric test has been simulated using this model. The examples used for model development correspond to MIP of samples from silty clay Campus Nord. The soil was prepared applying a compaction at the dry side. A suction of 1000 kPa was estimated from previous information on this soil. The PSD of the considered soil is bimodal as two main sizes are associated to big void volumes (see below). Figure 2 shows the oedometric results of the test and the simulated ones. The parameters considered for the calculations in Figure 2:
ro 105 nm ro 4 105 nm Ko 105 kPa o 10 kPa n 2.76 m 1.14
If the variation of void ratio is essentially due to radius change, i.e. the second term of equation (11) is neglected the radius variations can be calculated as: ri ei ri 2ei (12)
As there are pore volumes that may change from one class to anther, it is possible that some classes increase its associated volume, while others decrease it. As the model with binary saturation was deemed insufficient in other to obtain a powerful model, retention curve was modified. Following the work by Ramajo et al., [2002] it was proposed to incorporate a more sophisticated function for saturation versus suction at the pore scale level. The proposed relationship contains three ranges: Sri 1 for s seqi Sri 1 s seq.i i for seq.i s seq.i i (13) (14) (15)
The results obtained during compression are acceptable. However, the wetting after compression leads to swelling in the model, while collapse is expected in reality for this stress level. This implies that a further model modification is necessary. In order to reproduce collapse deformations, the equation for deformations has been modified and an additional term to obtain collapse deformations as a function of suction has been incorporated. vi p i s Ki Bsi (17)
where Bsi is a parameter to simulate collapse induced by changes in suction. The swelling deformations were already considered in first term as Bishop effective stress is considered (equation 5).
Tensin media total p, (kPa)
Deformacin volumtrica v (%)
Sri 0 for s seq.i i
100 0.0 0.2 0.4 0.6 0.8 1.0 1.2 1.4 1.6
Parmetros de K i ro = 100,000 nm Ko = 100,000 kPa n = 2.76
Modelo
1,000
10,000
Experiment
The shape parameter i corresponds to each class or size and increases the difficulty to desaturate for larger values, and this happens for smaller pores. The following function is proposed for this parameter, again as an inverse function of the radius of pores: r i o o ri
m
Model
Experimental
Parmetros de i r o = 400,000 nmo= 10 kPa m = 1.14
(16)
Figure 2. Oedometric curve for compression at controlled suction (experimental and simulate results).
953
Again this parameter is considered a function of the radius, in the following way: r Bsi o Bso ri
l
(18)
where Bso is the value corresponding to the characteristic radius ro and l is a shape parameter.
ro 105 nm ro 4 105 nm ro 4 105 nm Ko 105 kPa o 28.7 kPa Bso 104 kPa n 2.97 m 1.0 l 1.42
Figure 4. Calculated compression followed by collapse at different stress levels.
Figure 3 shows the modelling of the test including the collapse process developed during wetting. The following parameters have been considered in the calculations presented in Figure 3: Figure 4 shows the simulation of collapse at various stress levels. Finally, Figure 5 shows deformation points of different tests (three of them are in Figure 4: 300 kPa, 1500 kPa, 4500 kPa) that contain a phase of compression at constant suction (1000 kPa) plus a phase of collapse due to wetting until full saturation. According to experimental results collapse deformation has a maximum of 5% at a vertical stress of about 2000 kPa. After this value, collapse decreases slowly [Gens et al., 1995; Suriol et al., 1998]. Observation of the evolution of the pore size distribution from model calculations it is observed that the variations take place mainly for the larger sizes (Figure 6). This is attributed to the form of the functions considered for deformation. As the evolution of the porosimetry is calculated by means the changes in void volume and a process of void transfer among classes, the retention curve is also evolving during the test simulation. Therefore the resulting retention curve is a combination of a simple equation described above and a
100 0.0 Deformacin volumtrica v (%) 1.0 2.0 3.0 4.0 5.0 6.0 Parmetros de Bsi ro = 400,000 nm Bso = 10,000 kPa l = 1.42 Modelo Ex perimental 1,000 10,000
Figure 5. Calculated deformations obtained from various simulations at different stress level.
Figure 6. Simulated evolution of PSD during compression and collapse deformation for test in figure 3.
model
Figure 3. Comparison of model predictions and experimental results for an oedometric test at constant suction followed by inundation.
distribution of pores which undergoes variations due to deformations associated to changes in suction. Figure 7 shows the shape of the retention curve from the model. As the test began at 1000 kPa, this is the initial point of the curve which covers from this value to full saturation. The way this curve is obtained illustrates the difficulties to obtain retention curves in collapsible soils in reality because the changes in suction that are necessary for retention curve calculations affect
954
deformations significantly and lead to changes in void ratio and its distribution. This latter is crucial for determination of the retention curve. Of course, from a PSD it is possible to determine a retention curve directly assuming no deformations, but the wide range of suction that one needs to consider and the assumption of no deformations are contradictory. In addition to the retention curve, the water and air permeability are obtained using equation 3, but in a two phase condition. This means, that only the pores that are saturated contribute to water permeability while the pores that remain dry contribute to air permeability. The calculated permeability to water and air are represented in figure 8 as a function of suction. The values are relatively high as tortuosity and connectivity, which have not been introduced, could also play an important role. The variations are, however, as expected. Note that the air permeability for dry conditions (high suction) give a value lower than for water permeability at full saturated conditions (low suction) which is due to different density and viscosity. The air permeability does not change very much at the beginning of wetting which is expected as it is controlled by the large pores. In contrast near full saturation, the large pores undergo important changes of water content which affects more significantly the water permeability.
100 95
CONCLUSIONS
This paper presents basic developments for constitutive modelling from PSD curves (porosimetry). The simulation of an oedometric test with real PSD and these preliminary ideas is very challenging. Some parameters of the power equations for compressibility and retention curve equations had to be calibrated to get quantitative results. The preliminary model presented in this paper should be improved in the definition of the functions adopted for parameter dependency on pore radius and on the algorithm to recalculate the PSD curve after each increment of deformation. This later is necessary because when deformation is associated to pores, the volume changes by both length and radius. If a pore with an associated volume changes its radius it may change the class at which it belongs. The model is capable of reproducing collapse deformations via an additional term added to the deformation equation. The response in the case of swelling has not been analyzed in this paper and it probably requires another distribution of pores. It is foreseen through the effective stress considered. In this way the equation for volumetric strain for each pore radius (17) contains two terms that depend on suction. ACKNOWLEDGMENTS The work of R. Salgueiro has been supported by a grant of the Ministerio de Asuntos Exteriores de Espaa. REFERENCES
90 85 80 75 70 65 60 55 50 10 100
Suction s (kPa)
1000
10000
Figure 7. Retention curve (degree of saturation versus suction) calculated during suction variations under constant loading (volumetric deformations take place).
1000
suction (kPa)
100
10
1 1.0E-10
1.0E-09
1.0E-08
1.0E-07
1.0E-06
1.0E-05
1.0E-04
Permeability K (m/s) Kw Ka
Figure 8. Permeability to water and gas as a function of suction calculated during suction variations under constant load (volumetric deformations take place).
Bishop, A.W. (1959). The principle of effective stress. Teknisk Ukeblad, 106(39), 859863. Garcia-Bengoechea, I., Lovell Ch.W. & Altschaeffl, A.G. (1979). Pore distribution and permeability of silty clays, J. Geotech. Eng. Div., 105, 839856. Gens, A., Alonso, E., Suriol, J. & Lloret, A. (1995). Effect of structure on the volumetric behaviour of a compacted soil, Proc. 1st. Int. Conf. Unsat. Soils, Paris. Griffiths, F.J. & Joshi, R.C. (1989). Change in pore size distribution due to consolidation of clays, Gotechnique, 39, 159167. Koliji, A., Laloui, L., Cuisinier, O. & Vulliet, L. (2006). Suction induced effects on the fabric of a structured soil, Trans. Porous Med., 64, 261278. Ramajo, H., Olivella, S., Carrera, J. & Snchez-Vila, X. (2002). Simulation of gas dipole tests in fractures at the intermediate scale using a new upscaling method. Trans. Porous. Med., 46, 269284. Salgueiro Bustillos, Rene Fernando, (2009). Un modelo hidromecnico para suelo fino no saturado & compactado a partir de su porosimetra, Tesina de Mster de Ingeniera del Terreno e Ingeniera Ssmica, ETSECCPB, Barcelona.
955
Simms, P. & Yanful, E. (2001). Measurement and estimation of pore shrinkage and pore distribution in a clayey till during soil-water characteristic curve tests, Can. Geotech. J., 38, 741754. Suriol, J., Gens, A. & Alonso, E. (1998). Behavior of compacted soils in suction-controlled oedometer, Proc. 2nd. Int. Conf. Unsat. Soils, Beijing, 438443.
Thorn, R., Sivakumar, R., Sivakumar, V., Murray, E.J. & Mackinnon, P. (2007). Pore size distribution of unsaturated compacted kaolin: the initial states and final states following saturation, Gotechnique, 57, 469474. Van Genuchten, M. Th. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J., 44(5), 892898.
956
Hydro-mechanical coupling for unsaturated soils

D. Sheng & A.N. Zhou
Centre for Geotechnical and Materials Modelling, The University of Newcastle, NSW, Australia
ABSTRACT: This paper presents an alternative method to couple the hydraulic component with the mechanical component in a constitutive model for unsaturated soils. Some pioneering work on hydromechanical coupling is first reviewed. Generalised constitutive relations on coupled hydro-mechanical behaviour are then introduced. These generalised constitutive relations are incorporated into an existing model, the SFG model, to predict the effects of specific volume on the saturation-suction relationship. The coupled model is validated against experimental data for different soils. 1 INTRODUCTION dependency in a mechanical model that is independent of the hydraulic behaviour. Experimental data generally demonstrate the following points. 1. A SWCC obtained under a higher net mean stress tends to shift towards the higher suction (Ng & Pang 2000; Gallipoli et al. 2003b; Lee et al. 2005; Tarantino & Tombolato 2005). This means that the incremental relationship between degree of saturation (Sr) and suction (s) depends on net mean stress ( p ) or soil density. 2. When the suction is kept constant, isotropic loading and unloading can also change the degree of saturation of an unsaturated soil (Sharma 1998; Wheeler et al. 2003). This implies that the degree of saturation is related to stress or soil density when ds 0. One of the early models that fully couple the hydraulic and mechanical components of unsaturated soil behaviour is that by Wheeler et al. (2003). Other coupled models tend to emphasise the influences of the hydraulic component on the mechanical component (e.g., Vaunat et al. 2000; Sheng et al. 2004; Remero & Jommi 2008). The interaction between the mechanical and hydraulic components in the model by Wheeler et al. (2003) was realised through the use of the average soil skeleton stress and the modified suction. The average * soil skeleton stress ( ij ) is an amalgam of stress, suction and degree of saturation. The modified suction (s*) is a combination of suction and porosity. Therefore, the influence of hydraulic behaviour on the stress-strain-suction relationship is considered via the definition of the average stress. The influence of porosity on the hydraulic relationship between degree of saturation and modified suction is considered via the definition of the modified suction. The model by Wheeler et al. (2003) is one of the few models that are qualitatively tenable
Elastoplastic modelling of mechanical behaviour of unsaturated soils was pioneered by Alonso et al. (1990). Since then, a number of constitutive models have been proposed and reviews of these models can be found in, e.g. Gens (1996), Gens et al. (2006), Sheng et al. (2008b) and Gens (2010). The issue of interaction between the mechanical and hydraulic behaviour was perhaps first raised by Wheeler (1996) and then by Dangla et al. (1997). The first complete model that deals with coupled hydro-mechanical behaviour of unsaturated soils was perhaps due to Vaunat et al. (2000). A number of coupled models soon followed (e.g. Gallipoli et al. 2003a; Wheeler et al. 2003; Sheng et al. 2004). With respect to hydraulic behaviour of unsaturated soils, many models (van Genuchten 1980; Fredlund & Xing 1994) take advantage of the fact that the influence of suction on degree of saturation is more significant than the influence of deformation. The dependency of degree of saturation on suction is described by a Soil-Water Characteristic Curve (SWCC, also called soil-water retention curve). Such a curve is usually obtained under constant stress. Only until recently, the effects of deformation (or stress) on SWCCs have been considered (e.g., Gallipoli et al. 2003b; Sun et al. 2007b; Miller et al. 2008; Nuth & Laloui 2008; Man 2010). As pointed out by Wheeler et al. (2003), the mechanical behaviour of an unsaturated soil depends on degree of saturation even if the suction, net stress and specific volume are kept the same for the soil. Separate treatment of mechanical and hydraulic components in modelling unsaturated soil behaviour has certain limitations in reproducing some experimental observation. For example, it would be difficult to consider the saturation
957
in terms of coupling mechanical behaviour with hydraulic behaviour for unsaturated soils. However, the use of the modified suction and the soil skeleton stress, which is one of the advantages that make the model rigorously consistent in thermodynamics, can become a disadvantage as well, particularly in terms of quantitative prediction and the application of the model. For example, one of the difficulties in using this model is the required constitutive relationship between Sr and s*, which is not commonly available. It is also difficult to experimentally calibrate the synchronised evolution of the loading-collapse surface and the suction-increase and suction-decrease surfaces adopted in the model. In more recent models, the influences of mechanical properties on the hydraulic behaviour are usually modelled via the dependency of the SWCC on soil volume (Vanapalli et al. 1999; Gallipoli et al. 2003b), soil density (Sun et al. 2007b; Man 2010), volumetric strain (Nuth & Laloui 2008), or stress (Miller et al. 2008). Gallipoli et al. (2003b) suggested including a function of specific volume (v) in the SWCC equation of van Genuchten (1980). Sun et al. (2007b) proposed a hydraulic model in the following form: dSr se de ss ds/s (1) where ss is the slope of main drying or wetting curve, se is the slope of degree of saturation versus void ratio curve under constant suction. In theory, ssin the equation above can only be determined from constant-volume tests (de 0). However, such tests are not common. Man (2010) used a similar equation as (1). In his model both air entry value (sae) and the slope of main drying curve (ss) vary with void ratio. Nuth & Laloui (2008) assumed there is an intrinsic SWCC for a non-deforming soil and deformation of the soil can shift this intrinsic SWCC along the suction axis. The shift is governed by an air entry value that depends on volumetric strain. The models discussed above essentially adopt a hydraulic equation of the following form: dSr () ds () dv (2) This equation is not wrong, but the embedded Sr s relationship is for constant volume (dv 0). Therefore, it does not recover the conventional SWCC equations, which are obtained under constant stress. A change in suction at high degrees of saturation can cause a significant volume change albeit a constant net mean stress. It is difficult to obtain a SWCC for a constant soil volume. The vast available data on SWCCs that were obtained under constant stresses would be of limited use in these models. In addition, neglecting the volume change along SWCC can lead to inconsistent prediction of the degree of saturation
during undrained compression, an issue raised by Zhang & Lytton (2008). To overcome the limitations of the above models, a new approach for coupling hydraulic with mechanical behaviour is proposed in this paper. Two general constitutive equations respectively for volume and saturation changes are proposed. These generalised constitutive equations are then incorporated into existing models for unsaturated soils. The predictions of the proposed model are then compared with several sets of experimental data. 2 GENERAL CONSTITUTIVE LAWS
Under isotropic stress states, an increment in the net mean stress ( dp ) and an increment in suction (ds) can both cause volumetric strain (dv): d v A dp B ds (3)
-
where A and B are two general variables (A v /p , and B v / s ) and may be expressed as functions of mechanical variables (such as p , s, e, etc) and hydraulic variables (such as Sr, , etc). These parameters may also depend on the stress and hydraulic paths and can take different values on a loading (drying) and unloading (wetting) paths respectively. As suggested by Sun et al. (2007b) and Man (2010), the degree of saturation depends on suction, void ratio and stress path. Therefore, an increment in the degree of saturation (dSr) can be caused by an increment in suction and/or an increment in volumetric strain: dSr C ds D dv (4)
where C and D are two general variables like A and B (C Sr /s, and D Sr /v). Substituting equation (3) into equation (4) gives: dSr (C DB )ds DAdp Eds Fdp (5)
where E C DB and F DA. A noticeable difference between equation (4) and (5) is that the Sr s relationship in equation (4) is for constant volume and that in equation (5) for constant stress. In drying or wetting tests for determining SWCCs, the volume of the soil is not constant, but the net mean stress is. Therefore, a SWCC corresponds to equation (5) with dp 0: E SrSWCC s (6)
where SrSWCC defines the equation for the soil water characteristic curve. Therefore, it is parameter E, not parameter C, that corresponds to the slope of a SWCC, which is not necessarily a constant.
958
Equations (3), (4), (5) and (6) are general constitutive equations for coupled hydro-mechanical behaviour of unsaturated soils. These equations are generally true and transcend the differences in stress variables and constitutive models used to model the soil behaviour. Indeed, all existing models for coupled hydro-mechanical behaviour can be written in form of these equations. These equations are also valid for both drained and undrained conditions. For the undrained condition (dw 0, with w being the gravimetric water content), we always have eSr wGs const (7)
The function f(s) in equation (9) was set to s, i.e. f(s) s in Sheng et al. (2008a). This is the simplest form possible for f(s). Even with this simplest form, Zhou and Sheng (2009) showed that equations (9) and (10) can predict a large set of experimental data, both in terms of volume change and shear strength. Now the two variables A and B in the volume change equation (3) can be defined as follows: A vp
p f ( s ) ,
B vs
p f ( s )
(11)
where Gs is the specific gravity of soil and is usually treated as a constant. Equation (7) can be rewritten in the following form: dSr (Sr/n)dv (8)
It should be noted that both A and B are usually dual valued, due to elastoplasticity ( and ). 3.2 Soil-water characteristic curve
The constraint defined by equation (8) has to be imposed to deduce the variation of suction for a given variation of stress under undrained condition. However, these equations cannot directly be used to predict soil behaviour. We need to determine at least 4 variables involved: A, B, E and F. To do that, we need to seek the use of specific constitutive laws. 3 SPECIFIC CONSTITUTIVE RELATIONS
Extensive research has been done on the soilwater characteristic curves, first in the field of soil physics and later within geotechnical engineering. Numerous empirical equations have been proposed in the literature (see, e.g., van Genuchten 1980; Fredlund & Xing 1994). Most existing SWCC equations are for constant stress and can thus be directly used to find variable E in the generalised hydraulic model, i.e. equation (5). For example, van Genuchtens equation is defined as:
SrSWCC 1 s a
3.1 Relationship among volume, stress and suction A number of equations are available in the literature to describe the volume change. In this paper we adopt a volume change equation based on the SFG model (Sheng et al. 2008a): d v
(12)
vp
p f (s)
dp
vs ds p f (s)
where a, and are model parameters. In van Genuchtens equation and in that of Fredlund & Xing (1994), no clear threshold suctions like the air entry value are used. These equations are usually used for the drying curve only and hydraulic hysteresis is not included. 3.3 Relationship between saturation and stress
(9)
where vp and vs are soil parameters, and f(s) is a function of suction. Parameter vs must equal vp when the soil is fully saturated and it decreases with increasing suction. The following simple function is used for vs: , vs vp vp ssa s , s ssa s ssa (10)
where ssa is the saturation suction which is the unique transition suction between saturated and unsaturated states (Sheng et al. 2008a). Both vp and vs vary with stress path and can take different values on a loading and unloading paths respectively.
The influence of net mean stress on the degree of saturation is usually reflected by two related phenomena: 1) the SWCC is shifted as the initial void ratio of the soil changes; 2) the change of net stress under constant suction (ds 0) can result in a change of degree of saturation. These two effects are related, but not always equivalent. In some models for coupled hydro-mechanical behaviour of unsaturated soils, it is assumed that the air entry suction is a function of specific volume or volumetric strain (Sun et al. 2007b; Nuth & Laloui, 2008a; Man, 2010). As such, the SWCC is shifted with the air entry value. This approach of coupling mechanical properties with the SWCC is straightforward and simple to use. However, we are not
959
adopting such an approach because of the following reasons: 1. The volume change due to suction change is already included in most SWCC equations, since these equations are obtained from constant stress tests. We need only to consider the volume change caused by stress change when considering the effect of mechanical properties on the SWCC. 2. The intrinsic relationship between the degree of saturation and the volumetric strain must be satisfied. For example, constraints like equation (7) or (8) must hold for undrained tests. Setting the air entry value to a function of specific volume or volumetric strain does not necessarily lead to satisfaction of these constraints. 3. Shifting the air entry value may not sufficiently reflect the effect of the initial void ratio on the SWCC. The slope of the SWCC may also change (Vanapalli et al. 1999; Lee et al. 2005; Man 2010). 4. Common SWCC equations like that by van Genuchten (1980) and Fredlund and Xing (1994) do not use an air entry value. Based on the above reasons, we seek an alternative method to couple the mechanical equation with the hydraulic equation. To achieve the goal, we first study the change of degree of saturation during isotropic compression under constant suction. We also study the changes of suction and volume during undrained isotropic compression. To define the effect of stress on the degree of saturation, it is equivalent to find variable F in equation (5). Using (11), F can be written as: F DA D vp
Specific volume, v
2.4
(a) Preconsolidation
Prediction
2.3
2.2
2.1
2.0
vp = 0.21; vp = 0.06

1.9 10
1.0
100
1000
(b)
0.9
m=0 m=0.05 m=0.1 m=0.2
0.8
m=1.0
0.7
0.6
Prediction
0.5 10 100 1000
Net mean stress, p : kPa
Figure 1. Isotropic loading under constant suction: (a) specific volume vs mean stress, (b) degree of saturation vs mean stress.
between dv and dp . In addition, there are some other constraints that may be useful to determine the unknown variable D in equation (14): 1. For fully saturated soil, the degree of saturation is independent of net stress, i.e., D 0 when Sr 1; 2. D is positive ( dSr 0 when dp 0 , Figure 1). 3. Equation (8) must be satisfied for undrained condition. Replacing the equation (8) into equation (4) leads to: (Sr/n D) dv C ds, dw 0 (15)
p s
(13)
Replacing equation (13) into equation (5) and setting ds 0, we have: dSr Fdp D vp
p s dp
(14)
Equation (14) indicates that the change of degree of saturation (dSr) is related to the change of net stress ( dp ) when suction is kept constant (ds 0). The relationship between dSr and dp shown in equation (14) can be calibrated by isotropic compression tests under different suctions. Figure 1 is a replot from Wheeler et al. (2003) using data by Sharma (1998). It shows the volume and saturation changes of a bentonite-kaolin mixture during isotropic compression at constant suction (s 200 kPa). A comparison between the two data sets in Figure 1a and Figure 1b shows that the relationship between dSr and dp is quite similar to that between dv and dp . This suggests that the constitutive function between dSr and dp be similar to that
The above equation actually defines the variation of suction in terms of volumetric strain under undrained compression. According to the experimental data by Sun et al. (2008), both the volume and suction decrease during undrained isotropic compression of an unsaturated soil, i.e., dv 0 and ds 0. Parameter C in equation (4) is usually negative, because an increase in suction usually results in a decrease in saturation. Therefore, we have: D Sr/n (16)
960
The equal sign (D Sr/n) is valid only if ds 0. To meet the above constraints, variable D is set to the following function: D (Sr/n) (1 Sr)m (17)
where m is a fitting parameter. The last unknown variable F can now be specified as: F Sr S m m vp 1 Sr A r 1 Sr n n ps (18)
The solid curve in Figure 1a shows the calculated results by equations (3) and (11) for isotropic compression under constant suction, and the solid curves in Figure 1b the results by equations (5) and (18). The compression indices for loading and unloading are set as follows: vp 0.21, vp 0.06. The preconsolidation pressure is set to 40 kPa. Parameter m in equation (17) varies between 01.0, with m 0.05 giving the best fit of the data of Sharma (1998). Figure 1 demonstrates that equation (17) captures very well the observed behaviour of the soil during isotropic compression at constant suction. 4 VALIDATION OF THE MODEL
It can also be interpreted as the initial void ratio at the start of the SWCC tests. Equation (21) can be integrated numerically or analytically (for special values of m, e.g. m 0, 0.5, 1) to find the relationship between Sr and e0. Because equation (21) is in an incremental form, integration of the equation requires one specific SWCC that corresponds to a reference initial void ratio. In other words, the conventional SWCC equation is only used for the reference initial void ratio and the new SWCC for a new initial void ratio is obtained by integration of (21). Sun et al. (2007a, 2007b, 2007c) presented a series of isotropic compression tests on statically compacted pearl clay. Pearl clay is an industrial waste material with a moderate plasticity and very little expansive clay minerals. Figure 2 shows the results of isotropic compression tests by Sun et al. (2007a, 2007b). In these tests, pearl clay specimens were compacted to different void ratios (e0 1.171.73). The suction was kept constant during isotropic compression (s 147 kPa). The predicted compression curves are compared with the data in Figure 2a. All predictions are performed using equation (19) and (11) with one set of material parameters: vp 0.2, vp 0.06. The preconsolidation pressures for different initial void ratios are assumed as 25 kPa, 220 kPa, 420 kPa,
1.8
The two key equations of the model are: d v Adp Bds S dSr Eds r (1 Sr )m Adp n S S E B r (1 Sr )m ds r (1 Sr )m de n e (19)
Void ratio, e
e0=1.73
1.5
Test data Predictions
(20)
e0=1.39 e0=1.28 e0=1.17
1.2
Parameters A and B are defined by equation (11). Parameter E refers to the gradient of the SWCC and is defined by equation (6). Both continuous SWCC equations (e.g. van Genuchten 1980; Fredlund & Xing 1994) and those defined piecewisely (Sheng et al. 2008b) or incrementally (Lins et al. 2007; Pedroso & Williams 2010) can be used to find E. The void ratio in equation (20) refers to the value at the current stress and suction. Equation (20) clearly shows that the Sr s relationship for constant value (de 0) is more complex than the SWCC equation (dSr E ds). For constant stress, it is possible to derive the equation that relates Sr to the initial void ratio: Sr S (1 Sr )m r , dp 0 e0 e0 (21)
0.9
(a)
10 100 1000
0.7
Test data Predictions

0.6
0.5
(b)
0.4 10 100 1000
Mean net stress, p: kPa

Figure 2. Isotropic compression tests on pearl clay, (a) void ratio vs net mean stress (b) degree of saturation vs net mean stress (data after Sun et al. 2007a, 2007b).
The void ratio (e0) in equation (21) refers to the initial void ratio at the current stress and at Sr 1.
961
1.1
only one new parameter (i.e. m) introduced in the model. The model is validated against published experimental data.
e0=0.444 e0=0.474 e0=0.514 e0=0.517 Predictions
0.1 1
0.9 0.8 0.7 0.6
REFERENCES
Alonso EE, Gens A & Josa A. 1990. A constitutive model for partially saturated soils. Geotechnique, 40: 405430. Dangla P, Malinsky L & Coussy O. 1997. Plasticity and imbibition-drainge curves for unsaturated soils: a unified approach. In: Numerical Models in Geomechanics (NUMOG VI, Pietruszczak S & Pande GN, eds), Balkema Rotterdam, pp. 141146. Fredlund DG & Xing A. 1994. Equations for the soilwater characteristic curve. Can. Geotech. Journal, 31: 521532. Gallipoli D, Gens A, Sharma R & Vaunat J. 2003a. An elastoplastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Geotechnique, 53: 123135. Gallipoli D, Wheeler SJ & Karstunen M. 2003b. Modelling of variation of degree of saturation in a deformable unsaturated soil. Geotechnique, 53: 105112. Gens A. 1996. Constitutive modelling: application to compacted soils. In: Unsaturated Soils (Alonso EE & Delage P, eds.), Balkema, Rotterdam, vol. 3: pp. 11791200. Gens A. 2010. Soil-environmental interaction in geotechnical engineering, Geotechnique, 60: 373. Gens A., Snchez M & Sheng D. 2006. On constitutive modelling of unsaturated soils. Acta Geotechnica, 1: 137147. Lee IM, Sung SG & Cho GC. 2005. Effect of stress state on the unsaturated shear strength of a weathered granite. Canadian Geotechnical Journal, 42: 624631. Lins Y, Zou Y & Schanz T. 2007. Physical modeling of SWCC for granular materials. In: Theoretical and numerical unsaturated soil mechanics (T Schanz, ed), Springer, Weimer, pp. 6174. Man D. 2010. Predicting the dependency of a degree of saturation on void ratio and suction using effective stress principle for unsaturated soils. Int. J. Numerical & Analytical Methods in Geomechanics, 34: 7390. Miller GA, Khoury CN, Muraleetharan KK, Liu C & Kibbey TCG. 2008. Effects of soil skeleton deformations on hysteretic soil water characteristic curves: Experiments and simulations. Water Resources Research, 44: W00C06. Ng CWW & Pang YW. 2000. Influence of stress state on soil-water characteristics and slope stability. J. of Geotechnical & Geoenvironmental Engineering, 126: 157166. Nuth M & Laloui L. 2008. Advances in modelling hysteretic water retention curve in deformable soils. Computers & Geotechnics, 35: 835844. Pedroso DM & Williams DJ. 2010. A novel approach for modelling soilwater characteristic curves with hysteresis. Computers & Geotechnics, 37: 374380. Romero E & Jommi C. 2008. An insight into the role of hydraulic history on the volume changes of anisotropic clayey soil. Water Resources Research, 44, W12412.
Suction, s: kPa
10
100
1000
10000
Figure 3. SWCCs for specimens compacted at optimum water content (data after Vanapalli et al. 1999).
and 1200 kPa, respectively for e0 1.73, 1.39, 1.28 and 1.17. It is shown that the volume change of the pearl clay can be well predicted by the proposed model. Equation (20) is used to predict the change of degree of saturation during isotropic compression. The fitting parameter (m) is set 0.25. As shown in Figure 2b, the prediction is in a reasonable agreement with the data. Vanapalli et al. (1999) presented a series of drying tests to investigate the influence of stress history on the SWCC of a compacted till. The tested soil is a sandy till, with specimens compacted to different initial void ratios at optimum water content. Figure 3 shows the drying test results of these specimens, with different initial void ratios. The predicted SWCCs, shown as solid curves, are obtained from equation (21). The parameters used in the prediction are as follows: a 65, 1.0, 0.15, and m 0.03. The predicted SWCCs compares very well with the measured data, indicating that the proposed model can capture the effect of initial void ratio on SWCC. 5 CONCLUSIONS
In constitutive modelling of coupled hydromechanical behaviour of unsaturated soils, the mechanical component that defines the stressstrain relation and the hydraulic component that defines the water retention behaviour interact with each other. It is common to link the air entry value to the specific volume or volumetric strain. This paper proposes an alternative method to couple the mechanical component with the hydraulic component. It is based on the fact that the soilwater characteristic curve (SWCC) is obtained under constant stress instead of constant volume. The coupling between the mechanical equation and the hydraulic equation is then realised through an intrinsic relationship between the degree of saturation and the void ratio. There is
962
Sharma R. 1998. Mechanical behaviour of unsaturated highly expansive soil, University of Oxford, UK. Sheng D, Fredlund DG & Gens A. 2008a. A new modelling approach for unsaturated soils using independent stress variables. Canadian Geotechnical Journal, 45: 511534. Sheng D, Gens A, Fredlund DG & Sloan SW. 2008b. Unsaturated soils: From constitutive modelling to numerical algorithms. Computers & Geotechnics, 35: 810824. Sheng D, Sloan SW & Gens A. 2004. A constitutive model for unsaturated soils: thermomechanical and computational aspects. Computational Mechanics, 33: 453465. Sun DA, Sheng D, Cui HB & Sloan SW. 2007a. A density-dependent elastoplastic hydro-mechanical model for unsaturated compacted soils Int. J. Numerical & Analytical Methods in Geomechanics, 31: 12571279. Sun DA, Sheng D & Sloan SW. 2007b. Elastoplastic modelling of hydraulic and stress-strain behaviour of unsaturated soils. Mechanics of Materials, 39: 212221. Sun DA, Sheng D & Xu XF. 2007c. Collapse behaviour of unsaturated compacted soil. Canadian Geotechnical Journal, 44: 673686. Sun DA, Sheng D, Xiang L & Sloan SW. 2008. Elastoplastic prediction of hydro-mechanical behaviour of unsaturated soils under undrained conditions. Computers & Geotechnics, 35: 845852. Tarantino A & Tombolato S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Geotechnique, 55: 307317.
van Genuchten MT. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal, 44: 892898. Vanapalli SK, Fredlund DG & Pufahl DE. 1999. The influence of soil structure and stress history on soilwater characteristics of a compacted till. Geotechnique, 49: 143159. Vaunat J, Romero E & Jommi C. 2000. An elastoplastic hydromechanical model for unsaturated soils. In: Experimental evidence and theoretical approaches in unsaturated soils (A. Tarantino & C. Mancuso, eds.) pp. 121138. Rotterdam: Balkema. Wheeler SJ. 1996. Inclusion of specific water volume within an elsto-plastic model for unsaturated soil. Canadian Geotechnical Journal, 33: 4257. Wheeler SJ, Sharma RS and Buisson MSR. 2003. Coupling of hydraulic hysteresis and stress-strain behaviour in unsaturated soils. Geotechnique, 53: 4154. Zhang X & Lytton RL. 2008. Discussion on A new modeling approach for unsaturated soils using independent stress variables. Canadian Geotechnical Journal, 45: 17841787. See alo the reply by the authors. Zhou AN & Sheng D. 2009. Yield stress, volume change and shear strength behaviour of unsaturated soils: Validation of the SFG model. Can. Geotech. Journal, 46: 10341045.
963
Simulation of static compaction with unsaturated soil/water coupled F.E. analysis

M. Shibata, K. Kawai, S. Kanazawa & A. Iizuka
S. Tachibana S. Ohno
Saitama University, Saitama, Japan Kajima Corporation, Tokyo, Japan
M. Honda
Nikken Sekkei Civil Engineering Ltd., Tokyo, Japan
ABSTRACT: Although compacted soil is widely used for constructing earth structures, the mechanism of compaction has not been explained using principles of soil mechanics. Therefore, designing and constructing such structures depend on empirical methods. In this study, we considered compaction as un-drained compression of unsaturated soil, and have simulated the compaction process using principles of unsaturated soil/water coupling problem. The constitutive model proposed by Ohno et al. (2007) and the soil-water retention characteristic curve model proposed by Kawai et al. (2000) were used for these simulations. When compared with static compaction tests performed by Kawai et al. (2003), the simulation results showed good agreement. Consequently, the shape of the observed compaction curve could be explained. The results achieved in this study can be applied to construction management on geotechnical engineering sites. 1 INTRODUCTION Dry Density (t/m3)
D =100% ( dm ax )
Compa ction Curve

Ze ro A
Compacted soil has low compressibility and permeability, and high shear strength. These characteristics are suitable for construction of various earth structures and compaction is widely used on geotechnical engineering sites. However, empirical methods, such as compaction control by density (Figure 1) and Proctors Method (1933, Figure 2), are still used to design and construct compacted earth structures. This is because compaction has not been explained to date within the framework of soil mechanics. Compacted soil is unsaturated soil. Therefore, unsaturated soil mechanics are needed for elucidating the compaction mechanism. Recently, some constitutive models for unsaturated soil and soil-water retention characteristic curve models were proposed from experimental results. Therefore, a study of compacted soil is a subject in the field of unsaturated soil research. Kawai et al. (2003) conducted static compaction tests on silty clay (Figures 3 and 4). It was observed that compaction of specimens with higher moisture content caused a greater change in suction,
ir
oi
ds
Optimum Moisture Content
Cu
rv
D =90% (Standard)
Water Content (%)

Figure 1. Concept of D-value.
and required a smaller load in order to compact to the same level of dry density as that of specimens with lower moisture content. It was suggested that this behavior could be explained with unsaturated soil mechanics and accordingly the compaction curve shows maximum dry density at the optimum moisture content level.
965
In this study, compaction is regarded as the undrained compression of unsaturated soil and simulations of static compaction tests are performed with unsaturated soil/water coupled finite element analysis.
MATHEMATICAL MODEL USED FOR ANALYSIS
Here, the constitutive model for unsaturated soil and the Soil-Water Retention Characteristic Curve (SWRCC) model are introduced. 2.1 Constitutive model for unsaturated soil applied the effective degree of saturation
Some constitutive models for unsaturated soil have been proposed since Bishop (1960) expressed the effective stress for unsaturated soils. Ohno et al. (2007) expressed changes in stiffness of unsaturated soil with effective degree of saturation. In this model, application of the soil-water retention characteristic curve model can be used independent of the constitutive model. First, effective stress is expressed as follows: net ps1 where net pa1, ps Ses Sr Src 1 Src
2000
(1) (2) (3)
Figure 2.
Procters concept.
192
s pa pw , Se
Height of specimen (mm)
35
190
Suction s (kPa)
30
186 184
25
20 140
150
(a) Height change
Time (min)
160
170
180
190
182 140
Load (N)
150 160 170 180 190
188
1000
(b) Suction change
Time (min)
0 140
150
(c) Load change
Time (min)
160
170
180
190
Figure 3. Compaction on a lower moisture-specimen.

500 400
35
80
Height of specimen (mm)
Suction s (kPa)
Load (N)
70 80
30
60
300 200 100
40
25
20 20 60 70 80 60
0 60 70 80
(a) Height change

Figure 4.
Time (min)
(b) Suction change
Time (min)
(c) Load change
Time (min)
Compaction on a higher moisture-specimen.
966
Here, is effective stress tensor for unsaturated soil, net is the net stress tensor, 1 is the second rank unit tensor, is total stress tensor, s is suction, ps is suction stress, pa is pore-air pressure, pw is pore-water pressure, Sr is degree of saturation, Se is effective degree of saturation, and Src is degree of saturation at s . Volume change of soil at certain water content levels is expressed as: e e0 ln where p p sat (4) (5)
Se
Se 1
0
Figure 5.
100
psat
Se 0
apsat
p
exp [(1 Se)nln a]
Yield surface for unsaturated soil.

100
80 60 40 20
S* rf
Here, a and n are the fitting parameter to express the increase in consolidation yield stress according to de-saturation. The plastic volumetric strain is expressed as follows: p vp ln 1 e0 p sat (6)
(s1, Sr1)
Bo und ary ing d ry ve cu r
80 Srf(Sra) 60 40 20
Srfc
Bo un d
(s1, Sr1)
* Sra
ary
we
Src
Src
tti
ng
cu
The consolidation yield stress can be obtained from equation (6) as shown in the following equation: p p psat exp v MD where MD (7) (8)
rv e
300 400 500 600
100
200
(a) Drying process
Suction s (kPa)
300
400
500
600
100
200
(b) Wetting process
Suction s (kPa)
Figure 6.
SWRCC model proposed by Kawai et al.
1 e0
is the yield stress represented by mean Here, pc effective principal stress, M is q/p at the critical state, and D is the dilatancy coefficient. The following yield function can be obtained by applying equation (7) to the original Cam clay model.
suction history and SWRCC on wetting process does not correspond to that on the drying process. There are innumerable scanning curves. Kawai et al. proposed a SWRCC model which can express hysteresis, with application of similarity appearing for both the drying and wetting processes, respectively (Figure 6). 3 SIMULATIONS OF STATIC COMPACTION
f , , vp MD ln where
p q D vp 0 psat p
(9)
1 3 p : 1, q s : s, 3 2
1 s p1 A : , A I 1 1 3
The static compaction tests conducted by Kawai et al. (2003) were simulated with the soil/water coupled F.E. code (DACSAR-UA), applying the constitutive model for unsaturated soil and the SWRCC model as described above (2.2). 3.1 Analysis condition
Here, I is the fourth rank unit tensor. Equation (9) reduces to the original Cam clay model under saturated condition (Se 1). Figure 5 shows the concept of yield surface expressed by equation (9). 2.2 SWRCC model considering hysteresis
Soil-Water Retention Characteristic Curve (SWRCC) expresses the relationship between suction and soil moisture. It strongly depends on
Figure 7 shows the mesh used for simulation. Here, one-dimensional compression is assumed. The deformation of the bottom is fixed, and that of the right and left side are allowed only in a vertical direction. Drainage is not allowed from the boundaries. Kawai et al. controlled strain to conduct static compaction tests. However, it is difficult to express strain-controlled unloading in simulation. Therefore, loading and unloading
967
Load
Table 1. 0.075 a 10
Material parameter. k 0.010 n 1.0 M 1.333 k (m/day) 0.01 v 0.33 m 0.8
2 cm
m: Mualems modulus (1976).
Undrainage boundary 5 cm
1.23 Void ratio e (-) 1.2
Figure 7.
Mesh for simulation.
lower higher Water content
300
1.15
Load (kPa)
200
100
1.1 30
Figure 10.
: w=16% : w=24% : w=32% : w=40% 50 100 log p' (kPa) 500
0 Loading-50min Time t (min)

Figure 8.
Changes in void ratio during compaction.
Unloading-5min
Loading condition of compaction.
Srf=1.00
Dry
ing: A Wett
0.8
Logistic curve Eq. Srf - Src Sr= +S 1+exp(A+Blns) rc
value in relation to the initial degree of saturation on the primary wetting curve. Therefore, the distribution of initial water head in the vertical direction is unsteady. As seen in Figure 8, a leaving time is allowed before the loading so that the water head can reach equilibrium. In addition, cyclic compaction is also simulated. 3.2 Behavior of specimen under compaction Figure 10 shows the relationship between void ratio and mean effective stress. As shown in equation (2), suction stress is the product of an effective degree of saturation and suction. Consequently, the initial effective stress calculated from the primary wetting curve is greater on the specimen with higher water content. Specimens with lower water content, however, show higher yield stress with lower compressibility. Figure 11 shows the relationship between suction and the degree of saturation on the water characteristic curve during compaction. During the loading stage, the degree of saturation increases due to compression with a concomitant decrease in suction. On the other hand, during the unloading stage, the degree of saturation decreases due to expansion and the suction increases. These relationships, between the degree of saturation and suction, are dominated by SWRCC. Figure 12 shows the changes in suction during compaction. It is found that suction changes more drastically in specimens with higher water content. This tendency corresponds to the
ing: A=-
0.6
,B 34.7
=-24.0
=5.9
0.4
, B=4 .6
0.2
Src=0.15
0 0 200 400 600 800
Suction (kPa)
Figure 9. Water characteristic curve.
during compaction are expressed by increasing and decreasing load on the top of the specimen. Figure 8 shows the loading condition in the simulations. Figure 9 shows water characteristic curves used for simulation and Table 1 summarizes the material parameters. Initial void ratio of 1.20 and initial water content, every 2%, from 10 to 42%, are shown. The initial degree of saturation can be calculated from the void ratio and the water content. Initial suctions of all nodes are assumed to take a
968
1 Degree of saturation Sr (-)

Dry density d (g/cm3)
1.3
w=40% 0.8 0.6 0.4 0.3 50 100 Suction s (kPa) 200 w=32% w=24% w=16% 250
Zero air voids curve 1.28 1.26 1.24 1.22 1.2 0 10 20 30 40 Water content w (%) 50
Figure 11. Relationship between suction and degree of saturation during compaction.
Figure 13. Relationship between water content and dry density after compaction.
Suction s (kPa)
200
w =24% w =32% w =40%
250
1.29
w =16%
1.28 Drainage 1.27 Undrainage 1.26 39 40 41 42 Water content w (%)

of 42% water
100 50 Loading-50min Unloading-5min Time t (min)
43
content
Figure 12.
Changes in suction during compaction.
Figure 14. Compaction specimen.
static compaction test conducted by Kawai et al. Moreover, post-compaction suction is higher than the initial suction. Figure 13 shows the relationship between water content and post-compaction dry density. It is obvious that dry density reaches a peak value at a certain water-content as seen in the actual compaction curve. On compaction of the 42% water-containing specimen, the value of suction fell into the negative range (indicating the occurrence of positive porewater pressure). It is known that leakage occurs from soil mass on actual compaction. Therefore, zero total water heads were applied to all boundaries to satisfy the condition of drained boundaries when a positive pore-water pressure occurred. Here, since the drainage could not be calculated, the value of dry density was shifted horizontally to the zero air voids curve (Figure 14). This result is expressed by the open square in the plot. The dry density, on actual compaction, is represented by the space between open triangle (simulated under undrained condition) and square (simulated under drained condition) plots.
The shape of the compaction curve, obtained from simulations, is similar to the shape of the actual compaction curve. For example, the maximum dry density appears at the optimum water content and the gradient changes in the drier side of the optimum water content although there are some differences from actual compaction curve such as, the optimum water content is relatively high and the compaction curve bends at a sharp angle. It is obvious that we can express the shape of compaction curve qualitatively with the specific yield surface of unsaturated soil and hysteresis of SWRCC. 3.3 Effects of cyclic compaction
It is known that cyclic compaction increases dry density and strengthens soil. However, in classical saturated soil mechanics, such cyclic loading means cyclic stress changes in the elastic region and the final dry density is not dependent on the number of cyclic compaction loads. Here, the effect of the cyclic compaction is considered in the framework of unsaturated soil mechanics.
969
Figure 15 shows loading conditions of cyclic compaction via the relationship between time and load. Figure 16 shows changes in the void ratio. It is found that the volumetric strain accumulates due to cyclic compaction, particularly at a higher water content level. This result is explained by SWRCC (seen in Figure 17). The degree of saturation
300
Load (kPa)
200
100
0 Loading-50minUnloading-5minLeave for 20 min 5 Time t (min)

Figure 15.
1.22 1.2 Void ratio e (-) w =16% w =24% 1.15 w =32% 1.1 1.08 Loading-50minUnloading-5minLeave for 20 min 5 Time t (min) w =40%
increases while the suction decreases on loading, and the degree of saturation decreases and the suction increases on unloading. These relationships depend on the wetting and drying paths of SWRCC, respectively. Here, both the degree of saturation and suction gradually increase due to cyclic compaction since the hysteresis of SWRCC is considered (Figure 18). Saturation and suction conditions after the loading and unloading cycles, therefore, differ from the previous one. Consequently, the accumulation of strain, due to cyclic compaction, could be expressed without the use of any sophisticated model, such as the subloading surface model proposed by Hashiguchi (1980), used to express elasto-plastic strain in elastic region. Figure 19 shows a comparison of compaction curves originating from simple and cyclic compaction dependent on the number of cyclic compaction. The difference is particularly visible in the region where there is higher water content. However, it should be noted that the difference in the
Loading condition of cyclic compaction.

270 Suction s (kPa) w=16% w=24% w=32% w=40%
200
100 70 Loading-50minUnloading-5minLeave for 20 min 5 Time t (min)
Figure 16. Changes in void ratio during cyclic compaction.
Figure 18. Changes compaction.
in
suction
during
cyclic
1 Degree of saturation Sr (-)

1.3
w=40% 0.8 0.6 0.4 100 200 Suction s (kPa) w=32%
1.28 1.26 1.24 1.22 1.2 0
: Simple compaction : Cyclic compaction
w=24% w=16% 300
10
20 30 40 Water content w (%)
50
Figure 17. Relationship between suction and the degree of saturation during cyclic compaction.
Figure 19. Relationship between water content and dry density after compaction.
970
compaction curves is dependent on the hysteresis of SWRCC, and simulation results can be strongly influenced by the SWRCC model. 4 CONCLUSION
study is expected to be helpful for construction and maintenance of compacted earth structures. REFERENCES
Bishop, A.W. 1960. The principle of effective stress, Norwegian Geotechnical Institute, 32: 15. Hashiguchi, K. 1980. Constitutive equations of elastoplastic materials with elastic-plastic transition. Journal of Applied Mechanics, ASME, 47: 266272. Kawai, K., Karube, D. & Kato, S. 2000. The model of water retention curve considering effects of void ratio, Proc. Asian Conf. on Unsaturated Soils: 329334. Kawai, K., Nagareta, H., Hagiwara, M. & Iizuka, A. 2003. Suction changes of compacted soils during static compaction test, Proc. of 2nd Asian Conf. on Unsaturated Soils: 429434, 2003. Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media, Water Resources Research, 12(3): 514522. Ohno, S., Kawai, K. & Tachibana, S. 2007. Elasto-plastic constitutive model for unsaturated soil applied effective degree of saturation as a parameter expressing stiffness, Journal of JSCE, 63(4): 11321141 (in Japanese). Proctor, R.R. 1993. Four articles on the design and construction of rolled-earth dams, Eng. News Record, 111: 245248, 286289, 348351, 372376.
Compaction was simulated with the soil/water coupled F.E. code (DACSAR-UA), to which the constitutive model for unsaturated soil and soil-water retention characteristic curve model was applied. The simulations were successful in expressing the compaction behavior by assuming compaction as the un-drained compression of unsaturated soil. Moreover, the effect of cyclic compaction was explained from the hysteresis of SWRCC. In this study, the initial void ratio was assumed to be the same as that of specimens with different water content. It is generally true that low moisturespecimens keep a high void ratio. Therefore, the effects from an increase in the yield stress and initial void ratio appear in the actual compaction. Future work will need to focus on how to express the initial conditions. In this study, the compression of air within the soil has been neglected. Taking account of these conditions will enable us to better understand more practical aspects of compaction. This
971
Elastoplastic modelling of hydraulic and mechanical behaviour of expansive soils

D.A. Sun & W.J. Sun
Shanghai University, Shanghai, China
ABSTRACT: The Barcelona Basic Model (BBM) cannot predict the mechanical behaviour of unsaturated expansive soils, and the Barcelona Expansive Model (BExM) for unsaturated expansive soils is complicated and the micro parameters and the coupling function from micro-structural strain to macrostructural strain are difficult to be determined. This paper presents an elastoplastic constitutive model from the macroscopic observation for predicting the hydraulic and mechanical behaviour of unsaturated expansive soils, based on an existing hydro-mechanical elastoplastic model for non-expansive unsaturated soils. The model predictions are performed on oedometer tests on compacted unsaturated sand-bentonite mixture. The tests include swelling tests under a constant net vertical stress and one-dimensional compression tests under different constant suctions. The comparisons between measured and predicted results indicate that the model can quantitatively predict the hydraulic and mechanical behaviour of unsaturated expansive soils. 1 INTRODUCTION The model takes into consideration the coupled effect of the degree of saturation on the mechanical behaviour and void ratio on the water-retention behaviour. 2 COUPLED MODEL FOR UNSATURATED EXPANSIVE SOILS
Starting with the Barcelona Basic Model (BBM) proposed by Alonso et al. (1990), several elastoplastic constitutive models for unsaturated soils have been developed, which are based on experimental and theoretical studies. These models can reproduce most basic mechanical response of non-expansive unsaturated soils including collapse phenomenon. However, it cannot predict the mechanical behaviour of unsaturated expansive soils. Gens & Alonso (1992) presented a framework for describing the mechanical behaviour of unsaturated expansive soils. Alonso et al. (1999) presented the Barcelona Expansive Model (BExM) in which the deformations of two level structures (microstructure and macrostructure) were considered. The mechanical behaviour of the macrostructure follows the BBM, and that of the microstructure is adopted from the framework proposed by Gens & Alonso (1992). The BExM is complicated and the micro parameters and the coupling function from micro-structural strain to macro-structural strain are difficult to be determined. Meanwhile, the constitutive models can only predict the strength and stress-strain behaviour without incorporating the water-retention behaviour. This paper develops an elastoplastic constitutive model from the macroscopic observation for predicting the hydraulic and mechanical behaviour of unsaturated expansive soils, based on an existing hydro-mechanical elastoplastic model for non-expansive unsaturated soils (Sun et al. 2007).
Based on test results of non-expansive and expansive unsaturated soils (Sun et al. 2000; Zhan 2003), the changes in the compression curves with increasing suction are assumed in Fig. 1. That is, the compression index of non-expansive unsaturated soils increases with increasing suction, i.e. (s) (0), while the compression index of expansive unsaturated soils decreases with increasing suction, i.e. (s) (0). In Fig. 1, p is an isotropic net stress, is a swelling index for saturated and unsaturated soils, pn is an isotropic stress at which no swelling or collapse takes place when suction is reduced. 2.1 Stress-state variables and water retention behaviour The stress-state variables employed in the constitutive model are the average skeleton stress tensor ij and the suction s, and the strain state variables are the soil skeleton strain tensor ij and the degree of saturation Sr. The tensor ij is defined by
ij ij ua ij Sr s ij
(1)
973
ei ei(s)
ei(0) en pi
1 1
(0)
(s)
sr Sr
1 Wetting e1 e2 e1>e2 s
1 sr Drying
ln (p+sSr) pn (a) Non-expansive soils
ei ei(0)
1
ln s
Figure 2. Water-retention curves of expansive soils at different void ratios.
ei(s) en pi
1
(0) (s)
1 s=0 s2 s1
1 1
ln(p+sSr) pn (b) Expansive soils

en
p0y p'2y p'1y
(0)
Figure 1. Compression curves for non-expansive and expansive soils.
(s)
where ij is the total stress tensor, ua the pore-air pressure, and ij the Kronecker delta. The water-retention curves of unsaturated expansive soils at different void ratios are assumed in Fig. 2, which are similar with those of nonexpansive unsaturated soils (Sun et al. 2007). The water- retention curves can be rewritten as ds d Sr se d e s (2)
p'n
lnp'
Figure 3. soils.
Sketch of compression curves of expansive
where se is the slope of the Sr-e curve under constant suction larger than the air-entry value, and
tions. The initial yield stress will change with the imposed suction, and the Initial Yield Curve (IYC) is a curve in the s-p plane, which expresses the change in the initial yield stress with the suction. IYC can be defined by p y p0 y 0[(1 r )e s r ] p n p n
0

2.2
s for scanning curve sr for main drying/wetting curve

Initial Yield Curve (IYC)
(3)
(4)
Figure 3 is the sketch of compression curves of expansive soils. (0), (s1) and (s2) are the slopes of normal compression lines of saturated soil and unsaturated soils with suctions s1 and s2, respectively, p0y, p1 y and p2 y are the initial yield stresses for saturated soil and unsaturated soils with suctions s1 and s2. It can be seen from Fig. 3 that the initial yield stress increases with increasing suction, which has the same tendency with the LC curve in the BBM. However, the compression curve is located down with increasing suction, which is different from the BBM. The specimens with the same initial state are wetted or dried to different target suctions and then are compressed under different constant suc-
where py is the initial yield stress for unsaturated expansive soil with suction s; pn is an isotropic stress at which no swelling or collapse occurs when suction is decreased; r and are the material parameters. The IYC is used to judge whether the specimen reaches the yielding state. When the stress state is inside the IYC, the specimen is at the elastic state; when the stress state reaches the IYC, the elastoplastic deformation takes place and can be obtained by using the concept of the Equivalent Void Ratio Curve (EVRC) in the next section. 2.3 Equivalent Void Ratio Curve (EVRC) A line is drawn at the elastoplastic stage of compression curves in Fig. 4(a). The plastic void ratios
974
ep
ep(0) ep(s 2) ep(s 1) s=0 s2 s1 1
(s)
(a)
1
(0)
de
dp p
(9)
Substituting Eq. (9) into Eq. (8) and arranging gives

N
en
0 p'2 p'1 p0y p'2y p'1y p'4 p'3
dp
dp (Sr )ds p p se s
(10)
lnp'
p'n
s
s1
EVRC(s 1)
IYC
(b)
EVRC(s 1)
s2
EVRC(s 2) EVRC(s 2)
When the initial average skeleton stress p pnet s1Sr with suction s1 is less than the corresponding initial yield stress py, the strain increment consists of the elastic strain increment caused by the change of the average skeleton stress (p: p1p2) and the plastic strain increment caused by the change of suction (s: s1s2), as shown in Fig. 4. The elastic volumetric strain increment caused by p is given by
e d v
p'2
p'1 p0y p'2y p'1y p'4
p'3
p'
Figure 4.
Sketch of the model for isotropic stress state.
dp (1 e ) p
(11)
at different suctions are the same in the line. It can be seen that the average skeleton stress increases with decreasing suction. In order to describe this trend, the concept of Equivalent Void Ratio Curve (EVRC) is introduced when the stress state is inside the IYC, the EVRC is the line paralleled to p axis at suction s; when the stress state reaches the IYC, the EVRC can be defined by p0 y ( s ) py p n p n
( 0 )
The plastic volumetric strain increment caused by s is given by de p e p ( s2 ) e p ( s1 ) ( s2 ) ln p y ( s2 ) p n ( s1 ) ln p y ( s1 ) p n (12) Given the increment of the net stress and suction, the volumetric strain increment can be calculated by Eqs. (10), (11) and (12), and the increment in the degree of saturation Sr can be calculated by dSr (dp dp Srds)/s (13)
(5) (6)
(s) (0) [(1 ) e()s ]
where and are the material parameters for identifying the change of (s) with suction s. 2.4 Coupled model for isotropic stress state A simple constitutive model for isotropic stress states is presented here. Differentiating Eq. (1) gives the increment of the average skeleton stress dp dp Srds sdSr (7)
After the initial state reaches IYC as shown in Fig. 4, the plastic strain increment caused by the changes in average skeleton stress (p: p3 p4) and suction (s: s1 s2) is described by d vp
(0) dp0 y
(1 e ) p0 y
(14)
where p and p are the isotropic net stress and average skeleton stress. Substituting Eq. (2) into Eq. (7) gives ds dp dp Sr ds s se de s (8)
From Eq. (5), dp y p y p0 y dp0 y p y s ds (15)
Substituting Eq. (15) into Eq. (14) gives

p dv
where de is the elastic deformation under constant suction, that is
(0) (1 e ) p0 y
p y dpy ds s
p y p0y
(16)
975
During the elastoplastic deformation stage the increments in py is given by differentiating Eq. (1) according to dp y dpy Sr ds sdSr (17)
where py is an isotropic net stress in the elastoplastic range. Substituting Eqs. (2), (11) and (16) into Eq. (17) and arranging gives dp y where p y A se s( (0 ) ) p0 y p0 y B Sr A py s (19) dpy Bds (1 A) py se s p y (18)
(20)
initial suction s0 (about 2500 kPa) were first wetted to respective targeted matric suctions (i.e. 300 kPa, 600 kPa, 1200 kPa and 1500 kPa) under a constant vertical net stress (10 kPa). After the wetting tests, the specimens were compressed to the maximum vertical net stress of 2685 kPa under different constant suctions. Figure 5 shows the water-retention curve of sand-bentonite mixture measured by using the pressure plate apparatus and the filter papers at the ranges of lower and higher suctions, respectively. It can be seen the air-entry value of the sandbentonite mixture ever wetted to saturation is about 500 kPa. Figure 6 shows the Sr-e relationship during compression tests under different constant suctions. It can be seen that the relationship between the degree of saturation and void ratio can be considered to be linear and the gradients of the lines are almost the same at different suctions. Sun et al. (2009) and Sun (2009) show the more details about the laboratory tests.
Sr /%
Given the increment of the net stress and suction, when dpy 0, volumetric strain increment can be calculated by Eqs. (11) and (16) with Eq. (18); when dpy 0, volumetric strain increment can be calculated by Eqs. (11) and (12) with Eq. (10), and the increment in the degree of saturation can be calculated by dSr (dpy dpy Sr ds ) / s (21)
100
Pressure plate Filter paper
80
60
Dr W e tt in g
yi ng
The model for the isotropic stress can be extended to general stress by following Sun et al. (2007). 3 MODEL PREDICTIONS VERSUS EXPERIMENTAL RESULTS
40
20
10
100
s/kPa
1000
10000
Sun et al. (2009) have performed a series of laboratory tests on heavily compacted sand-bentonite mixture using a suction-controllable oedometer for unsaturated soils. Kunigel-V1-Na-bentonite (produced in Yamakuchi Prefecture, Japan) and Toyoura sand (a Japanese standard sand for soil testing) were used as testing materials, and were mixed with 3:7 by weight. The Kunigel bentonite contains about 48% Na-montmorillonite component and the others of quartz, feldspar, calcite, and pyrite, etc. The physical properties are D50 0.2 mm, and Uc 1.3. The tested sandbentonite mixture belongs to high expansive soil. Four tests named Test No.1No.4 were performed to measure the hydraulic and mechanical behaviour of the sand-bentonite mixture. The No.1No.4 specimens with almost the same initial void ratio varying from 0.55 to 0.56, initial water content varying from 9.2% to 10.0% and
Figure 5. Measured water-retention curve of sandbentonite mixture.
75
No. 1 2 3 4
s/kPa 300 600 1200 1500
Sr/%
60
45
0.5
0.6
Figure 6. Relationship between degree of saturation and void ratio during compression under different constant suctions.
976
In the following section, the model predictions are performed on the oedometer test results including the swelling tests at constant vertical net stress and one-dimensional compression tests under different constant suctions, using the above hydro-mechanical coupled elastoplastic model for unsaturated expansive soils. 3.1 Model parameters and their determination
0.65 No.1 No.2
After wetting No. s/kPa Exp. Pred. 1 300 2 600 3 1200 4 1500
0.6
No.3
No.4 0.55 500 Suction reduction 1000
The proposed constitutive model for unsaturated expansive soils requires 12 model parameters which can all be determined from element tests. The parameters (0) and can be obtained from the result of isotropic compression test on saturated soil with a loading-unloading-reloading cycle. The four parameters related with the IYC are r, , p0y and pn and two parameters about the compression index (s) are and , which can be determined from the results of isotropic compression tests at different constant suctions. There are three parameters to describe the water retention behaviour, i.e. sr, s and S0rw, which are determined by measured water-retention curve in Fig. 5. The parameter se reflects the hydraulic and mechanical coupling behaviour, and is determined by plotting e against Sr from results of above isotropic compression test on unsaturated soil at constant suctions, as shown in Fig. 6. Based on the experimental data, the relevant model parameters used for predicting the stressstrain relation and water retention behavior of sand-bentonite mixture were determined as follows: (0) 0.14, 0.02, r 0.25, 0.112, p0y 550 kPa, pn 2100 kPa, 0.467, 0.0042 kPa1, sr 0.15, s 0.03, S0rw 56.2%, se 0.5. The parameters sr, s and S0rw can be determined by the water-retention curve in Fig. 5 and se by the Sr-e relationship in Fig. 6. 3.2 Wetting tests under constant net stress
(a) e - s
s/kPa
60 No.1 55
After wetting No. s/kPa Exp. Pred. 1 300 2 600 3 1200 4 1500
Sr/%
50 45 40 500 No.2 No.3 No.4 Suction reduction 1000 s/kPa (b) Sr - s
Figure 7. Changes in void ratio and degree of saturation during wetting from initial state to targeted suctions.
Figure 7 shows the changes in void ratio and the degree of saturation before and after the wetting tests and the predicted changing process during wetting. Because the initial suction s0 is rather large, the average skeleton stress p of the initial state calculated by Eq. (1) is larger than the initial yield stress py. That is, the initial state is located outside the IYC. During the wetting test, the average skeleton stress is reduced. It belongs to the unloading process. The strain increment consists of the elastic volumetric strain increment caused by the decrease in the average skeleton stress and the plastic volumetric strain increment caused by the decrease in the suction, and can be calculated by Eqs. (11) and (12) with Eq. (10), respectively.
The changes in the saturation degree can be calculated with Eq. (13). From results of the wetting tests in Fig. 7, void ratio and the degree of saturation increase with decreasing the suction and the changes in void ratio and the degree of saturation before and after the wetting test increases with the range of suction variation. Simultaneously, the upward tendency of degree of saturation of Test No.1 is more obvious than that of other three. From the predicted results in Fig. 7(b), it can be seen that at the initial stage of suction decreasing the changes in the saturation degree trend to be the same for the four tests, while when the suctions decrease at smaller than 600 kPa, the increase tendency of the saturation degree of Test No.1 is more obvious. It can be analyzed from the water-retention curve shown in Fig. 5. When the suction decreases to 600 kPa during the wetting, the wetting curve changes from the scanning curve to the main wetting curve. The comparisons between predicted (Pred.) and measured (Exp.) results indicate that the coupled model provides good description of the changing process of void ratio and the degree of saturation during wetting at constant net stress.
977
3.3
Compression tests under constant suctions
Figure 8 shows the comparisons of predicted and measured results of the changes in void ratio (Fig. 8a) and degree of saturation (Fig. 8b) in onedimensional compression tests under constant suctions of 300 kPa, 600 kPa, 1200 kPa and 1500 kPa. When the net stress is small, the average skeleton stress p is less than the initial yield stress py on the IYC. The volumetric deformation belongs to the elastic stage. As the vertical net stress gradually increasing, the average skeleton stress is larger than the initial yield stress, the elastoplastic deformation takes place. The elastic and plastic volumetric strain increments can be calculated by Eqs. (11) and (16) with Eqs. (10) and (18), and the increment of degree of saturation can be calculated by Eqs. (13) and (21) with Eqs. (10) and (18). From the predicted results shown in Fig. 8 (a), it can be seen that the specimens behave from the elastic state to the elastoplastic state. With the increase in suction, the yield stress of the specimen gradually increases, and the compression index gradually decreases. It shows that the predicted results have the consistency with the test results. At the same time, the coupled model can also predict that the degree of saturation increases with increasing the vertical net stress even at constant suctions, which can be seen in Fig. 8 (b).
So the coupled model can predict not only the deformation but also the water-retention behavior of unsaturated expansive soils. 4 CONCLUSIONS
This paper presented a coupled hydraulic and mechanical elastoplastic constitutive model for expansive unsaturated soils, from the macroscopic observation based on the existing hydromechanical elastoplastic model for non-expansive unsaturated soils. The model takes into consideration the coupled effects of the degree of saturation on the mechanical behavior and void ratio on the water-retention behavior, and can predict the hydraulic and mechanical behaviour of expansive unsaturated soils. The model predictions are performed on oedometer tests on a compacted sandbentonite mixture. The tests include the swelling tests under constant vertical net stress, and onedimensional compression tests under different constant suctions. The comparisons between the measured and predicted results indicate that the model can quantitatively predict the hydraulic and mechanical behaviour of unsaturated expansive soils. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Geotechnique, 40(3): 405430. Alonso, E.E., Vaunat, J. & Gens, A. 1999. Modelling the mechanical behaviour of expansive clays. Engineering Geology, 54: 173183. Gens, A. & Alonso, E.E. 1992. A framework for the behaviour of unsaturated clay. Canadian Geotechnical Journal, 29: 10131032. Sun, D.A., Matsuoka, H., Yao, Y.P. & Ichihara, W. 2000. An elastoplastic model for unsaturated soil in threedimensional stresses, Soils and Foundations, 40(3): 1728. Sun, D.A., Sheng, D.C. & Sloan, S.W. 2007. Elastoplastic modelling of hydraulic and stress-strain behaviour of unsaturated compacted soils. Mechanics of Materials, 39(3): 212221. Sun, D.A., Sun, W.J. & Yan, W. 2009. Hydraulic and mechanical behaviour of sand-bentonite mixture, Proceeding of International Symposium on Unsaturated Soil Mechanics and Deep Geological Nuclear Waste Disposal, 2425 Aug. 2009, Shanghai, pp. 9097. Sun, W.J. 2009. Hydraulic and mechanical behaviour of unsaturated expansive soils and its elastoplastic modelling, Ph. D. Dissertation, Shanghai University, pp. 116132. Zhan, L.T. 2003. Field and laboratory study of an unsaturated expansive soil associated with rain-induced slope instability, Ph. D. Dissertation, Hong Kong University of Science and Technology, pp. 302354.
0.6
e
0.5
No. s/kPa Exp. Pred. 1 300 2 600 3 1200 4 1500
100
(a) e v
80 No. s/kPa Exp. Pred.
1 2 3 70 4 300 600 1200 1500
v/kPa
1000
Sr/%
60
50
40 0 1000
(b) Sr -v
v/kPa
2000
3000
Figure 8. Changes in void ratio and degree of saturation during compression under different constant suctions.
978
A constitutive model for cemented unsaturated soils and weak rocks

R. Tamagnini, M. Mavroulidou & M.J Gunn
Faculty of Engineering, Science and the Built Environment, London South Bank University, London, UK
ABSTRACT: The paper presents the formulation of a constitutive model for unsaturated cemented soils and weak rocks, based on principles of thermodynamics. The model gives some physically based description of the mechanics of unsaturated bonded soils and greatly simplifies their constitutive modelling. The main problems addressed in the model are reversibility and equilibrium (or mechanical instability) and the interactions between the mechanics of the granular skeleton and the bonding material. Destructuration implies the transformation of a continuous porous solid into granular matter. The exchange of mass between these two parts of the continuum (the intact continuous porous solid and the granular matter) is studied as a thermodynamic process. The energy during mechanical destructuration is related to the chemical energy involved during diagenetic or artificial cementation. The model was first used to fit laboratory data for a fully saturated calcarenite and a natural clay. This allowed the study of the effects of mechanical bonding independently of the effects of partial saturation. The constitutive equation was then hierarchically extended to unsaturated conditions, reformulating the effective stress and modifying the hardening term. Ongoing research is applying the model to simulate novel laboratory data for unsaturated lime-treated UK clays. 1 INTRODUCTION the elastoplastic stiffness with different degree of sensitivity. Mechanical destructuration can also produce brittle failure and softening in porous cemented rocks (Lagioia & Nova 1995). The constitutive approach that has been widely applied in the modelling of structured soils is kinematic hardening plasticity, as for example in Gajo & Muir Wood (2003). This approach has mainly been used for the modelling of structured clays. As regards calcareous weak rocks, the approach by Gens & Nova (1992) and Lagioia & Nova (1995) is widely considered as the reference framework. The basic features of this constitutive equation are: a) non associative plasticity, b) a difference between the stress space in which the model is defined and the Terzaghis effective stress being the control variable, c) the dependency of the hardening on a damage variable that is a function of the plastic strains, d) rate-independency of the model. The latter assumption tacitly implies that Terzaghis stress is not the effective stress in calcareous porous rocks because a large amount of strain can be recorded at constant mean effective stress during isotropic compression (see Figure 1). This feature will be discussed in the following sections. In this paper, the constitutive equation is derived from Thermodynamics. Ulm & Coussy (1998) approached the problem of cemented granular materials investigating the constitutive modelling
Chemically stabilised geomaterials (e.g. lime, cement or polymer stabilised soils) are increasingly used in geotechnical applications. The need has therefore emerged for predictive numerical tools able to describe the behaviour of chemically treated soils for a variety of boundary value problems. This should also include unsaturated chemically treated soils, especially as usually these soils are compacted when used in engineering applications. Such tools, able to model the process of artificial cementation (as well as the potential destruction of cementation bonds) in combination with unsaturated soil behaviour are not currently available to practicing engineers. Artificially cemented soils show similar mechanical behaviour to that of naturally bonded geomaterials, such as sedimentary weak calcareous rocks and structured clays. Bonding in soils was originally studied with regard to the behaviour of natural fully saturated soils (Burland 1990). Experimental results showed that geological history produces a diagenetic microstructure that affects the properties of the corresponding reconstituted material: the elastic domain of natural soils can be substantially increased. Microstructure and related effects can however be destructed upon mechanical loading. Mechanical debonding influences also
979
framework that is consistent with the principles of physics, the constraints of the basic postulates of thermodynamics are applied. The thermodynamics of cemented material were developed in a comprehensive and general way by Ulm & Coussy (1998) and extensively discussed in the book of Coussy (2004). The authors defined the free energy of the porous cemented skeleton as: s s ( ij , ijp , m ) U ( , m ) (1)
Figure 1.
Isotropic debonding.
where U free energy of the skeleton related to hardening/softening; m the mass of bonding material; the strain-like internal variable controlling the storage of energy during elastoplastic deformation; and ij and pij the total and plastic strain tensors of the granular matrix. The Clausius-Duhem inequality can be written as: s s p U ij p ij s ij ij ij e U 0 m m
of concrete. Their framework was evaluated in the modelling of weak rocks and then in structured soils and artificially cemented porous materials. The basic assumptions of this approach are implemented in the classical modified Cam clay that is the plastic driver for the deviatoric stress paths. Nova (1991) suggested that non-associative plasticity is a necessary condition for the modelling of mechanical instability; this paper shows that this statement is not true as the presented model describes debonding by an associative flow rule. The adopted thermodynamic approach reduces drastically the number of constitutive parameters (in the limiting case the number of parameters is the same of the classic modified Cam clay). This is a particularly appealing feature of the model for geotechnical applications. 2 THERMODYNAMICS
(2)
in which the derivatives are:
ij
s s p ij ij

U
Am
s m
(3)
All of the referred theoretical and experimental works are based on the postulate that bonding and structure are linked to the volumetric or volumetric plus deviatoric plastic strains, while destructuration is modelled as an irreversible process controlled by a damage or sensitivity internal variable. In this paper, the analysis of the mechanical behaviour starts from the process of cementation. Destructuration is defined as the recovering of an amount of the free energy that has been frozen during natural or artificial cementation. Structuration is explained as the generation of the cementing mass that is the transformation of a solution with hydrates into bonding materials. The reverse of this process is due to the destructuration by loading and the transformation of the cementing material into granular matter. In order to achieve a constitutive
The first equation defines the coupling between Terzaghis effective stress and the total and plastic strains. The second term is the hardening force of the granular matrix, generally expressed by the preconsolidation pressure in soil mechanics and Am is defined as the Affinity (Ulm & Coussy 1998) and has the meaning of a Gibbs potential. The Affinity is the thermodynamic force transforming the solution of water and hydrates into solid bonding material between the grains (dm 0) or during mass consumption (dm 0). The dissipation reads: p 0 s ij ij Am m (4)
This thermodynamic framework was originally formulated for the elastoplastic modelling of early age concrete (Ulm & Coussy 1998). In the present work the constitutive assumptions are adapted for the modelling of soils that are cemented by a bonding structure. Tamagnini (2004) proposed a hardening law for unsaturated soils that can be adopted and fulfils this framework. In the unsaturated soil model by Tamagnini (2004) the cementing mass rate was represented by the variation in the volume of water
980
(i.e. the degree of saturation) and the hardening force was defined as:

U pr exp( )exp( m )
(5)
Then the frozen energy is: U 1 pr exp( )exp( m ) (6)
where Us the skeleton internal energy; pout the pressure of the amount of bonding solid Nout that leaves the intact rock; Qs the heat and Ws the external work. In Equation 10 above, poutVoutNout is the work needed to get the bonding material out from the intact porous solid. Entropy variation is defined as: Ss Qs Ss Nout T (11)
where pr is the reference pressure; v/( ); and is a constitutive parameter Maxwells symmetry implies that: A U 2 m pr exp( )exp( m ) m m (7)
Substituting equation (11) into (10) and rearranging we obtain (for isothermal transformation) the change of the free energy: s 2 s1 d s dU s TdSs ij d ij (U ij ij TS ) Nout dm ij d ij s (12) s It is noteworthy that the free energy of the intact rock or soil (i.e. the continuum under study) is equal to the work rate but also to the variation of Gibbs potential s of the intact solid. The presence of the mass in the frozen hardening energy allows us to define a complementary evolution law for the variable m. Considering destructuration (i.e. the destruction of the bonding mass) as a function of the total strain (and not the plastic strain), the volume change rate of the intact rock or soil during debonding can be expressed as: dV V dm s vv s Vs V (13)
In the limit case when , Affinity can be expressed as: Am pr exp( )exp( m ) (8)
Equation (8) shows that the chemical potential that has cemented the matrix is also the mechanical force that drives the hardening or softening. Another interesting cross effect derives from the Maxwell symmetry:
ij 2 s A m ij m m ij
(9)
Equation (9) explains the stress reduction due to destructuration. This is generally associated with softening, brittle failure and loss of strength and stiffness. It is important to note that in this approach the destructuration is a function of the total strain tensor and not of the plastic strain tensor as commonly assumed in soil mechanics. This is an important feature of the model, because it allows accounting for the effects of damage in the trial predictor stage during numerical implementation. At the start of the numerical non-linear integration in the Gauss points the total strain rate is known, while the plastic strain is not. During cementation, the bonding material enters the solid matrix, which can therefore be considered as an open system. At the same time, during mechanical loading and destructuration, the bonding mass leaves the intact porous medium to become granular matter. We will now consider the variation of the energy of the matrix during destructuration as: U s U s 2 (U s1 U out Nout ) Qs Ws poutVout Nout (10)
where v specific volume. Equation (13) assumes that during debonding the measured strains equal the variation of the solid volume rather than that of the voids. In order to account for the destructuration induced by the deviatoric total strain Equation 13 is modified as: dm v v q s
(14)
Equation (12) is expressed as: d s ij v ij d ij ij d ij
(15)
This is the stress that will be implemented in the numerical algorithm. Equation 4 can be rewritten as:
p ij 0 s ij
(16)
981
It is interesting that in isotropic conditions (as shown in Figure 1), Equation 16 results into: vp p vp 0 s p (17)
Equation 17 shows that the destructuration is an internally reversible mechanical process: the void ratio before compression can be obtained by the dilatancy in the destructurated granular material and the granular material can be recemented restoring the initial condition before loading. Moreover, plastic work is completely frozen in the re-arrangement of the granular matter. Using the same rationale regarding mechanical bonding, the formulation can now be extended to cover unsaturated soil conditions. For an unsaturated soil, the free energy will be expressed as:
p s s ( ij , ij , S r ) U ( , Sr )
f D e d d f f p f d f ij c ij ij p pc dm f pc s pc
(23)
In Equation 23 the last term defines the variation of stress produced by the soil matrix destructuration (see Equation 9). The constitutive equation can be rewritten as: f f e e Dijkl Dklmn mn ij e d d ij Diikl f f kl e H D ij ijkl kl f e Dijkl kl f dm pc f f pc s e H D ij ijkl kl
b d ij d ij
(18)
and for bonded materials:

p s s ( ij , ij , m, S r ) U ( , m, Sr )
(24)
(19)
Terzaghis effective stress can now be replaced by Bishops stress (see for example Houlsby 1997):
ij ij ua ij Sr (ua uw ) ij
(20)
where ua and uw the air pressure and the water pressure respectively; and Sr the degree of saturation of the wetting fluid. 3 CONSTITUTIVE EQUATION: BONDED SATURATED SOILS AND POROUS ROCKS
Equation 24 corresponds to the thermodynamic expression in Equation 15: the first rate term is Terzaghis stress and the second is the rate of mechanical stress due to destructuration. Its thermo-dynamical meaning is reported in Equation 9. Note that in the isotropic case this results into: f 2 f f 2 f KH K K 1 p p p p d v d ii f f 3 H K p p f f K vpc p pc d v f f H K p p
The elastoplastic model is based on the previous theoretical assumptions and on the strain decomposition:
e p d ij d ij d ij
(25)
(21)
Accounting for the condition v and the definition of the hardening modulus H, equation (25) results into Equation 26. 1 0 d v d ii 3 H f K f p p
The plastic strains are obtained by the associative flow rule leading to the constitutive equation: f d ij D e d ij d ij (22)
(26)
where f the loading surface of the modified Cam Clay model; the plastic multiplier. The hardening rule is stated as defined in Equation 5. The consistency condition f f 0 can be expressed as:
Equation 26 is in agreement with the instability reported in isotropic condition after Lagioia & Nova (1995) (see the data in Figure 1). Equation 26 also fulfils Druckers postulate for mechanical instability based on the second order work.
982
It is important to note that the control stress is not Terzaghis stress but the sum of this stress and the enthalpic term rising during the release of the mechanical energy stored in the bonds. 4 CONSTITUTIVE EQUATION: EXTENSION TO UNSATURATED CONDITIONS
The model is now hierarchically upgraded following Tamagnini (2004), to account for unsaturated conditions. In this case Equation 23 is expressed as: f D e d d f f p f d f ij c ij ij p pc f f pcvbdSr pc dm pc pc The resulting constitutive equation is: f f e e Dijkl Dklmn ij mn e d ij d Diikl f f kl e H Dijkl ij kl f e Dijkl kl f pc dm f f pc e H Dijkl ij kl f e Dijkl f kl p dS f f pc c r e H Dijkl ij kl
b S d ij d ij d ij
(27)
namely a calcarenite (Lagioia & Nova 1995). The first test is an isotropic compression shown in Figure 1. The adopted constitutive parameters are: 0.17 0.03 M 1.6 0.3 the added parameter is that is obtained as the inverse of and it is equal to 7.1428. The preconsolidation pressure is equal to 2400 KPa. From Figure 1 it can be seen that the model simulates successfully the mechanical behaviour of the rock. Figures 2 and 3 simulate a drained triaxial test and consequent destructuration. The simulation has been obtained using the same parameters as for the isotropic case presented above except that the preconsolidation pressure was now slightly lower i.e. 2000 kPa. Note that heterogeneity in the properties of the calcarenite referring to the void ratio has been experimented by Coop & Atkinson (1993). However the possibility of the reduction of the plastic yield locus even when the stress is in the elastic condition could be taken into account.
(28)
Figure 2. Drained debonding.
triaxial
test
data,
structural
Note that in Equation 28 the first stress rate is Bishops stress as defined in Equation 20, the second is the enthalpy term due to destructuration and the third is the stress due to the capillary forces controlling the rate of the degree of saturation. 5 SIMULATIONS
The constitutive model for saturated bonding materials has been implemented in the FEM code CRISP (Britto & Gunn 1987). Based on this, a number of isotropic and triaxial tests were simulated. The first set of experimental data that has been simulated using the formulation developed above relates to the destructuration of a porous rock,
Figure 3. Drained triaxial test, compressibility.
983
Figures 4 and 5 show simulations of undrained triaxial tests using the constitutive parameters and the preconsolidation pressure as in the previous drained triaxial test. The results in Figure 4 show the softening simulated by the model. It is also shown that the deviatoric total strains affect the hardening. The fit of the model for this case is somewhat less good, however the overall behaviour of the material is captured. The last simulations are oedometric compression and wetting of structured clays using experimental data from Rao & Ravanisaddappa (2003). Figure 6 shows simulations and experimental data of two compression tests on the saturated bonded and remoulded material and the wetting at constant vertical stress equal to 200 KPa. The figure shows that in this material mechanical debonding does not produce the same effects as those in calcarenite. This implies that the sensitivity parameter
0.8
collapse due to wetting at constant stress
void ratio e
0.6
0.4
exp data bonded exp data remoulded exp data bonded unsaturated
0.2
1
verical net stress : n(KPa)
10
100
1000
Figure 6. Compression tests on saturated samples and wetting on unsaturated bonded samples after Rao & Ravanisaddappa (2002) and tests simulation.
1
0.8
void ratio : e
0.6
0.4
Figure 4.
Simulation of an undrained triaxial test.
exp data bonded unsaturated exp unsaturated remouldeded
0.2 1 10 100 vertical stress :n (KPa) 1000
Figure 7. Wetting on unsaturated bonded samples and remoulded bonded samples after Rao & Ravanisaddappa (2002) and tests simulation.
that controls the hardening due to cementation is equal to 1.0. The other parameters are 0.13 0.03 M 1.2 0.3 and parameter b that controls the softening due to saturation is 3.3. Figure 7 reports the comparison between the collapse due to wetting of the bonded and remoulded material. 6
Figure 5. Undrained triaxial test: stress-strain behaviour.
CONCLUSIONS
A constitutive model was formulated for bonded soils under both saturated and unsaturated
984
conditions. The model is based on thermodynamics and the classic theory of plasticity. The mechanical energy that is stored during cementation was directly related to the free energy controlling the hardening. It was thus shown that non-associated plastic flow is not necessarily needed to model mechanical instability. The formulation of the model for unsaturated soils is currently being used to simulate experimental data for unsaturated lime-treated London Clay produced at London South Bank University. Future work involves the implementation of the model into commercial software for the numerical analysis of boundary value problems involving unsaturated lime-treated soils. ACKNOWLEDGEMENTS This study is part of a larger research programme on the hydro-mechanical properties of lime-treated clays, funded by the UK Engineering and Physical Sciences Research Council (EPSRC) through grant EP/E037305/1. REFERENCES
Britto, A.M. & Gunn, M.J. 1987. Critical state soil mechanics via Finite Elements. Chichester: Ellis Horwood.
Burland, J.B. 1990. On the compressibility and shear strength of natural clays. Gotechnique 40(3): 329378. Coop, M.R. & Atkinson, J.H. 1993. The mechanics of cemented carbonate sands. Gotechnique 43(1): 5367. Coussy, O. 2003. Poro-mechanics, J. Wiley & Sons. Gajo, A. & Muir Wood, D. 2001. A new approach to anisotropic bounding surface plasticity: general formulation and simulations of natural and reconstituted clays. Int. J. Numer. Anal. Methods in Geomechanics 18: 314333. Gens, A. & Nova, R. 1993. Conceptual bases for a constitutive model for bonded soil and weak rocks. In: Anagnostopoulos et al (eds) Geotechnical engineering of hard soilssoft rocks, Rotterdam: Balkema: 483494. Houlsby, G.T. 1997. The Work Input to an Unsaturated Granular Material. Gotechnique 47(1): 193196. Lagioia, R. & Nova, R. 1995. An experimental and theoretical study of the behaviour of a calcarenite in triaxial compression. Gotechnique 45(4): 633648. Nova, R. 1991. A note on sand liquefaction and soil stability. In: Proc. Third Int. Conf Constitutive Laws for Engineering Materials: Theory and Applications, 712 January 1991, Tucson: 153156. Rao, S. & Revanasiddappa, K. 2002. Collapse behaviour of a residual soil. Gotechnique 52(4): 259268. Tamagnini, R. 2004. An extended Cam-clay model for unsaturated soils with hydraulic hysteresis. Gotechnique 54(3): 223228. Ulm, F. & Coussy, O. 1998. Coupling in early-age concrete: from constitutive modelling to structural design. Int. J. Solids Structures 35(3132): 42954311.
985
Constitutive model of unsaturated structured soils under cyclic loading

C. Yang
Centre for Geotechnical and Materials Modelling, The University of Newcastle, NSW, Australia Department of Geotechnical Engineering, Tongji University, Shanghai, China
M.S. Huang Y.-J. Cui
Department of Geotechnical Engineering, Tongji University, Shanghai, China Ecole Nationale des Ponts et ChaussesCERMES, Universit Paris-Est, Paris, France
ABSTRACT: Unsaturated structured soil has been one of research focuses in geotechnical mechanics. This paper aims to investigate the influence of structure degradation and suction variation on unsaturated soils subjected to cyclic loading. Within the framework of plastic bounding surface model, the structure damage theory and unsaturated soil mechanics, an elasto-plastic model for unsaturated soils under cyclic loading has been elaborated. Firstly, on the basis of Barcelona Basic Model, the water retention curve is employed to unite the suction with mechanical stresses under constant water condition. Secondly, the size of yielding surface is assumed to depend on the soil structure damage, and a law of structure degradation is given, linking the structure damage with accumulated plastic strain increments. Besides, by incorporating the concept of mobile mapping origin, the bounding surface model is modified to unify the mapping rule used during both loading and unloading, to simulate the phenomenon of hysteretic properties under cyclic loading. Comparisons between experimental results or published data and model simulations show that the elaborated model is capable of describing unsaturated soil behavior under cyclic loading. 1 INTRODUCTION unsaturated cemented soils subjected to cyclic transportation loads. Since most current constitutive models of soils are developed with respect to remolded soils, the phenomena of cementation damage in natural soils can not be described reasonably. It is necessary to establish some constitutive models which are capable of modeling the cementation damage of soils (Shen, 1996). Generally, the soil structure is attributed to the specific arrangement of solid particles and the cementation of sediments from the pore water between the interfaces of solid particles. Damage of cementation in structured soils under applied external loads is not only able to change the size and shape of initial yielding surface, but also to decrease the whole stiffness of soils (Burghignoli, Miliziano & Soccodato, 1998, Sharma & Fahey, 2003a&b). Once the cementation is damaged completely, the original structured soil approaches to the similar critical state with the correspondingly unstructured soils (Gens & Nova, 1993, Chai, Cui & Lu, 2005). Vaunat and Gens (2003) assumed that soils are composed of solid particles, bond, and void. The bond is taken as a brittle material, whose damage process symbolizes the deterioration of
The long-term deformation of foundations in the transportation system like the subway, expressway, high speed railway, has been one of the very important researches in the international geotechnical engineering. In China, the long-term performance of the metro has caused great differential settlements along the subway tunnel in Shanghai, with the maximum value over 10 cm (Ye Zhu & Wang, 2007). In Japan, the settlement in the embankment of one expressway has been accumulated up to 12 m during its initial 5 years operation (Yasuhara, Hirao & Hyodo, 1988). And in France, instability problems like the formation of sinkholes have been observed near the Northern French high speed railway, with maximum depth up to 1 m (Cui & Terpereau, 2004). Development of similar foundation differential settlements in transportation system will lead to serious safety problems and huge economic lost. With further in-situ measurements and laboratory experiments, some soils are probably within the unsaturated state for the sake of climate, ventilation, and the component of soil itself. Therefore, its important to further study the behaviors of
987
soil structure, and then the Modified Cam-Clay Model is incorporated to describe the constitutive behavior of structured soils. On the other side, Rouainia and Muir Wood (2002), Wei and Huang (2007) modified the function of yielding surface to cover the structural behavior of soils, and attributed the cementation damage to the development of plastic deformation of soils. Bounding surface model and dynamic hardening laws are employed to build up their both constitutive models on the basis of MCCM. It can be seen that the first method above (like Vaunat and Gens) is presented from the microscopic view to describe the cementation damage, but the second one is developed from the macroscopic view, both of which are representative of current researches on structured soils. Under cyclic loading, the deformation of soils is obviously non-linear and usually hysteretic in terms of strain vs. stress. Since the classical elasto-plastic theory assumes a complete elasticity within the yielding surface, its unable to account for the nonlinearity of modulus, change of pore pressure, the incurrence of hysteretic circles, the accumulation of plastic strains, and other characteristic behaviors under cyclic loading (Pastor, Zienkiewicz & Leung, 1985, Pastor, Zienkiewicz & Chan, 1985, Krieg, 1975). The multi-yielding surface theory on the basis of plastic hardening modulus provides a general framework to describe soil behaviors under cyclic loadings. However the multi-yielding surface model is somewhat complicated due to both the necessity to tracking all yielding surfaces in the stress space when modeling and the many parameters difficult to be determined from experiments. Simplifications are made by Krieg (1975), and Dafalias and Popov (1975), thus leading to the current bounding surface model. In this simplified model, the bounding surface and loading surface are defined, and the yielding surface between these two limiting surfaces is determined by interpolation functions. Also, the plastic modulus on the loading surface is a combined function of the distance between the current stress point and the conjugated stress point on the bounding surface. Furthermore, Dafalias and Herrmann (1982) developed the single yielding surface model by degrading the mobile loading surface into one point. As one of the effective tools in modeling soil behaviors under cyclic loading, the bounding surface model has been applied in various kinds of soils such as clays and sands (Pastor, Zienkiewicz & Chan, 1985, Liang & Ma, 1992, Li & Huang, 2007). Chai, Cui & Lu (2005) employed the bounding surface model and generalized plastic model to predict the softening and liquidation phenomena of French loess. On this basis, Yang, Cui, Pereira & Huang (2008) further established a constitutive model for unsaturated cemented soils under cyclic
loading, which describes the structure damage from the microscopic view, but which seems to be complicated in formulation and not easily applicable to engineering. Here this paper aims to simplify and improve the previous model (Yang, Cui, Pereira & Huang, 2008) from the macroscopic view so that to cover more effectively the behavior of unsaturated structured soils under cyclic loading. 2 FORMATION OF THE CONSTITUTIVE MODEL
2.1 Description of unsaturated behaviors of soils Experiments show that the increase in suction can enhance the strength and stiffness of unsaturated soils (Alonso, Gens & Josa, 1990, Wheeler & Sivakumar, 1995). Alonso, Gens & Josa (1990) established the first critical state unsaturated soil constitutive model with two stress state variables, net stress, which is the difference of total stress and pore air pressure, and suction, which is the difference of pore air pressure and pore water pressure. This model is named after Barcelona Basic Model (BBM), and favorably accepted in the unsaturated soil mechanics. This paper employs the Loading-Collapse yielding surface adopted in BBM, and incorporates the van Genuchten Soil Water Retention Curve (van Genuchten, 1980) to combine the suction together with external stresses to describe the hydro-mechanical behavior of unsaturated soils. (1) LC yielding surface
* ( s ) m m p0 p0 c c p p
m ( 0 ) m
(1)
* where p0 and p0 are the preconsolidation stresses in saturated and unsaturated state respectively, pc is the reference stress, m(s) is the deformation coefficient in unsaturated state, defined as
m ( s ) m (0 ) (1 rs )exp( s s ) rs
(2)
where, rs is a constant related to the maximum stiffness of soils, s is a constant controlling the rate of the change of soil stiffness with suction increase, m(0) is the deformation coefficient of soils in saturated state. (2) Soil water retention curve
m
1 Sr ( s ) n 1 s
(3)
where , n and m are fitting parameters from experiments.
988
2.2
Mechanical description of soil structure
To describe the damage of soil structure from a macroscopic view, it generally begins with the constitutive models of totally remolded soils like MCCM, and further to fabricate the enlarged yielding surface due to the existence of initial structure of soils. Subsequently, the evolution of the previous defined yielding surface is constructed to model the damage process of the natural soils (Rouainia and Muir Wood, 2002). This paper assumes the structure damage of soils only influences the size but not the shape of yielding surface. During loading, the structure of soils degrades gradually, and thus causes the yielding surface of the structured soils to contract towards that of totally remolded soils. Define a damage factor rd as rd p0 p0 (4)
tension strength of soils due to the soil structure and suction effects separately. 2.3 Bounding surface constitutive model Within the framework of bounding surface model and critical state theory, one elasto-plastic constitutive model for unsaturated structured soils can be obtained, which not only considers the collapse behavior and soil-water retention curve of unsaturated soils, but incorporates the structure degradation. Formulation of the model is specified as follows. (1) Choice of stress variables Regarding the structure of soils, the damage stress variables, ( p, q )T are given as p p pbc ; q q (7) Notes: in this paper the net stress, deviatoric stress, and suction are denoted as p, q, and s, respectively. (2) Bounding surface equation According to the critical state theory, the bounding surface equation can be written as
* p F ( ,s , p0 , d ) q 2 M 2 ( p ps )( p0 p )
where p0 and p0 are the yielding stresses in structured and unstructured soils separately. And assume the damage factor, rd to be the p exponential function of plastic damage strain d , and can be expressed as
p rd 1 (rd 0 1)exp( kd d )
(5)
(8)
where rd0 is the initial structural factor, kd is the exponent describing the degrading rate of structure with plastic deformation, and the plastic damage strain is given as the function of absolute accumulation of plastic strains, as
p p 2 p 2 d (1 A) (d p ) A (d q ) p d p
where M is the slope of the critical state line. (3) Hardening law During the plastic loading, the hardening parameter p0 can be given by the plastic volumetric strain, and expressed as dp0 p0 1 em p d pm m ( s ) m (9)
(6)
where is the rate of plastic volumetric strain, p is the rate of plastic deviatoric strain, and A is a d q parameter expressing the proportion of the influence p p of d p and d q on the structure damage of soils. Therefore, the constitutive model of unsaturated structured soils can be depicted in Figure 1 in p-q-s stress space. pbc and ps are the components of
where em is the void ratio, m, m are the slopes of normal consolidation line and unloading-reloading line. Furthermore, the associated flow rule is adopted. With the consistency condition, the flow rule, and BS the hardening law, the plastic modulus H L on the yielding surface during loading or unloading can be deduced. (4) Imaging rule and mobile mapping center From the bounding surface theory (Dafalias and Herrmann, 1982), the traditional linear mapping rule is adopted in this paper. The plastic modulus HL at the current stress point is proportional to the corresponding modulus at the conjugate stress point on the bounding surface, thus the plastic deformation within the bounding surface described. Considering the incorporation of unsaturated aspect and structure degradation behaviour, in this paper the mapping rule is defined in the stress space of ps , qs (See Fig. 2). The mapping rule in the traditional bounding surface model is modified to simulate the hysteretic circles during the unloading-reloading process.
Figure 1. Three-dimensional view of yield surfaces in (p, q, s) stress space.
989

p pOL /U 2 q qOL /U 2
p
(13)
Finally, the rate of plastic strains d during the general loading can be gained according to the flowing rule. 2.4 Determination of model parameters
In this model, most parameters can be obtained from common experiments. But some other parameters about the structure degradation like rd0, kd, A and bounding surface model like rL/U0 cant be determined easily. In the following simulations these parameters are determined by parameter optimization method. 3 MODEL VALIDATIONS
In order to testify the efficiency of the model established above, two samples under the simple shearing and the cyclic triaxial shearing with constant water content are simulated separately with the model proposed above, and results are compared with the laboratory experiments and relevant data in previous documents.
Figure 2. Mapping rule in bounding surface model.
According to Masing, the unloading can be considered as the inverse loading with respect to the traditional loading. Therefore, both loading and unloading can be unified into one general loading, further leading the concept of mobile mapping center in the bounding surface model. To be specific, during traditional loading, the mapping centre can be taken as the left interaction point POL of the yielding surface and the abscissa ps (See Fig. 2a). And during the inverse loading (unloading), the mapping center moves correspondingly to the beginning point of the inverse loading (See Fig. 2b). Therefore, the specific mapping rule can be given as follows
BS 0 H L /U H L /U rL /U
3.1 Simple shearing test under constant water content With reference to experimental data in Alonso, Gens & Josa (1990), the sample is sheared under constant net stress, p 150 kPa, and constant water content. Although tests in Alonso et al. (1990) are simulated at constant net stress and constant suction (s 100, 200, 300 kPa), their data are still very useful for the validation of this model. Values of parameters are listed in Table 1. And the comparison between the simulation of this paper and the know data is depicted in Figure 3. It can be seen that with the cooperation of external applied stress, suction and degree of saturation, the calculated values of suction and soil strength decreases with soil compression during shearing, in accordance with conclusions among others (Alonso, Gens & Josa, 1990, Cui & Delage, 1993). Meanwhile, when compared with totally remolded soils, the strength of structured soils is bigger, and its decreasing range of suction is smaller.
Table 1. Parameters used in simulation of the shear test. p0* (kPa) 150 pc (kPa) 100 s0 (kPa) 300 e0 0.90 rL0 1.4 1.25e-2 um 0.35 ks 0.60 rd0 1.35 m(0) 0.20 rs 0.75 kd 3.5 m 0.02 n 0.5 A 0.25 s 8e-3 m 3 M 1.0 Sr 0.01 pbc (kPa) 10
(10)
where rL/U is the mapping exponent in unified loading, and varies with the structure degradation as follows. rL /U rd rL /U 0 (11) where rL/U0 is the mapping exponent of equivalent remolded soils, 0 and are the distances from the imaging point on the bounding surface and the current stress point to the imaging center, listed as
0
pI pOL /U 2 qI qOL /U 2
(12)
990
Table 2. Parameters used in simulation of the cyclic triaxial shear tests. p0* (kPa) 700 pc (kPa) 25 s0 (kPa) 1000 e0 0.93 m 0.012 pbc (kPa) 10 um 0.25 rL0 1.36 rd0 1.35 m(0) 0.17 rU0 1.85 kd 3.5 s 0.01 s 0.01 A 0.25 M 1.1 rs 0.75 n 5.75 ks 0.02 Sr 0.16 m 0.06
Figure 3. Response of model to shear tests with constant water content and constant mean net pressure.
But the structured soil finally tends to the same critical state as the totally remolded soils. This result also can be found in papers of Leroueil & Vaughan (1990), Gens & Nova (1993), and Chai, Cui & Lu (2005). 3.2 Cyclic triaxial shearing tests of unsaturated French loess under constant water
French loess is known as the typical unsaturated cemented soils (Cui & Terpereau, 2004). Here the initial water content of studied soils is 23%. The sample is first consolidated with 25 kPa confined stress, and sheared with single or multiple levels of cyclic loading. Values of parameters are listed in Table 2. First one single level of cyclic shearing between 0 and 75 kPa has been simulated with this model. Results in terms of deviatoric stress vs. deviatoric strain, and axial strain vs. cyclic loading number are shown in Figure 4. It seems that this model is capable to offer fairly nice predictions on the strain accumulation and also the hysteresis under cyclic loading. To further validate the efficiency of this model in predicting the plastic strain accumulation under cyclic loading, multiple levels of cyclic loading is carried out. The initial deviatoric stress varies between 0 and 15 kPa, afterwards the peak value of the deviatoric stress in each level is increased by 15 kPa. In each stress level, the cycling of 100 times is applied till the sample is destroyed. Results in Figurwe 5 shows that the model proposed in this paper can predict the deformation of soils under cyclic loading.
Figure 4. Response of model to single level cyclic triaxial shear tests with constant water content.
Figure 5. Axial strains of samples with constant water content under multi-level cyclic loading.
991
CONCLUSION
The mechanical behavior of unsaturated structured soils under cyclic loading has been studied in this paper. Within the framework of bounding surface model, one elasto-plastic model is proposed with the combination of structure damage theory and unsaturated soil mechanics. Meanwhile, the mobile mapping origin is introduced to modify the traditional mapping rule in bounding surface model, unifying the loading and unloading into one generalized loading. Comparisons between model predictions and experimental results or know data show that this model is capable to simulate the mechanical behavior of unsaturated structured soils mentioned in this paper both qualitatively and quantitatively well. However, the influence of deformation on the soil water retention curve is not included in this paper. And some parameters are difficult to be measured directly. Therefore, more efforts are required for the improvement and validation of this model. REFERENCES
Alonso E.E, Gens A. & Josa A. A constitutive model for partially saturated soils[J]. Gotechnique, 1990, 40(3):405430. Becker T. & Meissner H. Direct suction measurements in cyclic triaxial test devices. Proc. of the 3 st Int. Conf. on Unsaturated Soils (UNSAT 2002) Recife, Brazil: Swets and Zeitliger, 2:459462. Becker T. & Li T. Behavior of unsaturated soils subjected to cyclic loading. Proc. of Int. Workshop on Unsaturated Soiols, Weimar, Germany: Springer, 2003:347364. Burghignoli A, Miliziano S. & Soccodato F.M. The effect of bond degradation in cemented clayey soils[C]. Proc. Symp. on Geotechnical Engineering of Hard Soils Soft Rocks. 1998: Balkema: 465472. Chai H.Y, Cui Y.J. & Lu Y.F. Simulation of loess behaviors under cyclic loading[J]. Chinese Journal of Rock Mechanics and Engineering, 2005, 24(23):42724281. Cui Y.J. & Delage P. On the elasto-plastic behavior of an unsaturated silt[C]. Unsaturated Soils, Proceedings of the 1993 ASCE National Convention and Exposition, Dallas, TX, USA, 1993:115126. Cui Y.J., Terpereau J.M., et al. A geological and geotechnical characterization of the loess in Northern France [C]. A.W. Skempton Memorial Conference, 2004, 1:417428. Dafalias Y.F. & Herrmann L.R. Bounding surface formulation of soil plasticity[A].Soil Mechanics Transient and Cyclic Loads: Constitutive Relations and Numerical Treatment[M], 1982:53282. Dafalias Y.F. & Popov E.P. A model of nonlinearly hardening materials for complex loading. Acta Mechanica, 1975, 21:173192. Gens A. & Nova N. Conceptual bases for a constitutive model for bonded soils and weak rocks[C]. Proc. Symp. on Geotechnical Engineering of Hard Soils Soft Rocks. Balkema, 1993:485494.
Krieg R.D. A practical two-surface plasticity theory. Journal of Applied Mechanics, ASME, 1975, 42:641646. Leroueil S. & Vaughan P.R. The general and congruent effects of structure in natural soils and weak rocks. Gotechnique, 1990, 40(3):467488. Li X.Z. & Huang M.S. A bounding surface model for creeping soft clay under cyclic loading[J]. Chinese Journal of Geotechnical Engineering, 2007, 29(2):249254. Liang R.Y. & Ma F. Anisotropic plasticity model for undrained cyclic behavior of clays: theory. Journal of the Geotechnical Engineering, 1992, 118(2):229245. Meissner H, Becker T. Material behavior of unsaturated soils subjected to cyclic loading. Proc. of the 3 st Int. Conf. on Unsaturated Soils (UNSAT 2002) Recife, Brazil: Swets and Zeitliger, 1:227232. Pastor M, Zienkiewicz O.C. & Chan A.H.C. Generalized plasticity and modeling of soil behavior[J]. International Journal for Numerical and Analytical Methods in Geomechanics, 1990, 14:151190. Pastor M, Zienkiewicz O.C. & Leung K.H. Simple model for transient soil loading in earthquake analysis(II): non-associative models for sands[J]. International Journal for Numerical and Analytical Methods in Geomechanics, 1985, 9:477498. Rouainia M. & Muir Wood D. A kinematic hardening constitutive model for natural clays with loss of structure. Gotechnique 2000, 50(2):153164. Sharma S.S. & Fahey M. Cyclic deformation characteristics of cemented calcareous soil[C]. Deformation characteristics of Geomaterials, ISLyon-03. Lyon, France: Balkema, 2003:169177. Sharma S.S. & Fahey M. Degradation of Stiffness of Cemented Calcareous Soil in Cyclic Triaxial Tests[J]. Journal of Geotechnical and Geoenvironmental Engineering, 2003, 129(7):619629. Shen Z.J. An Elasto-plastic Damage Model for Cemented Clays[J].Chinese Journal of Geotechnical Engineering, 1996, 18(1):9597. van Genuchten M Th. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils[J]. Soil Sci. Soc. Am. J. 1980, 44: 892898. Vaunat J. & Gens A. Bond degradation and irreversible strains in soft argillaceous rock[C]. Proceedings of the 12th Panamerican Conference on Soil Mechanics and Geotechncial Enginnering, 2003:479484. Wei X. & Huang M.S. An anisotropic bounding surface model for natural structured clays[J]. Chinese Journal of Geotechnical Engineering, 2007,29(8): 12251229. Wheeler S.J. & Sivakumar V. An elasto-plastic critical state framework for unsaturated soil[J]. Gotechnique, 1995, 45(1):3553. Yasuhara K, Hirao K. & Hyodo M. Partial-drained behavior of clay under cyclic loading[A]. Proc 6th Int1 Conf on Numerical Methods in Geomechanics[C], Balkema, 1988: 659664. Ye Y.D, Zhu H.H. & Wang R.L. Analysis on the current status of metro operating tunnel damage in soft ground and its causes [J]. Chinese Journal of Underground Space and Engineering, 2007, 3(1):157160, 166.
992
Hydraulic conductivity function for unsaturated cracked soil

L.M. Zhang, J.H. Li & D.G. Fredlund
Department of Civil and Environmental Engineering, The Hong Kong University of Science and Technology, Clear Water Bay, Hong Kong
ABSTRACT: This paper presents a method to predict the hydraulic conductivity function for cracked soil considering crack volume changes during drying-wetting processes. The cracked soil is viewed as an overlapping continuum of a crack network system and a soil matrix system. The hydraulic conductivity functions for the two pore systems can be determined based on their pore size distributions and combined to give the hydraulic conductivity function for the cracked soil at a particular state. The hydraulic conductivity function for the cracked soil at different states along a crack development path can be obtained and combined to give the hydraulic conductivity function for the cracked soil considering crack volume changes. An example is presented to illustrate the prediction of the hydraulic conductivity functions for a cracked soil. The hydraulic conductivity is significantly affected by the presence of cracks. When the cracks are still open under saturated conditions, the hydraulic conductivity function is bi-modal. When the cracks close completely under saturated conditions and open at low suctions, the saturated permeability is dominated by the soil matrix initially. The hydraulic conductivity at low suctions may be dominated by that of the crack network, whereas the hydraulic conductivity at high suctions is controlled by the soil matrix because water cannot pass through the cracks with large apertures. 1 INTRODUCTION The objective of this research is to propose a procedure to determine the permeability function for an unsaturated cracked soil considering crack volume change during a drying or wetting process. The cracks are considered as a pore series of known geometry in the soil at different states.
Cracks provide preferential pathways for water flow, which significantly increases the hydraulic conductivity of the soil mass and, in turn, pore water pressures in the soil mass. As a result, the shear strength of the soil will decrease. As water infiltration into the cracked soil often starts from unsaturated conditions, the water retention ability and hydraulic conductivity under unsaturated conditions are important factors. The water retention ability and hydraulic conductivity of cracked soil vary with the soil suction under unsaturated conditions. A water retention curve or a Soil Water Characteristic Curve (SWCC) and a coefficient of permeability function are often used to represent such variations. The water retention ability and hydraulic conductivity of cracks depend on the crack geometry. Cracks in soils will deform during a drying or wetting process. The hydraulic conductivity of a cracked soil will vary with changes in both crack volume and suction. When ignoring the crack volume change, Zhang and Fredlund (2003) suggested a method to determine the water retention curve for a discontinuous material based on the pore size distributions of the fractures and the rock matrix. How to determine the water retention curve and the permeability function for a cracked soil considering crack volume change is still a challenging problem.
TENSORIAL PERMEABILITY FUNCTION FOR IDEALIZED NON-DEFORMABLE CRACKED SOIL Saturated coefficient of permeability of desiccation cracks
2.1
The saturated coefficient of permeability of desiccation cracks here refers to the coefficient of permeability of saturated cracks when they are fully saturated and the volume change of cracks is negligible. The desiccation crack network can be generated based on digital photos obtained from a field survey. The saturated permeability tensor of the crack network can then be calculated using a computer program, FRCN that simulates water flow in a saturated crack network (Li 2009; Li et al. 2009). The calculated permeability tensor is the saturated permeability tensor of the desiccation crack network without considering its volume change.
993
2.2
Unsaturated permeability function for desiccation cracks
ks() kc() (1 wc)km()
(4)
The coefficient of permeability for a crack network is not a constant. It depends on the saturated permeability and the volumetric water content, or the suction, . Therefore, the permeability function for desiccation cracks represents the relationship between the coefficient of permeability and suction. There are two approaches to obtaining the permeability function for an unsaturated soil: empirical equations (Richards 1931; Brooks and Corey 1964), and statistical models (Childs and Collis-George 1950; Burdine 1953; Fredlund et al. 1994). These approaches can be applied to cracks if the cracks are viewed as an equivalent porous medium. The permeability function can be predicted when the saturated permeability tensor for the cracks, ksc and the water retention curve are available. The procedure to predict the permeability function for desiccation cracks is described as follows taking the Brooks and Corey method (1964) as an example. Three soil parameters can be identified from a water retention curve for desiccation cracks: c the air entry value of the cracks, aev , the residual c degree of saturation, Sr , and the pore size distribution index, c. These parameters can readily be visualized if the saturation condition is expressed c in terms of an effective degree of saturation, Se ,
c Se
In general, there is a complementary permeability (m) tensor, kij , for the soil matrix in addition to the per(c ) meability tensor, kij , for the crack network. If the soil-water characteristics are isotropic, the tensorial form of permeability function for the cracked soil is,
(s) (c ) (m) kij ( ) kij ( ) (1 wc )kij ( )
(5)
PREDICTION OF PERMEABILITY FUNCTION CONSIDERING CRACK VOLUME CHANGE Desiccation-induced crack volume change
3.1
S c Src 1 Src
(1)
where Sc is the degree of saturation of cracks and assumed to be isotropic. The coefficient of permeability for the cracks c with respect to the water phase, kw , can be predicted c from the Se curve as follows (Brooks and Corey 1964):
c c ksc for aev kw c c c c c kw ks (Se ) for aev
(2)
where ksc is the saturated coefficient of permeability for cracks with respect to the water phase; c is an empirical constant, which is related to the pore size distribution index:
c
2.3
2 3c c Prediction of permeability function for non-deformable cracked soil
(3)
Both laboratory tests and field experiments were conducted to study desiccation-induced cracks. In this research, a field survey was conducted to study the desiccation-induced cracks at a site focusing on changes in crack pattern and geometry with water content (Li 2009). The test results showed that the crack development follows three stages during a drying process. In stage 1, water content decreases slowly but few new cracks appear. In stage 2, new cracks appear and develop quickly as water content continues to decrease. In stage 3, water content decreases slowly and the cracks tend to be steady. In a wetting process the aperture of cracks becomes smaller when the water content increases. Cracks tend to close to a certain extend for highly plastic soils due to self-healing or clogging. On the other hand, cracks keep open for non- or low-plastic soils (Eigenbrod 1996; Rayhani et al. 2008). Obviously the crack development is closely related to the water content, and in turn the soil suction. If the crack porosity is used as an indicator to represent the crack development, the crack development can be described by a relation between the crack porosity (nc) and the soil suction . Figure 1 shows a simplified relation between crack porosity and soil suction. The initial crack porosity is nb when the suction is small. When there are no cracks initially or cracks close after wetting, the initial crack porosity nb can be zero. Cracks start to develop after the suction is larger than b. Cracks develop and the crack porosity increases with the increase of suction. Finally cracks tend to be steady when the suction reaches s and the crack porosity at the steady state is ns. Figure 1 shows a linear relationship between crack porosity and suction. More realistic relationships can be established when sufficient data are available. 3.2 Permeability function for cracked soil considering crack volume change Although cracks are deformable as shown in Fig. 1, the cracks in a short period during the
The permeability function for the cracked soil can be obtained following the procedure for the saturated permeability for a cracked soil (Li et al. 2009). The permeability function for the cracked soil is,
994
w
ks3
nc
s
Air-entry value
k3 nc3
(3, k3)
3
Crack development path
r
b
Residual water content

s
ks2
Log()
k2 k nc2 ks1 (1, k1)
(2, k2)
Porosity
2
ns
nc1
1
nb
b s
Log()
Figure 2. Relationship between permeability of a cracked soil and soil suction considering crack volume changes.
Figure 1.
Crack development path.
drying-wetting process can be assumed to be nondeformable. The permeability functions at various states can be combined to give a permeability function for the cracked soil considering crack volume changes. Figure 2 illustrates a proposed procedure to obtain the permeability function considering crack volume changes. The procedure is as follows: 1. At a certain state corresponding to a particular suction, obtaining the pore size distributions for the cracks and the soil matrix. 2. Estimating the water retention curves for the cracks and the soil matrix from the pore size distributions at the specified state. 3. Calculating or measuring the saturated permeability tensors for the cracks and the soil matrix at the specified state. 4. Predicting the permeability functions for the cracks and the soil matrix at the specified state based on the water retention curves and the saturated permeability tensors; 5. Combining the permeability functions for the cracks and the soil matrix to obtain the permeability function (or soil water characteristic curves) for the cracked soil corresponding to the cracks at the specified state. Figure 2 shows three permeability functions at different states in the -k plane. The permeability functions are for the cracked soils containing non-deformable cracks at crack porosity nci. The saturated coefficient of permeability of the cracked soil with
crack porosity nci is ksi, and the suction is i at nci. 6. Obtaining the crack development path which shows the relationship between crack porosity and suction. An assumed crack development path is shown in the -nc plane in Fig. 2. The crack development path intersects with the permeability functions at suctions i, which are the suctions when the cracks with crack porosity nci are present. Then a pair of (i, ki) (or (i, i)) can be found on the corresponding permeability function. 7. Combining these pairs of (i, ki) (or (i, si)) in each permeability function curve, which yields a permeability function curve for the cracked soil considering crack volume changes. 4 WORKED EXAMPLE
A cracked soil ground was documented at four different times. Details of the field work are reported by Li (2009). The predominant particles of the cracked soil were silt (34%) and clay (35%), with some sand (23%) and gravel (8%). It was a silty clay in the Unified Soil Classification System. The crack pore size distributions measured in field at the four snapshots are shown in Fig. 3. The crack porosity values for the cracked soils on the four observation days were 0.41%, 0.95%, 2.61%, and 3.10%, respectively. The water contents during the four days were also measured in field. The
995
Crack porosity (%)
0.3 0.2 0.1 0

0
0.1
0.3 0.2 Aperture (mm)
0.4
0.5
Crack porosity (%)
0.3 0.2 0.1 0 0 0.1
0.19
0.27
Aperture (mm)
0.36
0.44
0.53
Crack porosity (%)
0.6 0.4 0.2 0
0.1
0.2
0.29
0.39
0.49
0.58
0.68
0.77
0.87
0.97
Aperture (mm)
Crack porosity (%) 0.8 0.6 0.4 0.2 0.0 0 0.1 0.2 0.29 0.38 0.47 0.57 0.66 0.75 0.85 0.94 1.03 1.31
The saturated permeability tensors for the crack network are calculated using the numerical program, FRCN (Li 2009). The crack network surveyed on 12 November 2008 was imported into the numerical model. Flow through the crack network was simulated and the permeability tensor for the saturated crack network was obtained. The saturated permeability values are 7.0 103 m/s along the north-south direction, 9.0 103 m/s along the west-east direction, and 1.35 102 m/s along the vertical direction for the crack network. In the horizontal direction, the major principal saturated permeability is 1.1 102 m/s along 34 and the minor one is 5 103 m/s in the direction perpendicular to the major principal direction. The tensorial form of the permeability function for the crack network can be estimated using the Brooks and Corey method (1964) as, 7.0 10 3 2.9 10 3 0 c 3 3 2.9 10 9.0 10 0 Se 2 0 0 1.35 10
Aperture (mm)
Figure 3. Crack pore size distributions at the four snapshots: (a) 02 Nov. 2007; (b) 04 Nov. 2007; (c) 06 Nov. 2007; (d) 12 Nov. 2007.
(c ) kij

c
(6)
corresponding matric suction can then be obtained using the SWCC for the soil, which was also measured. The crack development drying path for the cracked soil was obtained based on the crack porosity data and the corresponding suction measurements. When the soil was saturated no cracks appeared. When the suction of the soil reached about 47 kPa some cracks appeared and the crack porosity was 0.41% as shown in Fig. 3(a). After then cracks developed rapidly and reached a crack porosity of 2.61% as shown in Fig. 3(c) at a suction of 8300 kPa. The cracks nearly reached a steady state with a crack porosity of 3.10% at a very high suction as shown in Fig. 3(d). In order to obtain the hydraulic properties during a wetting process, a crack development wetting path is assumed. The crack porosity on the wetting path is larger than that on the drying path. Using the capillary law, the water retention curves for the four crack networks can be predicted based on their pore-size distributions shown in Fig. 3. Take the cracked soil surveyed on 12 November 2007 as an example. The crack pore size distribution is shown in Fig. 3(d). The drying water retention curve for the crack network can be obtained based on the pairs of suction and volumetric-water content values using the Brooks and Corey method. The wetting water retention curve for the crack network is also obtained using the Pham et al. (2005) scaling method. The distance between the wetting curve and the drying curve is assumed to be 20% log-cycle and the slope of the wetting water retention curve is assumed to be the same as that of the drying curve.
The drying and wetting permeability functions for the crack network can be obtained from the drying and wetting SWCCs, respectively, in which c is c 3.6 obtained from the Se relation. The hysteresis permeability function along the north-south horizontal direction is shown in Fig. 4. The drying SWCC curve for the soil matrix was measured in the laboratory using a special pressure plate extractor (Kwong 2009). The drying SWCC is used to predict the wetting SWCC by the scaling method. The distance between the drying curve and the wetting curve is assumed to be 50% log-cycle following Pham et al. (2005). The slope is assumed to be the same as that of the drying curve. The
1.0E-01 1.0E-02 1.0E-03 1.0E-04 1.0E-05 Coefficient of permeability (m/s) 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 1.0E-15 0.01 Drying Wetting
0.1 Suction (kPa)
10
Figure 4. Estimated permeability functions along a horizontal direction for the crack network shown in Fig. 3(d).
996
C oeffic ient of per m eability) ( m /s
hysteresis permeability function for the soil matrix can then be obtained using the hysteresis SWCC and the saturated permeability for the soil matrix. The values of the saturated coefficient of permeability of the soil matrix were measured in the laboratory, which are 6.0 107 m/s along the north-south direction, 3.0 106 m/s along the west-east direction, and 3.5 106 m/s along the vertical direction. Assuming the soil matrix is transversely anisotropic, a tensorial form of permeability function for the soil matrix is, 0.6 10 6 0 0 m 6 0 3 10 0 Se 6 0 0 3.5 10
1. 0E -01 1. 0E -02 1. 0E -03 1. 0E -04 1. 0E -05 1. 0E -06 1. 0E -07 1. 0E -08 1. 0E -09 1. 0E -10 1. 0E -11 1. 0E -12 Drying Wet t ing
(m) kij

m
(7)
1. 0E -13 1. 0E -14 1. 0E -15 0.01
0.1
10 S uct ion (kP a)
100
1000
10000
The porosity of the soil matrix is 44.3%. The hysteresis permeability function along the north-south direction is shown in Figure 5, in m which m is 12 based on the Se relation for the soil matrix. The tensorial form of the permeability function for the cracked soil can be obtained using eq. 5, 7 10 3 2.9 10 3 0 c c 3 3 kij 2.9 10 9 10 0 Se 2 0 0 1.35 10 6 0.6 10 0 0 m 6 0.97 0 3 10 0 Se 6 0 0 3.5 10
Figure 6. Horizontal permeability functions for the cracked soil shown in Fig. 3(d).

m
(8) The hysteresis permeability function for the cracked soil can also be obtained by combining the hysteresis permeability functions for the cracks and for the soil matrix using eq. 5. Figure 6 shows the permeability function along the north-south
1.0E-05 1.0E-06 1.0E-07 Coefficient of permeability (m/s) 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 1.0E-15 0.01 Drying Wetting
0.1
10 Soil suction (kPa)
100
1000
10000
Figure 5. Estimated horizontal permeability functions for the soil matrix.
direction for the cracked soil. When the cracks are still open under saturated conditions, the hydraulic conductivity function is bi-modal. The significant difference in permeability results from the large permeability of the cracks in the low suction range, which is several orders of magnitude larger than that of the soil matrix in the same suction range. The cracks increase the coefficients of permeability of the cracked soil greatly at low suctions. The permeability functions for the idealized cracked soils at the other three snapshots in Figs. 3(a)(c) are obtained following the same procedure and shown in Figure 7. As the cracks in the cracked soil grow during the drying process, the saturated permeability increases if the cracks do not deform. Following the crack drying process shown in Fig. 3, the drying permeability function considering crack volume change can be obtained using the drying permeability functions of the idealized cracked soils at different snapshots. Similarly, the wetting permeability function can be obtained using the wetting permeability functions at different snapshots following an assumed wetting path. The hysteresis permeability functions considering crack volume changes are shown in Figure 8. When the suction is low the permeability during drying is similar to that for the soil matrix. On the other hand, the permeability during wetting is much higher than that for the soil matrix. The difference between the two cases is caused by different crack development paths. On the drying path, cracks open at a larger suction (47 kPa), which indicates that cracks do not contribute significantly to the permeability until the suction reaches 47 kPa. On the wetting path some cracks
997
1.0E-02 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 (a) 1.0E-15 0.01
Drying Wetting
0.1
10 Suction (kPa)
100
1000
10000
1.0E-02 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 (b) 1.0E-15 0.01
Coefficient of permeability (m/s )
Drying Wetting
permeability of the cracked soil is dominated by that of the crack network in the low suction range in the wetting process. When the soil suction is large, the crack aperture becomes large. However the large cracks are incapable of conducting water at large suctions. The small pores in the soil matrix are instead the primary flow paths at large suctions. Therefore, the permeability at high suctions is dominated by the permeability of the soil matrix and the permeability is similar to that for the soil matrix both in the drying process and in the wetting process. 5 SUMMARY
0.1
10 Suction (kPa)
100
1000
10000
1.0E-02 1.0E-03 1.0E-04 1.0E-05 1.0E-06 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 (c) 1.0E-15 0.01
Drying Wetting
0.1
10 Suction (kPa)
100
1000
10000
Figure 7. Permeability functions for the cracked soil shown in: (a) Fig. 3(a); (b) Fig. 3(b); (c) Fig. 3(c).
1.0E-03 1.0E-04 1.0E-05 1.0E-06 Coefficient of permeability (m/s ) 1.0E-07 1.0E-08 1.0E-09 1.0E-10 1.0E-11 1.0E-12 1.0E-13 1.0E-14 1.0E-15 0.01 Drying PF for cracked soil Wetting PF for cracked soil Drying PF for soil matrix Wetting PF for soil matrix
0.1
10 Suction (kPa)
100
1000
10000
Figure 8. Permeability Functions (PFs) for the cracked soil considering crack development.
This paper presents a method to predict the permeability function for cracked soil considering crack volume changes during drying-wetting processes. First, the cracked soil element is analyzed without considering the crack volume change. The cracked soil is viewed as an overlapping continuum of a crack network system and a soil matrix system. The pore size distributions for the two pore systems can be established. Then the water retention curves and permeability functions for the two pore systems are estimated using their pore size distributions. The estimated water retention curves for the two pore systems can be combined to give the SWCC for the cracked soil at that state. Similarly, the permeability function for the cracked soil at that state can be obtained. Secondly, the SWCC and permeability function for the cracked soil at different states along a crack development path can be obtained. Finally, the permeability function for the cracked soil at different states can be combined to give the permeability function for the cracked soil considering crack volume changes. Using the information from a field crack survey, an example is presented to illustrate the prediction of the hysteresis permeability functions for the cracked soil. The permeability may be dominated by that of the crack network at low suctions, but by that of the soil matrix at high suctions. Although changes in crack geometry are considered in this paper, a coupled analysis is not intended. The cracks are simply considered as a pore series of known geometry in the soil at different states. ACKNOWLEDGMENT This research was substantially supported by the Research Grants Council of the Hong Kong Special Administrative Region (Project No. 622206).
are still open at low suctions (0.1 kPa) so that the cracks contribute significantly to the permeability in the low suction range. Because the permeability of the crack network in the low suction range (e.g., 1 kPa) is several orders of magnitude larger than that of the soil matrix (e.g., in eq. 8), the
998
REFERENCES
Brooks, R.H. & Corey, A.T. 1964. Hydraulic Properties of Porous Media. Colorado State University Hydrology Paper No. 3: 27. Burdine, N.T. 1953. Relative permeability calculations from pore size distribution data. Trans. of AIME 198: 7178. Childs, E.C. & Collis-George, N. 1950. The permeability of porous materials. Proceedings of the Royal Society of London 201 A: 392405. Eigenbrod, K.D. 1996. Effects of cyclic freezing and thawing on volume changes and permeabilities of soft fine-grained soils. Can. Geotech. J. 33(4): 529537. Fredlund, D.G., Xing, A. & Huang, S.Y. 1994. Predicting the permeability function for unsaturated soils using the soil-water characteristic curve. Can. Geotech. J. 31: 533546. Kwong, C.P. 2009. Experimental Study of Water Infiltration in Cracked Soil. MPhil thesis, Hong Kong University of Science and Technology.
Li, J.H. 2009. Field Experimental Study and Numerical Simulation of Seepage in Saturated/unsaturated Cracked Soil. Ph.D. thesis, Hong Kong University of Science and Technology. Li, J.H., Zhang, L.M., Wang, Y. & Fredlund, D.G. 2009. Permeability tensor and REV of saturated cracked soil. Can. Geotech. J. 46: 928942. Pham, H.Q., Fredlund, D.G. & Barbour, S.L. 2005. A study of hysteresis models for soil-water characteristic curves. Can. Geotech. J. 42: 15481568. Rayhani, M.H.T., Yanful, E.K. & Fakher, A. 2008. Physical modeling of desiccation cracking in plastic soils. Engineering Geology 97: 2531. Richards, L.A. 1931. Capillary conduction of liquids through porous medium. Physics 1: 318333. Zhang, L.M. & Fredlund, D.G. 2003. Characteristics of water retention curves for an unsaturated fractured rock mass. Proceedings, Second Asian Conference on Unsaturated Soils, 1012 Nov. 2003, Osaka, Japan, 425429.
999
Parameter identification in BBM using a parallel asynchronous differential evolution algorithm

Youliang Zhang & Xia-Ting Feng
State Key Laboratory of Geomechanics and Geotechnical Engineering, Institute of Rock and Soil Mechanics, Chinese Academy of Sciences, Wuhan, China
Domenico Gallipoli Charles Augarde
Department of Civil Engineering, University of Glasgow, Glasgow, UK School of Engineering & Computing Sciences, Durham University, Durham, UK
ABSTRACT: The application of the Barcelona Basic Model (BBM) to practical geotechnical problems requires the selection of accurate parameter values based on laboratory or field tests. This is often an arduous task due to lack of good quality experimental data, cost of lengthy laboratory testing campaigns, difficulties in the interpretation of test results and unavailability of well-established calibration methods. In-situ techniques are usually preferred to laboratory tests as they offer a quicker and less expensive alternative while overcoming limitations associated to the potential disturbance of the samples retrieved from the field. In this context, the paper presents some preliminary work carried out within a wider project to devise a methodology for the definition of constitutive model parameter values through the back analysis of pressuremeter tests. In particular, a parallel asynchronous differential evolution algorithm using a client-server model is proposed for parameter calibration in BBM. A validation example is also presented, which demonstrates the ability of the algorithm to identify up to eight parameters simultaneously. 1 INTRODUCTION to the lack of reliable methodologies for interpreting test results. A new algorithm, based on Differential Evolution (DE) (Storn 1997), is here employed to backanalyse pressuremeter test results. The proposed algorithm makes use of a stochastic global optimization method capable of dealing with nonlinear and multimodal minimization problems. DE has shown superior performance in real-world applications (Storn 1999) but the algorithm is computationally expensive due to slow rate of convergence. This work presents a possible solution that overcomes high computational demands by developing a parallel asynchronous version of the conventional differential evolution algorithm. 2 2.1 PROBLEM DEFINITION Introduction of BBM
The Barcelona Basic Model (BBM) was introduced by Alonso et al. (1990). Since then, it has been widely used in geotechnical analyses due to its ability to describe many aspects of the mechanical behaviour of unsaturated soils. However, the successful use of any constitutive model in geotechnical engineering depends not only on the ability of the model to capture soil behaviour, but also on the selection of accurate parameter values based on field or laboratory data. For BBM, extra difficulties are met due to its high nonlinearity, relative complexity and large number of parameters. To date, most research has focused on the extension and improvement of the original BBM and relatively little work has been devoted to developing methodologies for parameter calibration (e.g. Wheeler et al. 2002). In this paper, a method is described that allows simultaneous identification of parameter values in BBM by interpreting the results obtained from pressuremeter tests. The pressuremeter technique is widely used for in-situ characterization of saturated soils, however it has been rarely applied to unsaturated soils mainly due
The BBM is an elastoplastic constitutive model, which extends the Modified Cam-Clay model to the case of partly saturated soils. In BBM, soil behaviour is described for triaxial stress states in terms of mean net stress p, deviator stress q and
1001
suction s. The elliptic yield function in the stress space (p, q, s) is defined as: F q 2 M 2 p ps p0 s p (1)
d se
G dq 3
(8)
where M is the slope of the critical state line, p0(s) is the isotropic preconsolidation stresses at suction s and ps is a parameter that defines the linear dependency of apparent cohesion on suction: ps k s (2) The preconsolidation pressure p0(s) depends on suction according to: p* s p0 s p 0 c p
c
where s describes the elastic compressibility in the v-ln (s pat) plane (where pat is atmospheric pressure) and G is the shear modulus In BBM, there are therefore nine parameters (G, M, k, , s, (0), r, , pc), of which six (, s, (0), r, , pc) are mainly associated with the isotropic behaviour of the soil and three (G, M, k) are mainly associated with deviatoric behaviour. 2.2 Pressuremeter tests in unsaturated soils
0
(3)
* where p0 is the isotropic preconsolidation stress in saturated conditions, pc is a reference pressure, is the elastic compressibility in the v-ln p plane while (0) and (s) are the slopes of the virgin compression lines in saturated conditions and at suction s, respectively. The slope of the virgin compression line (s) is related to suction through the following expression:
( s ) (0) (1 r )exp s r
(4)
where r is a constant setting the range of possible slopes when suction is increased from zero to infinity and is a parameter controlling the rate at which the slope of the virgin compression line changes with suction. Isotropic hardening is controlled by plastic volumetric strain only:
* dp0 1 e d vp * p0 0
(5)
Currently, laboratory tests still remain the most common route for defining parameter values in BBM. Typical laboratory tests include a combination of suction-controlled oedometeric compression, triaxial loading and direct shearing, which enable exploration of a variety of stress paths. Laboratory tests are, however, laborious, timeconsuming, costly and require fairly sophisticated equipment due to the need of controlling suction. Furthermore, uncertainties arise due to the disturbance of the samples retrieved from the field, which may no longer represent the in-situ soil. Due to these drawbacks, in-situ tests, and particularly pressurementer tests, are commonly seen as providing more useful information than laboratory experiments and are therefore increasingly employed for the determination of engineering parameters in unsaturated soils. Figure 1 shows three experimental pressuremeter curves obtained by Schnaid et al. (2004) during tests in a Brazilian unsaturated residual soil. Schnaid et al. (2004) also carried out simultaneous measurements of soil suction by using field tensiometers installed at variable radial distances from the pressuremeter borehole. These measurements showed that the in-situ suction is little altered by the expansion of the pressuremeter and the entire
where e is the void ratio and vp is the plastic volumetric strain. A non-associative flow rule relates shear and volumetric plastic strains according to: d sp 2q 2 p d v M 2 p ps p0 (6)
where sp is the plastic shear strain and is a constant related to M, , and (0). e Elastic volumetric and shear strains, i.e. v and e s , are given by:
e d v
d p s ds p s patm
(7)
Figure 1. Typical experimental pressuremeter curves together with numerical simulations.
1002
test can be regarded as taking place at constant suction. The suction values measured in the field during the three pressuremeter tests are reported in Figure 1. Figure 1 also shows typical curves from finite element simulations of pressuremeter tests at different values of suction by employing BBM. The good match between experiments and simulations confirms the potential for inverse analysis of pressuremeter data. In this paper, some preliminary results are presented from a wider research aimed at selecting parameter values in BBM by matching finite element simulations of pressuremeter tests to experimental data. This can be regarded as the solution of an optimization problem, which is formulated as: minimize subject to f x
like genetic algorithm or particle swarm optimization and demonstrates faster convergence. DE contains an initialization stage and an evolutionary stage with three evolution operators: mutation, crossover, and selection. The pseudo code of DE is outlined below.
cs,i cm,i
i 1
(9) (10)
x min x x max
where f(x) is the objective function, which measures the quality of the match between experimental and simulated curves (the value of f(x) is termed fitness in evolutionary algorithms). The value of the objective function depends on the values of model parameters collected in the vector x. In Equation 9 and 10, im is the cavity strain obtained from experimental data, is is the cavity strain from finite element simulation at the same applied pressure as im , xmin and xmax are the lower and upper bounds of the parameters vector x, N is the total number of experimental points used for calculating the objective function. In this work, the unknowns are the n BBM parameters to be identified, which are collected in a n-dimensional vector of floating-point numbers: x x1, , xi , , xn
T
(11)
where xi is the ith BBM parameter to be defined. In the next section, the original differential evolution algorithm is briefly described together with a novel parallel asynchronous version developed during this work. 3 3.1 PARALLEL DIFFERENTIAL EVOLUTION Original differential evolution algorithm
During the initialization stage, a set of NP individuals (i.e. a set of NP vectors x) is generated randomly and uniformly distributed across the search space in order to create a diversified population. After initialization, the algorithm proceeds in an iterative manner by producing new generations from the old population of NP vectors. Three operators are applied to the population during the evolutionary stage in the order of mutation, crossover, and selection. The random mutation operator of DE, which makes the population robust, generates a trial vector by adding a weighted difference of vectors to a target vector. The most popular mutation operator is defined as: v i , m x r 0, m F x r1, m x r 2, m
(12)
Differential Evolution (DE) is a simple and powerful population-based evolutionary algorithm proposed by Storn & Price (1995). The algorithm uses a floating-point encoding strategy, has fewer parameters than other evolutionary algorithms
where m is the generation (or iteration) number; i is the individual index; r0, r1, and r2 are random indices in the range of [1,Np] and r0 r1 r2; vi,m is the trial vector; F is a positive scale factor usually taken a value between (0,1). A uniform crossover operation is then applied to increase the population diversity: vi, j , m if ui , m x i, j , m rand () CR otherwise (13)
1003
where CR[0,1] is a user-defined crossover probability and j[0,n] is the index of a specific component in the n-dimensional vector x. The selection operator specifies which trial vector enters into the next generation. In DE, the trial vectors will replace the parents if their fitness is smaller than the parent vectors so that individuals always improve from one generation to the next. The selection criterion can be expressed as follows: u if i, m xi , m 1 x i, m 3.2
population (clients) Client 1 Client 2

...
Client n
Data back Request for service Server
Data back Request for service
Figure 2.
Client-server model for parallel DE.
f u i , m f xi , m (14) otherwise

Parallel asynchronous differential evolution algorithm
While DE is effective in solving real-world problems, its computational cost is remarkably high due to the large number of objective function evaluations. Millions of evaluations of the objective function have to be conducted to attain the global optimum. In order to reduce computational costs, evolutionary algorithms like DE can be implemented in a parallel form, where each individual of the population performs consecutive evaluations of the objective function independently from other individuals. To make full use of this feature, individuals are assigned to separate processors where function evaluations are carried out in parallel. Parallelization can be achieved in either a synchronous or asynchronous manner. In the synchronous model, DE moves to the next generation only when all individuals have completed the current one. A drawback of this model is that the slowest individual determines the speed of the algorithm and, in non-linear simulation-based optimization, the computational time can vary significantly between individuals. In contrast, in the asynchronous model, each individual moves to the next iteration immediately after it finishes the current one without having to wait for other individuals. This feature is particularly advantageous when load unbalances between individuals are very large as no processor remains idle while other processors are busy. The client-server model (Zhang et al. 2009a, 2009b) is adopted for the implementation of the asynchronous approach as shown in Figure 2. The client-server model consists of two parts: the server and the particle clients. The server resides on one processor, which is the centre for data sharing and job coordination. The server stores individuals data, listens to clients queries for information or data, processes the queries and returns data back
to the clients. The server is also in charge of termination of all clients when convergence criteria are met. The clients are instead responsible for undertaking the actual numerical simulations and, hence, producing new generations of individuals. Each client communicates with the server, and there is no communication between clients. The number of clients depends on the population size. The Message Passing Interface (MPI) (Snir et al. 1996) is chosen as the parallel programming library. There are two programs which reside on the server and clients separately. The algorithm for the server program is described as follows:
MPI_Isend and MPI_Irecv are non-blocking MPI communication functions, which suit the asynchronous algorithm. The algorithm for the client program is described as follows:
1004
VALIDATION EXAMPLE
The above parallel asynchronous DE algorithm is here used for selecting parameter values in BBM. The parallel computer Shenteng7000, which in 2008 ranked 19th among the top 500 supercomputers in the world, was used for the computation. It is a Linux cluster hosted at Supercomputing Center of the Chinese Academy of Sciences in Beijing, China. There are 1288 IBM HS21SM blade servers. Each blade server has 2 Intel XeonE5450 quadcore processors with 32GB memory. The Shenteng7000supercomputers peak performance can reach 157 trillion floating point operations per second, with actual speed exceeding 106.5 trillion floating point operations per second. In order to validate the effectiveness of the proposed algorithm, three constant-suction finite element simulations of pressuremeter tests, produced by using a known set of parameter values in BBM, were taken as the experimental data. The optimization algorithm was then tested to check whether the same set of parameter values could be found via the optimization procedure. The algorithm was used to find the values of eight BBM parameters, i.e. x [G, M, k, , (0), r, , pc]T, by exploring the 8-dimensional search space defined by the parameter ranges given in Table 1. This wide range of the searched parameters is used to cover a large variety of soils and demonstrate the performance of the search algorithm. The numerical example in Gallipoli 2000 was modified as a validation example in this paper. During the creation of the experimental data, the values of BBM parameters were chosen equal
to x [3.019e6 kPa, 0.9, 0.5, 0.025, 0.13, 1.5, 1.0e-5 Pa1, 2.0e6 Pa). This model parameters values were determined using compacted unsaturated Speswhite Kaolin (Gallipoli 2000). In addition, the value of the elastic compressibility index s was taken equal to 0.02 in all simulations. The initial vertical and horizontal net stresses were equal to 60 kPa, while the initial value of the hardening parameter p0* was set to 15.6 kPa in all finite element simulations. This initial status is corresponded to a normally consolidated condition. Pressuremeter tests were simulated by a 2D axisymmetric FE model using eight-noded quadrilateral elements with pore water pressure and pore air pressure on the corner nodes. The cavity pressure was applied incrementally in steps of 10 kPa. The total number of loading steps was 50. The three experimental curves at suction levels of 0 kPa, 100 kPa, and 200 kPa, generated by using the predefined set of BBM parameter values, are depicted in Figure 3 while the specifications of the DE algorithm are listed in Table 2. The result of the search is summarized in Table 3 where the lowest values of the objective function (fitness) are shown, together with the corresponding parameter vector, for different cumulative numbers of Function Evaluations (FEs). Inspection of Table 3 indicates that the parameter values found by the algorithm approach very closely the target values after approximately 9000 evaluations of the objective function and
Table 1.
BBM parameters and their ranges. Minimum value 1.0e5 0.7 0.2 0.01 0.06 0.9 1.0e-6 1.0e4 Maximum value 2.0e7 1.5 0.8 0.1 0.25 1.8 8.0e-4 2.0e7
Parameters G (kPa) M K (0) R (Pa1) pc (Pa)
Figure 3. Pressuremeter curves generated by using predefined parameters in BBM. Table 2. Specifications of DE. Value 120 0.7 0.5 4.0e-3
Parameters Population size F CR Convergence tolerance
1005
Table 3. BBM parameter values at different stages during the search. FEs 1 538 1012 2208 3994 4999 6048 7247 7963 8898 9028 Fitness 2.167e0 4.639e-1 1.645e-1 7.200e-2 3.691e-2 3.226e-2 2.229e-2 1.302e-2 7.224e-3 3.986e-3 3.451e-3 Unknown vector x [2.570e7, 1.015, 0.701, 0.027, 0.142, 1.178, 2.865e-5, 4.208e6] [1.523e6, 0.934, 0.515, 0.011, 0.121, 1.140, 2.531e-5, 3.553e4] [3.512e6, 0.957, 0.378, 0.029, 0.155, 1.166, 4.408e-5, 4.755e4] [2.732e6, 0.942, 0.365, 0.024, 0.143, 1.113, 3.327e-5, 9.780e4] [2.852e6, 0.931, 0.379, 0.028, 0.141, 1.172, 1.524e-5, 1.894e6] [2.821e6, 0.951, 0.386, 0.026, 0.146, 1.183, 1.425e-5, 2.316e6] [3.016e6, 0.920, 0.456, 0.022, 0.134, 1.365, 1.145e-5, 5.280e6] [3.048e6, 0.915, 0.474, 0.024, 0.134, 1.420, 1.048e-5, 3.014e6] [3.001e6, 0.905, 0.496, 0.025, 0.131, 1.480, 1.027e-5, 2.481e6] [3.049e6, 0.902, 0.491, 0.025, 0.131, 1.477, 1.014e-5, 6.434e5] [3.019e6, 0.895, 0.507, 0.026, 0.129, 1.519, 9.928e-6, 1.993e6]
convergence properties of the algorithm. Inspectioon of Figure 4 indicates that fast convergence occurs during the early stages of the search but the decrease of the objective function becomes significantly slower towards the middle stages of the search process. Figure 5 also shows a near-linear relationship between elapsed time and cumulative number of function evaluations. 5
Figure 4. Convergence properties of parallel asynchronous DE.
CONCLUSIONS
Figure 5. Dependency of elapsed time on cumulative number of function evaluations for parallel asynchronous DE.
an elapsed time of about 5.0 hours. The average time for a single objective function evaluation was 116.7 seconds and it can therefore be estimated that, if a serial algorithm was used, the computation time could be over 292.7 hours. This confirms the advantages of the parallel algorithm, in terms of computational efficiency, with respect to the standard serial algorithm. A semi-logarithmic plot of f(x) versus the cumulative number of Function Evaluations (FEs) is shown in Figure 4, which demonstrates the
The paper presents a parallel asynchronous version of a differential evolution optimization algorithm for finding parameter values in BBM by inverse analysis of the results from pressuremeter tests. A client-server model is devised for the asynchronous parallel implementation of the algorithm. This has been validated through application to fictitious pressuremeter tests, generated by finite element simulations using a known set of BBM parameter values. By following the proposed methodology, the values of eight BBM parameters were simultaneously identified with very good accuracy from the analysis of the fictitious pressuremeter data. The proposed algorithm was run on the computer cluster Shenteng7000 at the Supercomputing Center of the Chinese Academy of Sciences in Beijing, China. Results show that computational time was reduced significantly by parallelization in comparison with a standard sequential implementation. ACKNOWLEDGEMENTS The authors gratefully acknowledge the Hundred Talents Programme of the Chinese Academy of Sciences and the financial support of the UK Engineering and Physical Sciences Research Council (EPSRC) through grant ref. EP/C526627/1.
1006
REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3):405430. Gallipoli, D. 2000. Constitutive and numerical modelling of unsaturated soils. Doctoral thesis, University of Glasgow. Schnaid, F., de Oliveira, L. & Gehling, W. 2004. Unsaturated Constitutive Surfaces from Pressuremeter Tests, Journal of Geotechnical and Geoenvironmental Engineering 130(2):174185. Snir, M., Otto, S., Huss-Lederman, S., Walker, D. & Dongarra, J. 1996. MPI: The Complete Reference. Cambridge:MIT Press. Storn, R. 1999. System Design by Constraint Adaptation and Differential Evolution, IEEE Transactions on Evolutionary Computation 3(1):2234.
Storn, R. & Price, K. 1997. Differential EvolutionA Simple and Efficient Heuristic for Global Optimization Over Continuous Spaces, Journal of Global Optimization 11(4): 341359. Wheeler, S.J., Gallipoli, D. & Karstunen, M. 2002. Comments on use of the Barcelona basic model for unsaturated soils. Int J Numer Anal Methods Geomech 26: 15611571. Zhang, Y., Gallipoli, D. & Augarde, C.E. 2009a. Simulation-based calibration of geotechnical parameters using parallel hybrid moving boundary particle swarm optimization. Computers and Geotechnics 36 (4): 604615. Zhang, Y., Gallipoli, G. & Augarde, C.E. 2009b. Parallel Hybrid Particle Swarm Optimization and Applications in Geotechnical Engineering. In Z. Cai et al. (Eds.): ISICA 2009, LNCS 5821. Berlin:Springer.
1007
Use of results from undrained tests for the constitutive modeling of unsaturated soils
Xiong Zhang
Department of Civil and Environmental Engineering, University of Alaska Fairbanks, Fairbanks, AK, USA
ABSTRACT: At present, Controlled Suction Tests with Simple Stress Paths (CSTs/SSP) are commonly used to characterize unsaturated soil behavior and develop constitutive models for unsaturated soils. This kind of tests several disadvantages: it requires advanced laboratory equipment, usually laborious, timeconsuming due to lengthy equilibrium time. It is not uncommon for a simple isotopic CST/SSP usually takes several weeks or months. In addition, each CST/SSP can only provide limited information and is unable to capture the coupled behavior of the soil. Due to the lengthy testing process, sometimes the test results, especially for the water content change, are not reliable or simply not available. As a result, CSTs/SSP cannot be justified for routine engineering purpose and significantly limited the application of unsaturated soil mechanics in engineering practice. This paper discusses use of results from undrained tests for the constitutive modeling of unsaturated soils. 1 INTRODUCTION
s s3 s2 s1 (a) s p
Alonso et al. (1990) proposed the first elasto-plastic model for unsaturated soil, which is later called the Barcelona Basic Model (BBM). Since then many elasto-plastic models have been developed. Behavior of unsaturated soils is very complicated. As a result, in the development of constitutive models for unsaturated soils, Controlled Suction Tests with Simple Stress Paths (CSTs/SSP) are commonly used to characterize unsaturated soil behavior for the constitutive modeling purpose. In such tests, only one stress variable, either p or s or shear stress q, is changed, while the other stress variables remain unchanged to eliminate the coupling effects between different stress variables and make the test results easy to analyze. Figure 1 shows the stress paths of controlled-suction tests to determine the model parameters for the BBM proposed by Alonso et al. (1990). CST/SSP requires advanced laboratory equipment. Three methods are commonly used for suction control: the axis translation technique (Hilf 1956), the osmotic technique (Delage et al. 1992; Dineen and Burland 1995), and humidity control with salt solutions (Al Mukhtar et al. 1993). Development of suction-controlled triaxial apparatus using any of the three techniques involves major modifications of traditional triaxial apparatus for saturated soils (e.g., Sivakumar 1993; Cui and Delage 1996; Mun 2005). CST/SSP is usually laborious to perform due to extra work related to control of suction and measurements of changes in the water content.
(b) q s
(c)
Figure 1. Proposed stress paths to determine model parameters for the BBM (Alonso et al. 1990).
CST/SSP is time-consuming due to lengthy equilibrium time. It takes about five days to two weeks to bring a soil to a specific initial condition (e.g. Sivakumar 1993; Sharma 1998; Hoyos 1998). It takes a much longer time to perform a CST/SSP.
1009
A simple isotopic CST/SSP usually takes several weeks or months. For example, it took Sivakumar (1993) 959 days to complete 30 CSTs/SSP for a compacted speswhite kaolin, with an average of 32 days/test. It took Sharma (1998) 801 days to complete 20 CSTs/SSP for two compacted expansive soils under isotropic conditions with an average of 40 days/test. Although problems associated with unsaturated soils exist all over the world, CST/SSPs cannot be justifiable for routine engineering projects. It is therefore important to develop alternative, cost-effective ways to characterize elasto-plastic behavior for unsaturated soils. This paper discusses possible use of results from undrained tests for the constitutive modeling of unsaturated soils. 2 USE OF UNDRAINED TESTS FOR THE CONSTITUTIVE MODELING OF UNSATURATED SOILS
short testing period, and the test results are more reliable. Since a constitutive model must be able to predict soil behavior under any condition, including undrained loading conditions, it is both theoretically reasonable and practically feasible to replace CST/SSP with undrained tests for the constitutive modeling of unsaturated soils. The only problem is that, at present, there is no method available to take full advantage of results from undrained tests for the constitutive modeling purposes. The newly proposed modified state surface approach (Zhang and Lytton 2009a and 2009b) can be used for the purpose. Due to the space limit, this paper only shows the application of the isotropic conditions.
MODIFIED STATE SURFACE APPROCH
A constitutive model has to be able to predict soil behavior under all possible stress paths. On the other hand, results from any test reflect constitutive behavior of the soil along the specific stress path in the test and should be able to be used for constitutive modeling purposes. Compared with CST/SSP, some tests for unsaturated soils are easier to perform. An example is the unconfined compression test, which has been routinely performed in industrial labs and undergraduate classrooms. Often, an unsaturated soil specimen is used, although the test results are usually analyzed based on the theory of saturated soil mechanics. The test can be done in less than one hour, using any routine triaxial test apparatus for saturated soils. From the viewpoint of constitutive modeling, an unconfined compression test is an undrained triaxial test with simultaneous changes of p, q, and s. The variations of suction during the test are not externally specified, but internally determined by the coupling requirement that water content remains constant. In an undrained test, suction in the soil specimen will instantaneously reach equilibrium everywhere, which has been experimentally verified (Bishop et al. 1960; Rahardjo et al. 2004; Thu et al. 2006). Although having a more complicated stress path, an undrained test has the following advantages over CST/ SSP: The tests are fast, cheap, easy to perform, and more representative of field conditions. With simultaneous change of p, s, and q, each test can provide at least three times the information as that from a CST/SSP, which means fewer tests are needed to provide the same amount of information. There is much less possibility of disturbance due to a
Zhang and Lytton (2009a and 2009b) proposed a Modified State Surface Approach (MSSA) for the constitutive modeling of unsaturated soils. The principle of the MSSA can be illustrated by Fig. 2 (Zhang et al. 2010). Fig. 2a shows an isotropic loading-unloading-reloading stress path under an arbitrary constant suction. The soil specimen has an initial condition of point D and an initial yield * curve of LC1 with a preconsolidation stress of p0 at s 0 kPa. The yield stress at s s2 is p0 at point E. The soil is loaded from D to E to V, unloaded from V to D, and then reloaded to F. Fig. 2b illustrates a typical soil response in the v-lnp plane when the hysteresis is neglected. The following observations can be made from the process: 1. Regardless of stress path and stress history, the shape and position of the virgin consolidation curve EVF are always the same for the soil in the v-lnp plane. Plastic loading only changes the range of the virgin consolidation curve. For example, the initial virgin consolidation curve for the soil is EVF. After loading from D to E to V, the virgin consolidation curve for the soil is VF. 2. During an elastic loading or unloading process, for example, from D to E, from V to D, or from D to V, the shape and position of the unloading-reloading curve remain unchanged in the v-lnp plane. 3. During plastic loading process, the shape of the unloading-reloading curve remains unchanged in the v-lnp plane ( is a constant), but its position will change. Specifically, the unloadingreloading curve will move downward in parallel with the original unloading-reloading curve. The range of the elastic zone also expands due to the increase in the preconsolidation stress.
1010
When there is a plastic loading, similar to the previous discussion, the shape and position of the virgin consolidation curve EVF are always the same in the v-p-s space regardless the previous stress path and stress history. It can also be proven that when there is unloading, any unloading stress path must fall on a lower elastic surface parallel to surface ABEHGDA. For example, if there is an unloading process from V, the unloading stress path VD must fall on the surface DUVW, which is in parallel with surface ABEHGDA. Multiple normal compression tests can be performed at any arbitrary suction levels. Consequently, the virgin consolidation curves at different suction levels will also form a plastic surface in the v-p-s space such as BEHUXYZ in Fig. 2c. The location and shape of the plastic surface will always remain the same in the v-p-s space and the plastic surface is unique. The uniqueness of the state boundary surface is a fundamental assumption made in the constitutive modeling of elasto-plastic soil behavior. The uniqueness of the state boundary surface for unsaturated soils has been experimentally verified by Wheeler and Sivakumar (1995). The plastic surface BEHUXYZ in Fig. 2c is actually the shape of the state boundary surface when the shear stress is equal to zero (isotropic conditions). Similarly, in the v-p-s space, the following criteria can be made for the elastic and plastic surfaces: 1. The shape and position of the plastic surface BEHUVWX are always the same for the soil in the v-p-s space. Plastic loading only changes the range of the plastic surface. 2. During an elastic loading or unloading process, the shape and position of the unloadingreloading elastic surface and the plastic surface remain unchanged in the v-p-s space. The volume change of any isotropic elastic loading or unloading stress path must fall on the elastic surface in the v-p-s space. 3. During a plastic loading process, the shape of the unloading-reloading elastic surface remains unchanged ( and s are constants), but its position will change. Specifically, the unloadingreloading elastic surface will move downward in parallel with the original unloading-reloading elastic surface. The volume change of any isotropic plastic loading stress path must fall on the plastic surface in the v-p-s space. 4. The yield curve is the interception of the unloading-reloading elastic surface and the plastic surface. The above criteria can be used to represent unsaturated soil behavior under isotropic loading conditions including stress path independency. Zhang and Lytton (2009a) derived the mathematical
Figure 2. Principle of the MSSA (Zhang et al. 2010). (a) Stress paths for normal compression tests. (b) Volume change for a normal compression test at constant suction. (c) Three dimensional representation of volume change of the soil.
4. The yield point is the interception of the unloading-reloading curve and the virgin consolidation curve. Let us consider two other arbitrary stress paths from D to E, i.e. 1 and 2 as shown in Fig. 2a, in the elastic zone. Since stress paths 1 and 2 are in the elastic zone, the specific volume changes are stress path independent and the results for the two stress paths should be the same. Zhang and Lytton (2009a and 2009b) proved that volume of the soil in the elastic zone is a surface in the v-p-s space. Fig. 2c shows the specific volume change for the stress paths in Fig. 2a in the v-p-s space. All the elastic volume changes such as stress paths 1 and 2 and DE are in the same surface of ABEHGDA.
1011
expressions of the elastic surfaces for the BBM and successfully used the MSSA to represent many unsaturated soil behavior including the stress path independency under isotropic conditions. Fig. 3 shows the elastic and plastic surfaces used in the BBM. They include an elastic surface AEFG and the plastic surface which is made up of two parts: a plastic collapsible surface EFSCDHR and a plastic expansive surface GFSB. ve C1 ln p s ln(s pat) p s pat v N 0 s ln s ln c pat p e C3 ln p s ln(s pat) (1) (2) (3)
ANALYSIS OF AN UNDRAINED TEST USING MSSA
where, (s) (0)[(1 r)exp(s) r]; r parameter controlling the slope of the virgin compression line, parameter controlling the slope of the virgin compression line for s0, (0) slope of the virgin compression line associated with the mean net stress at saturation (s 0); pc reference stress, N(0) a constant, s slope of the virgin compression line associated with soil suction, and C3 constant. Note that Fig. 3 was plotted in scale based on the real data of Case 3 in Alonso et al. (1990), with soil parameters of (0) 0.2, 0.02, s 0.08, s 0.008, r 0.75, 12.5 MPa-1, and pc 0.10 MPa. From the initial conditions of the soil, i.e. p 0.15 MPa, s 0 MPa, and v 1.9, C1, N(0), and C3 in Equations 1, 2, and 3 were calculated as 0.844, 1.033, and 0.694, respectively.
Let us qualitatively analyze an isotropic undrained test using the MSSA as shown in Figure 4. Figure 4a schematically shows the stress path of an undrained test. A soil specimen has an initial condition of point G in the elastic zone. The corresponding initial position of the LC yield curve is represented by HEB (LC1). Under an undrained loading process, the soil followed the stress path from G to E in the elastic zone, followed by a plastic undrained loading from E to C. During this process, the suction decreases due to the increase in the mechanical stress. Finally there is an undrained unloading from C to G. Due to the irrecoverable plastic deformation, the final suction is less than the original suction as shown in Figure 4a. Figure 4b shows the corresponding changes in the specific volume is plotted in the v-p-s space. According to the MSSA, GF is on the initial elastic surface ADGHEB, and FC is on the plastic surface BEHIFC. The results from the undrained test do provide information regarding the shape of the elastic and plastic surfaces and can be used to determine the model parameters as shown in Eqs. (1) and (2) if the BBM model is used to simulate the soil behavior. For isotropic condition, only one
s s3 G E C p (a) v G D A s G B s E H I s3 s2 s1 (b) ln p Yield points H LC1 I LC2
s2 s1
D G A p1 B
F C
Figure 3. Shape of the state boundary surface for the BBM (Zhang and Lytton 2009a).l.
Figure 4. Analysis of an undrained test using the MSSA. (a) Stress path for an undrained loading and unloading test. (b) Changes in the specific volume in the v-p-s space for the undrained test.
1012
undrained test is sufficient to determine the model parameters in the BBM. If controlled suction tests with simple stress paths (CSTs/SSP) are used, four tests are needed fro the same purpose as shown in Figures 1a and 1b. Figure 5 shows the results of an undrained test reported by Sun et al. (2008) on a compacted Pearl Clay. Figure 5a shows the stress path for the undrained test. Figure 5b shows changes in the specific volume during the undrained test. Point E in Figure 5 is the yield point corresponding to E as shown in Figure 4. The data points in the range from G to E in Figure 5 are used to best fit Eq. 1 using the leastsquares regression. It was found that C1 1.175, 0.0258, and s 0.0285. The data points in the range from E to C in Figure 5 are used to best fit Eq. 2 using the least-squares regression. It was found that N(0) 1.0697, (0) 0.47778, r 6.452, 0.8735 MPa-1, and pc 0.5216 MPa. Figure 6 shows comparisons between experimental results reported by Sun et al. (2008) and the prediction based upon the above model parameters. It can be seen that the predicted changes in the specific volume match the measured data very well. Only one undrained test is needed to determine the model parameters for the BBM if the MSSA is used. By contrast, if Controlled Suction Tests with Simple Stress Paths (CSTs/SSP) are
Figure 6. Comparison between experimental and predicted data.
used, four tests are needed fro the same purpose as shown in Figures 1a and 1b. This means an undrained test can provide more information for determining the model parameters in the BBM. In addition, compared with controlled suction test, it takes much less time to perform an undrained test. Consequently, use of results from undrained tests for the constitutive modeling of unsaturated soils will significantly reduce the time and efforts needed for characterizing soil behavior. Due to limited space, this paper only discusses constitutive modeling of unsaturated soil behavior under isotropic conditions. The author has extended the same approach to the triaxial stress conditions, which will be presented in another paper. 5 CONCLUSION
Based upon the newly proposed MSSA, test results from an undrained isotropic compression test is used for the constitutive modeling of unsaturated soils. Only one test is needed to determine the model parameters for the BBM and the prediction matches the observation very well. Compared with controlled-suction tests, undrained test can provide more information for the constitutive modeling purpose. In addition, the time and efforts needed for performing an undrained test is much less. It is therefore concluded that the approach proposed in this paper is a very promising, cost-effective alternative to the existing controlled-suction tests in the constitutive modeling of unsaturated soils. REFERENCES
Figure 5. Analysis of an undrained test using the MSSA. (a) Stress path for an undrained loading and unloading test. (b) Changes in the specific volume in the v-p-s space for the undrained test. Al Mukhtar, M., Robinet, J.C. & Liu, C.W. 1993. Hydro-mechanical behaviour of partially saturated low porosity clays. Engineering Fills. Clark, B.G., Jones, C.J.F.P. & Moffat, A. I. B. (Eds.). London: Thomas Telford, 8789.
1013
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Geotechnique, 40(3), 405430. Bishop, A.W., Alpan, I., Blight, G.E. & Donald, I.B. (1960). Factors controlling the shear strength of partly saturated cohesive soil. In: Proc., ASCE Research Conf. on Shear Strength of Cohesive Soils, Univ. of Colorado, Boulder, CO, pp. 503532. Cui, Y.J. & Delage, P. (1996). Yielding and plastic behaviour of an unsaturated compacted silt. Geotechnique, 46(2), 291311. Delage, P., Suraj, D., Silva, G.P.R. & Vicol, T. 1992. Suction controlled testing of non saturated soils with an osmotic consolidometer. 7th International Conference Expansive Soils, Dallas, pp. 206211. Dineen, K. & Burland, J.B. 1995. A new approach to osmotically controlled oedometer testing. In: Proc. of the 1st Conference on Unsaturated Soils Unsat 95, Vol. 2, Paris: Balkema, pp. 459465. Hilf, J.W. 1956. An Investigation of Pore-Water Pressure in Compacted Cohesive Soils. Ph.D. dissertation, Technical Memorandum No. 654, US Department of the Interior, Bureau of Reclamation, Design and Construction Division, Denver, Colorado. Hoyos, L.R. 1998. Experimental and Computational Modeling of Unsaturated Soil Behavior under True Triaxial Stress States. Ph.D. dissertation, Georgia Institute of Technology, Atlanta, Georgia. Mun, B. 2005. Unsaturated soil behaviour under monotonic and cyclic stress states. Ph.D. Dissertation, Department of Civil Engineering, Texas A&M University, College Station, TX.
Rahardjo, H., Heng, O.B. & Leong, E.C. 2004. Shear strength of a compacted residual soil from consolidated drained and the constant water content triaxial tests. Can. Geotech. J. 41, 116. Sharma, R.S. 1998. Mechanical behaviour of unsaturated highly expansive clays. D.Phil. thesis, University of Oxford, UK. Sivakumar, V. 1993. A critical state framework for unsaturated soil. Ph.D. Thesis, University of Oxford. Sun, D.A., Sheng, D., Xiang, L. & Sloan, S.W. 2008. Elastoplastic prediction of hydro-mechanical behaviour of unsaturated soils under undrianed conditions. Computers and Geotechnics, 35: 845852. Thu, T.M., Rahardjo, H. & Leong, E.C. 2006. Shear strength and pore-water pressure characteristics during constant water content triaxial tests. Journal of Geotechnical and Geoenvironmental Engineering, 132(3), 411419. Zhang, X., Liu, J. & Li, P. 2010. A new method to determine the shapes of yield curves for unsaturated soils. Journal of Geotechnical and Geoenvironmental Engineering, ASCE. Volume 136, Issue 1, pp. 239247. Zhang, X. & Lytton, R.L. 2009. A modified state surface approach on unsaturated soil behavior study (I): Basic concept. Canadian Geotechnical Journal, 46(5), 536552. Zhang, X. & Lytton, R.L. 2009. A modified state surface approach on unsaturated soil behavior study (II): General formulation. Canadian Geotechnical Journal, 46(5), 553570.
1014
Analytical and numerical modelling
A driver for the integration of coupled hydro-mechanical constitutive laws for unsaturated soils
F. Cattaneo
Universit degli Studi di Brescia, Brescia, Italy
G. Della Vecchia C. Jommi
Politecnico di Torino, Turin, Italy Politecnico di Milano, Italy
ABSTRACT: Coupling between the hydraulic and the mechanical behaviour is emerging as a fundamental interpretative tool when dealing with the response of unsaturated soils to generalised hydro-mechanical histories. Different approaches have been proposed to deal with the coupled constitutive response of soils in unsaturated conditions. All of them involve high non-linearities of the constitutive laws, and numerical integration algorithms are mandatory for their use. The formulation of a constitutive driver is presented for the numerical simulation of the local hydro-mechanical response of unsaturated soil. The formulation allows for developing numerical integration routines for the constitutive equations, ready for their straightforward implementation in finite element codes. The method is illustrated for a particular, but sufficiently general, class of retention and stress-strain models. Conditioning of the global system of governing equations is analysed, and possible definitions for residuals in its iterative solution are discussed. 1 INTRODUCTION a constitutive driver may be easily coupled with best-fit procedures to seek for the optimum set of constitutive parameters for a given soil. The formulation of the constitutive drivers proposed in the literature in general refers to saturated conditions (Bardet & Choucair 1991, Alawaji et al. 1992, Mattsson et al. 1999). Here a possible formulation for unsaturated soil models is presented. Due to hydro-mechanical coupling, the fluid mass balance equations must be solved together with the momentum balance, to provide for any type of external control. The formulation of the numerical driver was conceived to allow for a straightforward implementation of the routines developed for the integration of the non linear constitutive laws into classical finite element codes. In the following, boldface notation is used for vectors and tensors. 2 FORMULATION OF THE COUPLED PROBLEM
Many models have been proposed in the last years to simulate properly the mechanical behaviour of soils in unsaturated conditions. With reference to the class of models developed in the framework of elastoplasticity, the constitutive equations for the soil skeleton may be written in the general form (e.g. Gens et al. 2006) . . c Dc( s), (1) where c is a suitable constitutive stress rate tensor, . Dc is the tangent elastic-plastic tensor, is the total . strain rate tensor and s is a measure of that part of the strain rate which is due to suction variations. Suction, s, is defined as the difference between the gas and the liquid pressures. The choice of the appropriate constitutive stress, and of the way in which the suction induced strains are calculated, in general distinguish the different models one from the other. In the solution of a boundary value problem integration of the constituve law is local in space, and it is performed at the quadrature points of finite elements (Simo & Hughes 1998). A constitutive driver simply reproduces this operation. Its role at the development stage is to verify the model performance along any kind of load path, without the need for spatial discretisation. Besides,
In geotechnical laboratory testing, stress and strain (derived from imposed forces and displacements) may be the mechanical control variables. As for the fluid phases, mass or pressure variations can be imposed. In the following, ul and ug will indicate the pressure of the liquid and the gas phases
1017
respectively, while mw and ma represent the water mass and the air mass contained in the porous volume. Any combination of the previous quantities may characterise a specific laboratory test. From the numerical viewpoint, a standard integration procedure would imply possible mixed control of static (, ul, ug) and kinematic (, mw, ma) variables. Partitioning consequently the constitutive matrix would lead to a direct formulation for the unknown variables of the specific problem under consideration. As the driver was conceived to provide integration routines ready for their implementation in typical finite elements codes, where displacements and fluid pressures are adopted as primary variables, a strain-pressure driven approach was preferred. Therefore, the input variables for the integration routines are always strain and fluid pressure increments, for any loading path. The final values of fluid masses and stresses are the dependent variables to be found iteratively. Let vector c collect the controlling (driving) variables, and vector s the dependent ones: c := ul ug s := mw , ma (2)
laws must be introduced. The hydraulic constitutive law, i.e. the retention domain, can be expressed in a sufficently general way by a class of models linking the degree of saturation Sr to the suction s: ug ul and to the void ratio e. For such kind of models, the increment of the degree of saturation in the incremental step depends directly on suction and total volumetric strain increments. Therefore, Sr can be directly calculated from the increments of the driving variables: Sr Sr n 1 Sr n Sr(Sr n , ug , ul , ) Sr n. (5)
As for the mechanical response, the constitutive driver is able to integrate, over a general loading path, models defined in terms of different constitutive stresses. However, for the sake of brevity, only models that adopt the average soil skeleton stress (Jommi 2000), := ( ug ) Sr(ug ul ) ,
(6)
will be discussed here (in the previous expression is the Kronecker's delta). For the class of models introduced, the total stress increment can be written as: n 1 n n 1 (Sr n 1uln 1 (1 Sr n 1 )ug ) n .
n 1
and let n 1 be a time step bounded by tn and tn1. Both driving and dependent variables, c and s, are assumed to be completely known at time tn. The system of equations describing the generic laboratory constraints during the time step n 1 can be formally written following Bardet & Choucair (1991) as: s c = V. (3) The elements of the matrices and are constant during the general loading path, and can be partitioned as: E 0 := 0 l 0 0 0 S 0 0 := 0 w g 0 0 0 0 . a (4)
(7)
The constitutive stress at the end of the step is obtained by integrating the mechanical constitutive equations from the given increment of driving variables (strain and fluid pressures) and from the known increment of degree of saturation (Eq. 5):
n 1
= ( , , ul , ug , Sr ).
n
(8)
2.2 Water mass balance A representative volume element of unsaturated soil may contain water both in the liquid and in the gas phase. Neglecting the latter contribution, the water mass reads: mw = nSr w , (9)
Matrices E and S define the mechanical constraints of the test in the same way as proposed by Bardet & Choucair (1991), but involving total stress instead of effective stress. The other nonzero terms refer to fluid phases: for example, if the pressure of the liquid phase is imposed l 1 and w 0, while if the water mass increment is imposed l 0 and w 1. The constraints for the gas phase, g and a, are defined in similar way. 2.1 Soil behaviour
where n is the porosity and w is the water density. The water mass increment can be written as:
n 1 n mw = n n 1Sr n 1 w mw .
(10)
The classical exponential law may be assumed to rule the dependence of water density w on the liquid pressure: u n 1 ul0 n 1 0 w = w exp l . Bw (11)
To describe properly unsaturated soils response, both the hydraulic and the mechanical constitutive
1018
Assuming as reference state (denoted by the superscript 0) the relative atmospheric pressure (ul0 0) and a temperature of 20C, the reference water mass density is w0 998.207 kg/m3. 2.3 Air mass balance Air is present both in the gas and in the liquid phase, that is dissolved in water (superscript g and d respectively):
g d ma = ma ma .
quadratic convergence rate, if the initial guess x0 is sufficiently close to the solution x* and the Jacobian matrix is not singular (Quarteroni et al. 2007). The solution at the end of the step n 1 is found by iteration, starting from an initial guess c0 0, and repeating the following steps, until the theoretical condition R 0 is reached for the residual: 1. Integration of the retention model to find the increment of the degree of saturation
n 1,i n 1,i Sr n 1,i = Sr (uw ua ), i
(12)
Dry air in the gas phase can be written as:

g ma = n(1 Sr ) a ,
Sr i Sr n 1,i Sr n
(13)
where a is the dry air density. The latter can be written as a function of gas pressure by the ideal gas law:
n 1 a 0 = a n 1 ug patm
2. Determination of the constitutive stress, adopting a suitable integration procedure (Vaunat et al. 2000, Cattaneo et al. 2010)
n 1,i
= ( n, i, Sr i )
patm
(14)
3. Determination of the dependent variables increment by Equations (7), (10) and (18) 4. Check for the constraints si ci V = R 5. If R 0 (operatively, R TOL) the driving variables increment is updated by solving the linear system c
i 1
with a0 1.204 kg/m3, at the same reference state defined for water. The volume of air dissolved in water can be determined by the ideal gas law and Henrys law. The volume of dissolved gas Vd can be linked to the volume of liquid water Vw by the coefficient of solubility h, h Vd /Vw, (15)
s = c c
i
n 1,i
R
1
which depends only on temperature. At the reference temperature, T 20C, its value is 1.868 102 (Fredlund & Rahardjo 1993). Multiplying by a, with reference to the total volume VTOT: V Vd a = h w a VTOT VTOT
l ma = hnSr a .
and the iterative procedure is restarted from step 1. Step 5 requires the evaluation of the derivatives of the dependent variables increment with respect to the driving variables increment: mw s c ma ul mw ul ma ul ug mw . ug ma ug
(16)
Therefore, the air mass can be written as: ma = (1 (1 h )Sr )n a , and its increment reads
n 1 n ma = (1 (1 h )Sr n 1 )n n 1a ma .
(19)
(17)
(18)
Note that, from Equation (7), ij hk =

ij n 1 (uln 1 ug ) ij n
SOLUTION STRATEGY
hk
Sr n 1 , hk
(20)
Introducing Equations (7), (10) and (18) in Equation (3), the governing system of non-linear equations is obtained, which can be solved by Newtons method. Newtons method provides
and, since ij = ij ij , it follows that

ij
n 1 ij
hk

hk
ij Sr . Sr hk
(21)
1019
ij with respect to hk and The derivatives of Sr are the tangent consistent moduli as defined by Simo & Taylor (1985). The introduction of dissolved gas in Equation (12), besides providing a more correct description of the physical transfer mechanisms, avoids singularity of the matrix (19) for Sr 1 (e.g. Vaunat et al. 1997). In this way the algorithm proposed is able to simulate naturally the transition from the saturated to the unsaturated condition. Water vapour was neglected here, but its introduction in the structure of the driver is rather straightforward. It is worth noting that the structure of the routines which provide the integration of the constitutive laws (steps 1 and 2 of the previous procedure) is the same as that adopted in the solution of boundary value problems with a classical displacement-pressure based finite element approach. Therefore, at the end of the model development stage, they can be transferred to a FE code in a straightforward way. Finally, the solution procedure may be interfaced to an external routine, able to seek for the optimum set of constitutive parameters via an identification algorithm. For the sake of brevity, the latter topic is not discussed here. 4 NUMERICAL EVALUATION
Figure 1.
Stress path of the first test.
To discuss the performance of the constitutive driver, the results of two different ideal tests are shown. In the discussion two aspects are evaluated, namely the condition number of the matrix governing the iteration procedure (step 5 of the solution procedure), s c
n 1,i
,
(22)
Figure 2.
Oedometer test: condition number.
and possible alternative choices to bound the residual R. Test 1 is an oedometer compression path on an initially unsaturated soil (0A), followed by a suction controlled saturation stage at constant vertical stress (AB) (Fig. 1). The second path is a standard triaxial strain-controlled compression test performed at constant water mass and constant gas pressure. The hydro-mechanical model proposed by Romero & Jommi (2008) is adopted to describe the behaviour of the soil, with the constitutive parameters calibrated for Boom clay. The condition number of the matrix in Equation (22) is shown as a function of the cumulated number of iterations in Figures 2 and 3 for the two tests, respectively. This matrix, introduced to update the increments of the driving variables,
Figure 3.
Triaxial test: condition number.
1020
is composed of terms having significantly different orders of magnitude. This difference leads to very high condition numbers, which could possibly give serious problems in terms of accuracy of the solution, whenever iterative algorithms were adopted. It has to be remarked that the condition number tends to decrease during the iterative procedure of a single loading step, but it increases again at the beginning of the following loading step, as the bandwidth of the condition number in the figures evidences. Nonetheless, the problems associated to high condition number can be easily overcome, by adopting standard scaling numerical algorithms. In Figures 2 and 3 the result of diagonal scaling via matrices Dr and Dc, performed with standard routines of the LAPACK library, is shown. The linear system Ax b is scaled into: ( Dr ADc )y Dr b , with
1 y Dc x,
be modified by multiplying the relative tolerance RTOL by the absolute values of V:

2 1 N Ri r2 a = N i =1 ATOL RTOL Vi 1/2
<1,
(26)
where the vector V is defined as: V : c n s n . (27)
Finally, a residual measure, r3, may be simply evaluated by scaling it with the imposed constraint increments: r3 = | Ri | TOL. i =1 | Vi |
N
(28)
(23)
giving condition numbers significantly lower, which eliminate any problem in the solution accuracy. A key point arising from the structure of the nonlinear system is the definition of a suitable measure for the residual R, which should be sufficiently general and effective for any combination of external driving and dependent variables. The exit condition of the iterative algorithm dominates the velocity and the accuracy of the solution of the nonlinear system via the Newtons method. In order to obtain satisfactory solutions, both in terms of convergence and accuracy, different possibilities were exploited. A first simple possibility is to prescribe the residual measure r1, defined as the Euclidean Norm of the vector R, to be less than a fixed tolerance TOL: r1 R TOL. (24)
The convergence rate of the different definitions in the solution of a single step is shown in Figures 4 and 5 for the two tests, respectively. It is worth noting that any of the definitions leads to quadratic convergence, characteristic of the Newtons method. The results suggest that the definition r2a seems to be the more efficient one. Nonetheless, the simple definition r1, based on the norm of the residual, provides comparable convergence characteristics, in spite of the different order of magnitude of the controlled variables. The rate of convergence of relevant dependent variables to the external constraints in the iterative solution of a single step is presented in Figures 68. It can be noticed that the convergence rate is almost independent from the residual definition, and that the accuracy depends only on the number of iterations. Extremely low errors can be reached for any variable, in some cases dominated by the machine precision (e.g. the water mass increment evolution presented in Figure 7).
Taking into account that the external controls may prescribe constraints differing by several orders of magnitude, a second possible definition for the residual measure, r2, was attempted, playing on both the relative and the absolute errors of the relevant variable increments:
2 1 N Ri r2 = N i =1 ATOL RTOL Vi 1/2
<1.
(25)
The latter definition, which is suggested in some finite element codes, may loose its significance when the imposed constraint Vi is equal to zero. In this case the contribution of the i-th variable to the residual is due only to the absolute tolerance ATOL. In order to avoid this problem, the definition may
Figure 4.
Oedometer test: evolution of the residuals.
1021
Figure 5.
Triaxial test: evolution of the residuals. Figure 8. Triaxial test: convergence to null radial stress increment condition.
CONCLUSIONS
Figure 6. Oedometer test: convergence to the imposed vertical stress.
Figure 7. Triaxial test: convergence to null water mass increment condition.
A constitutive driver to integrate the hydromechanical response for rate-type problems involving unsaturated soils was presented. Among all the possible control variables in unsaturated soil testing, the driver was written considering the strain and the fluid pressures as driving variables. In this way the routines written for the integration of the constitutive equations are ready to be implemented in finite elements codes. Retention models involving a dependence of the degree of saturation on the suction and on the void ratio, and stress-strain relations based on the average skeleton stress as constitutive stress were considered here. For this kind of models the integration technique was explicitly described. Nonetheless, the structure of the driver allows for a straightforward extension to constitutive models conceived following different approaches. The simulation of two hydro-mechanical paths were presented, in order to evaluate possible critical situations arising from high condition number and from the choice of suitable residual measures. Numerical results show that the conditioning problem can be tackled by a simple matrix scaling. A simple definition of the residual as the norm of the calculated variables proved to be sufficiently effective, although the introduction of separate relative and absolute tolerances seems to enhance the convergence rate. Nonetheless, the quadratic convergence rate of Newtons method, as well as the rate of convergence of the single variables to the imposed value, do not depend on the definition of the residual, but only on the correct definition of the consistent tangent matrix.
1022
REFERENCES
Alawaji, H., Runesson, K., Sture, S. & Axelsson, K. 1992. Implicit integration in soil plasticity under mixed control for drained and undrained response. International Journal for Numerical and Analytical Methods in Geomechanics 16(10): 737756. Bardet, J.P. & Choucair, H. 1993. A linearized integration technique for incremental constitutive equations. International Journal for Numerical and Analytical Methods in Geomechanics 15(1): 119. Cattaneo, F., Della Vecchia, G., Jommi, C. & Maffioli, G. 2010. A comparison between numerical integration algorithms for unsaturated soils constitutive models. Accepted for publication. Proc. 7th European Conference on Numerical Methods in Geotechnical Engineering, Trondheim. Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils. New York: John Wiley & Sons. Gens, A., Sanchez, M. & Sheng, D. 2006. On constitutive modelling of unsaturated soils. Acta Geotechnica 1(3): 137147. Jommi, C. 2000. Remarks on the constitutive modelling of unsaturated soils. In A. Tarantino & C. Mancuso (eds.), Experimental evidence and theoretical approaches in unsaturated soils: 139153. Rotterdam: Balkema.
Mattsson, H., Axelsson, K. & Klisinski, M. 1999. On a constitutive driver as a useful tool in soil plasticity. Advances in Engineering Software 30(8): 511528. Quarteroni, A., Sacco, R. & Saleri, F. 2007. Numerical Mathematics (2 ed.). Berlin: Springer-Verlag. Romero, E. & Jommi, C. 2008. An insight into the role of hydraulic history on the volume changes of anisotropic clayey soils. Water Res. Res. 44: doi:10.1029/2007 WR006558. Simo, J.C. & Taylor, R.L. 1985. Consistent tangent operators for rate-independent elastoplasticity. Computer Methods in Applied Mechanics and Engineering 48(1): 101118. Simo, J.C. & Hughes, T.J.R. 1998. Computational Inelasticity. Berlin: Springer-Verlag. Vaunat, J., Cante, J.C., Ledesma, A. & Gens, A. 2000. A stress point algorithm for an elastoplastic model in unsaturated soils. International Journal of Plasticity 16: 121141. Vaunat, J., Gens, A. & Jommi, C. 1997. A strategy for numerical analysis of the transition between saturated and unsaturated flow conditions. In Pande & Pietruszczak (eds.), Numerical Models in Geomechanics NUMOG VI: 297302. Rotterdam: Balkema.
1023
Numerical simulation of volume change of a swelling soil in the northeast of Brazil

L.M. Costa, S.R.M. Ferreira, L. do N. Guimares & I.D.S. Pontes Filho
Federal University of Pernambuco, Recife, Pernambuco, Brazil
ABSTRACT: The incidence of vertisoils has been observed in the semi-arid region of the northeast of Brazil One of the most important characteristic of these soils is the presence of highly expansive clay and the property of contracting and fissuring during drying process, and swelling during wetting. This paper presents a numerical analysis of the volume change behavior of a vertisoil from Petrolndia-PEBrazil, due to changes in the water content at different stress level. A coupled hydro-mechanical formulation, implemented in the computational code CODE_BRIGHT was applied to simulate conventional and suction controlled edometric tests performed with this soil. The constitutive model used is the double structure generalized plasticity model proposed by Sanchez et al. (2005). Comparison between experimental data and output results of numerical simulations shows a good agreement, checking the capability of the Model and the computer code to deal with hydro-mechanical behavior of expansive soils. 1 INTRODUCTION model proposed by Sanchez et al. (2005), based on the general framework proposed by Gens & Alonso (1992) and considering the improvements suggested by Alonso et al. (1999). Two levels of structure are considered. The macrostructure behavior is described by the Barcelona Basic Model (BBM), developed by Alonso et al. (1990). Other mechanisms, not included in the BBM, which occur in the microstructure, at clay particle level, can take place in expansive soils inducing plastic strains. So the formulation presents the definition of laws for the macrostructural level, the microstructure level and the interaction between both structural levels. 2.1 Macrostructural model BBM considers two independent stress variables, the net stress, (ij paij), and matric suction, s (pa pw). It is an elastoplastic strain-hardening model, which extends the concept of critical state for saturated soils to unsaturated conditions including a dependence of the yield surface on matric suction. The yield surface is expressed by
* f p, q, s, p0 q 2 M 2 p ps p0 p 0
One of the most important characteristic of vertisoils is the presence of high expansive clay and the property of contracting and fissuring during drying process, and swelling during wetting. The soil volumetric instability (contraction and swelling or colapse) due to changes in water content is complex and it is influenced by several factors, among them the soil type (origin and formation), climate conditions and stress state (fabric, suction and applied stress). In expansive soils, volume change due to applied stress or suction is governed by the various phenomena occurring at microstructural level, which means, in individual clay particles and their vicinity. Gens & Alonso (1992) presented a conceptual basis for modeling expansive soil, where two different levels are considered: the microstructural level, at which swelling of active minerals takes place and the macrostructural level responsible for major structural rearrangement. This paper presents an application of a double structure model, based in Gens & Alonso (1992), and implemented in a computational code to analyze the volume change characteristic, due to changes in water content of a vertisoil from Petrolndia, in Pernambuco, a state in the northeast of Brazil. 2 CONSTITUTIVE MODEL
(1) (2) (3) (4)
with p m max(pa, pw); m(1 2 3)/3 q 1 3 ps ks
The constitutive model adopted in this paper is the double structure generalized plasticity
where 1, 2 and 3 are the total principal stress, M is the slope of the critical state line, p0 is the apparent
1025
unsaturated isotropic preconsolidation stress for suction s, p0* is the saturated preconsolidation stress and k describes the increase of the apparent cohesion with suction. The net mean stress p is defined in the manner indicated in equation 2 to facilitate the transition from unsaturated to saturated states. For isotropic conditions yield states associated with suction are described by means of a yield function defined in the space (p, s), which is named the Loading-Collapse yield surface (LC). It explains the collapse upon wetting and the increase of apparent preconsolidation stress p0 with suction through the following relationship
* s p0 p0 c c p p
strain depends on a microstructural effective stress (p ) defined by p = p s (10)
0
(5)
where, is the elastic stiffness parameter against changes in p, (0) is the slope of virgin compression line for saturated isotropic loading, pc is a reference stress and (s) is the slope of virgin compression line for isotropic loading at a constant suction s defined as (s) (0)[(1 r)exp(s) r] (6) where controls the rate of stiffness increase with suction and r is a limiting value of soil stiffness for very high suction. A non associated plastic potential is defined by
* g p, q, s, p0 q 2 M 2 p ps p0 p
where is a constant. Another assumption adopted in this formulation is the hydraulic equilibrium between microstructure and macrostructure, therefore only one suction variable should be considered. In the (p, s) plane, a line corresponding to a constant microstructural effective stresses is named Neutral Line (NL), since no microstructural strain takes place along it. The neutral line divides the (p, s) plane into two parts, defining the microstructural stress paths indicated in Figure 1. The increment of the microstructural elastic strain is expressed as a function of the increment of the microstructural effective stress
vm
p p s Km Km Km
.
(11)
where the subscript m refers to the microstructural level, the subscript v refers to the volumetric component and Km is the microstructural bulk modulus, that in this paper is computed by the following law Km 2.3 e m p m Interaction between structural levels
(12)
(7)
where is established in such a way that under K0 loading lateral strains are zero. The hardening parameters, p0* depends on the rate of volumetric plastic strain. The hardening law is given by
* dp0 1 e d p * p0 0
Microstructural effects induce irreversible macrostructural deformations, which are considered proportional to microstructural strain according to interaction functions. Two interaction functions are definided: fc for microstructural compression paths and fs
(8)
where e is the void ratio. Elastic strains are induced by changes in net mean stress, deviatoric stress and suction according to the expression d e
s NL Microstructural compression Microstructural swelling
s dp 1 ds dq 1 e p 3G 1 e s patm
(9)
where G is the shear modulus, s is the elastic stiffness parameter against changes in suction and patm is the atmospheric pressure. 2.2 Microstructural model The microstructural behaviour is assumed elastic and volumetric. The microstructural volumetric

p
Figure 1. Definition of microstructural swelling and contraction paths.
1026
expansion accumulates fs
Table 1. Parameters used in simulation.
compression accumulates equilibrium state
Parameters defining BBM for macrostructural 0.009 s 0.002 (0) 0.10 r 0.80 (Mpa1) 10.0 p0*(MPa) 0.22 pc (MPa) 0.10 Parameters defining the law for microstructural level 1.0 m (MPa1) 0.006 m (MPa1) 0.012
fc Macroporosity develops due to drying Micropores invade macrocropores
p/po
Interaction functions fc0 0.10 fc1 1.5 nc 0.50 fs0 1.50 fs1 3.70 ns 2.0 emacro 0.80 emicro 0.30
Figure 2. Interation mechanisms between micro and macropores.
15 10 5 0 -5 -10 -15 10 100 1000 10000 Vertical stress (kPa) 5 Volumetric strain (%) Test 2 0 -5 -10 -15 10 100 1000 10000 Vertical stress (kPa) Laboratory Simulation Laboratory Simulation Test 1
for microstructural swelling paths. For isotropic loading the interaction functions depend on the ratio p/p0. According to Alonso et al. (1999) any suitable function for fs and fc consistent with the physical ideas presented in Figure 2 can be adopted. In this paper the interaction functions are expressed by fc fc 0 fc1 p p0 3 fs fs 0 fs1 p p0
ns nc
(13) (14)
NUMERICAL SIMULATION
Ferreira & Ferreira (2009) performed suction controlled oedometric tests of undisturbed soil samples at natural water content of a vertisoil from Petrolndia. The soil presented the following characteristics: wL 60%, PI 30%, wc 19%, w 17.41% and d 15.05 kN/m3. The initial degree of saturation was 59, 24% and suction 5.0 MPa. Two stress paths were followed in the tests. In the first one the specimen was loaded in stages, with a ratio of loading v /v 1, up to a vertical pressure (102040801603206401280 kPa). In the second path, after strain stabilization due to vertical pressure, soil suction was reduced up to zero and swelling strains measured. In the first test the specimen was flooded at 10 kPa whereas in the second test the inundation occurs at 160 kPa. The model described above was implemented in the Finite Element code CODE_BRIGHT (Olivella et al., 1996; Sanchez et al., 2005), which was used to simulate the suction controlled oedometric tests. The parameters used in the simulation are listed in Table 1. Laboratory data and the numerical simulation results are presented in Figure 3.
Figure 3. Comparison between laboratory data and simulation results.
In Test 1 the sample was flooded under a vertical stress of 10 kPa and swelling deformation was measured, a volumetric strain of 10% was registered, while in Test 2, inundation occurs under a higher vertical stress (160 kPa) the swelling volumetric strain was lower (about 2%). Comparison between laboratory and simulation results shows a very good agreement with experimental data. Loading under controlled suction and swelling due to suction reduction under a determined vertical stress is very well reproduced for both tests.
1027
CONCLUSIONS
The numerical results achieved demonstrate the capability of the double structure generalized plasticity model proposed by Sanchez et al. (2005) to reproduce the experimental behavior of expansive soils. ACKNOWLEDGEMENTS The authors wish to thank the financial support provided by CNPq/Brazil. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Geotechnique, vol. 40, no 3, pp. 405430.
Alonso, E., Vaunat, J. & Gens, A. (1999) Modelling the mechanical behaviour of expansive clays. Engineering Geology, vol. 54, pp. 173183. Ferreira, S.R.M. & Ferreira, M.G.V.X. 2009 Mudanas de volume devido variao do teor de umidade em um vertissolo no semi-rido de Pernambuco. Revista Brasileira de Cincia do Solo, vol. 33, pp. 779791. Gens, A. & Alonso, E.E. 1992. A framework for the behaviour of unsaturated expansive clays. Canadian Geotechnical Journal, vol. 29, pp. 10131032. Olivella, S., Gens, A., Carrera, J. & Alonso, EE. (1996) Numericalformulationforasimulator(CODE-BRIGHT) for the coupled analysis of saline media. Engineering Computations, vol. 13, no 7, pp. 87112. Sanchez, M., Gens A., Guimares, L.N. & Olivella, S. 2005. A double structure generalized plasticity model for expansive materials, Int. J. Numer. Anal. Meth. Geomech., vol. 29, pp. 751787.
1028
A new analytical solution for one dimensional transient flow in unsaturated soils
H. Farshbab Aghjani
School of Civil Engineering, Iran Universiy of Science and Technolgy, Iran Former MSc Student of Amirkabir University of Technology (Tehran Polytechnic), Tehran, Iran,
A. Soroush
Department of Civil and Environmental Engineering, Amirkabir University of Technology, Tehran, Iran
ABSTRACT: Transient flow of water in unsaturated soils is a complicated phenomenon. This phenomenon is formulated via a Partial Differential Equation (PDE) known as the Richards Equation. In the last decades, researchers have presented analytical solutions to solve this PDE. The existing analytical solutions can be applied only to problems with simple geometry and boundary conditions and they dont cover all possible transient flow problems. This paper introduces a new analytical scheme for solving the Richards Equation in homogenous and one-dimensional media. In this scheme, this equation is linearized by applying the exponential form of hydraulic conductivity and SWCC functions and performing the Kirchhoff Transforms. Then by exploiting the Green Function approach, a closed-form relation, which comprises sum of three integral terms, is derived. Moreover, the boundary and initial conditions are defined as arbitrary functions in this solution. Therefore, this new solution is appropriate for modeling all types of unsaturated seepage problems. 1 INTRODUCTION various boundary and initial condition (Kavetski et al., 2002; Clement et al., 1994). In this method, the soil medium is divided to elements and nodes. Then differential equation is transformed to discrete equations. These discrete equations are solved by numerical methods such as finite differences and finite elements, and consequently the solution including soils suction head can be obtained in discrete locations (nodes) and discrete intervals. Therefore, a numerical solution is an approximation of the explicit PDE. Also, by implementing numerical solutions, some numerical errors can be developed due to employment of implicit equation at nodes of discretized medium. In order to obtain the accurate solution and reduce discretization error, the convergence to the true solution and the value of error should be controlled (Cellia et al. 1990, Kavetski et al. 2001). For solving the PDE using analytical approach, some assumptions and simplifications should be made. Thus at first step, the Richards Equation should be transformed to a linear form by assuming an appropriate mathematical form for SWCC and HCF. Then by employing the analytical techniques for solving partial differential equation available in mathematics, the solution of transient
The transient flow of water in unsaturated porous media is known as one the most prominent phenomenon in unsaturated soil mechanics. During transient flow of water within a soil medium, the unsaturated features and properties of soil such as matric suction and moisture content are affected. In order to investigate the variations of soil properties, this phenomenon is modeled by mathematical means. That is, based on the mass continuity law, unsaturated transient flow is formulated as a Partial Differential Equation (PDE) presented by Richards (1931) and known in the literature as Richards Equation. For establishment of Richards Equation, the material variables including Soil Water Characteristic Curve (SWCC) and unsaturated Hydraulic Conductivity Function (HCF) are implemented. Because these functions depend on soils suction, Richards Equation is a nonlinear partial differential equation. The solution of Richards Equation can be attained by analytical, semi-analytical and numerical methods. The numerical method has vast application for the variety of transient flow problems with
1029
flow problem can be derived as a closed-form mathematical expression. The applicability of analytical solution is limited for some simple flow problems which have simple geometry, homogenous medium and simple initial and boundary conditions (Ju and Kung, 1997; Arampatzis et al., 2001; Kavetski et al., 2002). Nevertheless, these kinds of solutions have some advantages. For instance, they can be employed to examine the accuracy of the numerical solution and can be expressed in dimensionless form which appears simpler than the dimensional form. Also they are helpful for studying flow behavior in porous media. In the semi-analytical method, the differential equation is solved in two steps. The initial step is using analytical techniques such as Laplace transforms and the next step which includes the inversion to the real answer is performed by numerical approaches. In the past few years, researchers have presented a verity of analytical solutions for Richards Equation by utilizing the exponential-form of SWCC and HCF (Warrick 1975, Srivastava & Yeh 1991, Fityus & Smith 2001, Menziani et al. 2005). However, most of them are limited to semiinfinite or finite media. Also for initial conditions, their assumption is that the initial matric suction profile is constant or corresponds to steady state flow. Moreover, the boundary condition only includes particular forms such as constant matric suction or steady infiltration with constant flux at medium boundary. Consequently, the existing analytical solutions in literature attributed to some limited boundary and initial conditions and they cannot cover all general states for unsaturated transient flows. This paper presents a new analytical solution for solving the Richards Equation which can be applied for all boundary and initial conditions. In the new solution, at first step Rihards Equation is transformed to a linear form by employing the Kirchhoff Transform and using the exponential form for SWCC and HCF. Then by exploiting the Green Identify and Green Function approach, an analytical relation is derived as sum of the three integral terms. In these integral terms, the boundary and initial conditions are defined as alternative functions. Therefore, contrary to the previous analytical solutions in the literature, the new solution is capable of modeling all boundary and initial conditions. 2 TRANSIENT FLOW IN UNSATURATED SOIL
phase. The main factor which causes water to seep into the soil is the gradient in total potential (Lu & Likos 2004). The required potential for water movement in soil comprises gravitational potential which is equal to elevation of arbitrary point (defined positive upward from the arbitrary axis) and water pressure potential. When soil is unsaturated, a suction force is established within soil medium due to existence of water meniscus between soil grains. Thus, the pressure head turns to suction form which has a negative value. Therefore, the total head in any point of one-dimensional vertical column of unsaturated soil is equal to the sum of matric suction head (hm) and elevation head (z), i.e. h hm z. For practical applications, Darcys Law can be employed for determining the discharge velocity (V) of water in unsaturated state as follows: h Vk. z (1)
where h is the total head (defined previously), k is the unsaturated hydraulic conductivity which is a function of soil suction and z is the elevation of arbitrary point in soil column. In transient conditions, water flows in such a manner that continuity law is satisfied. Thus the governing mathematical equation for one-dimensional transient water flow in unsaturated soil can be expressed as the following partial differential equation which is presented by Richards (1931) and known as Richards Equation:
C . hm h k 1 t z z h . hm k 1 C . m t z z
(2)
where and C are the volumetric water content and specific moisture capacity, respectively. Because hydraulic conductivity and moisture capacity depend on matric suction head, the Richards Equation is a non-linear PDE. For linearization of this equation, the following exponential function, proposed by Gardner (1958), is used for k and C: k ( h ) kS . exp( . hm ) C ( h ) ( s r ) . . exp( . hm ) (3-a) (3-b)
Flow of water in unsaturated soils occurs through pores spaces occupied by a continuous liquid
In this function, ks , s , r , and are saturated conductivity, saturated volumetric water content, residual volumetric water content and pore size distribution factor, respectively. These parameters can be obtained by fitting of experimental test results. For more simplifications, the Kirchhoff
1030
transformation with the following non-dimensional time and space variables are introduced: K .
hm
SOLVING PDE
k ( hm ) . dhm ks . exp( . hm ),
(4-a) (4-b)
. ks . t Z . z, T ( s r )
Consequently, by substituting Equations 3 and 4 in Richards Equation (Equation 2) and implementing the non-dimensional transformed variables, the simplified operator form of the PDE governing the transient flow of water in unsaturated soils is obtained as following: 2 K K K 0 Z 2 Z T 2 (K ) 0 L( K ) Z 2 Z T
(5)
Equation 5 is parabolic type of PDE, and is valid in the domain of m; temporally and spatially defined as: Time domain : 0 T T ; Space domain : 0 Z L (6)
Besides, the boundary and initial conditions (BC and IC) of transformed PDE are: Boundry Condition : a . K b .
Z 0 a1, b1, B1(0,T ) , Z L a2 , b2 , B2 ( L1,T )
K B (Z ,T ) Z (7) (8)
InitialCondition : T 0 K f (Z )
The aforesaid boundary condition is introduced in general state which is recognized as the Robin boundary condition in mathematics. By selecting the appropriate values for a and b, all possible cases existing in flow problems can be covered. For instances, if a flux with value of q infiltrates into a soil medium at its upper border, then the mathematical form for boundary condition in the transformed form is defined in the following form: K K q Z (9)
In order to attain the analytical solution of the PDE, some approaches are introduced in the literature. For instances, the most efficient analytical methods are involved as separation of variables, variable transformation such as Laplace transform, Fourier transform and Green Function (Zauderer 1983, Evans et al. 1999). Each method has its own limitations and advantages. Because the time interval of unsaturated flow equation is bounded in general state, the method of Laplace transform cannot be applied. Also according to parabolic form of PDE and inhomogeneous BC and IC, the method of variables separation seems inappropriate to solve this PDE. The Green Function approach is one of the most comprehensive techniques for solving the PDE which can be applied to majority of problems such as inhomogeneous equations with arbitrary inhomogeneous initial and boundary conditions (Zauderer 1983). Thus in this paper, the Green Function approach was employed to solve analytically the partial differential equation of unsaturated transient flows. In order to solve the PDE by Green Function, the differential equation should have the self adjoining property. The self adjoining property allows utilizing the Green Second Identity and deriving an integral term relation. Most of engineering differential equations such as wave equation and Poisson equation cover this property. However, the parabolic PDE is devoid of self adjoining property and thus another complementary differential operator should be set up. In the literature of unsaturated soil mechanics, the Green Function approach has not been considered yet, so the adjoin operator of Equation 5 is not available. To this end, a new operator denoted as L* which is the adjoining operator for Equation 5 is defined in the domain of m as: Operator L* : 2 2 Z T Z 2G G G * L (G ) Z 2 Z T
(10)
It should be noted that the infiltration flux q can be constant or variant with time. Thus, for evaluating the water flow in unsaturated soil medium and determining profiles of matric suction, it is enough to solve the transformed flow equation within its spatial and temporal domain by imposing the initial and boundary conditions.
Now, the term of K.L*(G) G.L*(K) is calculated and integrated over the m domain. As L* is an adjoining operator of Equation 5, the result of integral is introduced in the following relation by applying the Green Second Identity (Zauderer 1983, Haberman 1983):
K L (G ) G L( K ) dZdT K G K G GK GK dZdT Z Z T Z

*
(11-a)
1031
K L (G )dZdT G L ( K )dZdT
*

I

III Z L
G . K T 0
Z L
T T
. dZ . K (Z , T ) . dZ
K
K T T G G GK dT GK T 0 dZ (11-b) Z Z Z 0

II IV

Z 0 Z L
G (Z , T )
For determining the outcome of aforesaid integrals, some specific conditions including boundary and end condition are attributed to G under L* operator:
L (G )
*
Z 0 Z L Z 0
G ( Z , 0 ) . K ( Z , 0 ) . dZ
(17) (18)
G (Z , 0) . f (Z ) . dZ
G Z
2
G Z
G T
( Z ) . (T ),
(12-a)
G L( K ) dZ dT 0
BoundryCondition : (a b ) G b End Condition : G (Z ,T T ) 0
G 0, (12-b) Z (12-c)
By substituting the results of expressions I, II, III and IV from Equations 14, 16, 17 and a18 in Equation 11-b, the following solution is obtained for non-dimensional Richards equation (Equation 5), (if a2 b2 0): K ( , ) T T B ( L,T ) . G ( L,T ) . dT a b2 Z T0 2
T T
The right-hand side of Equation 12-a, which is the resultant of L* operator, is the Dirac Delta Function. This function has some prominent properties in mathematics. Here, the following properties of this function are employed:
f ( x ) . ( x ) . dx f ( )
(13-a) (13-b)
(x )
B (0,T ) . G (0,T ) . dT a b Z 1 1 T0
0
for x for x
Z L Z 0
G (Z , 0) . f (Z ) . dZ
(19-a)
By implementing Equation 13-a, the result of integral term I in Equation 11-b can be written as:
In case of (a2 b2 0), the solution becomes as: K ( , ) T T B ( L,T ) . G ( L,T ) . dT b2 T0
T T
K . L (G ) dZ . dT K (Z ,T ) (Z ) (T ) dZ dT K (, )
*
(14)
In addition, considering the relevant boundary conditions attributed to G function, the result of term III in Equation 11-b is calculated in the following expression: (a b) . G b . G b . G 0 G Z a b Z
Z L
B (0,T ) . G (0,T ) . dT b1 T0
Z L Z 0
G (Z , 0 ) . f (Z ) . dZ
(19-b)
(15)
. G . K . K Z G Z G K Z 0 Z L T T 1 . G . dT B ( Z , T ) dT (16) T0 a b Z Z 0
Thus, by substituting the aforesaid expression obtained for K in the following equation, the closed-form solution of Richards Equation at a point with coordinate in the unsaturated column and during time can be derived: K ( , ) Ln ks hm ( , ) (20)
The results of other integral terms in Equation 11-b (III and IV) are calculated in the following equations:
The initial and boundary conditions in the righthand side of Equations 19-a and 19-b are expressed
1032
as arbitrary functions. Thus the new analytical solution can be applied for solving all transient flow problems in one-dimensional unsaturated porous media. In Equation 19, G(Z,T) is recognized as Generalized Function or Green Function of PDE. In order to complete the solution, G(Z,T) should be determined by solving Equation 12-a, with the boundary and end conditions defined in Equations 12-b and 12-c, respectively. 4 DETERMINING GREEN FUNCTION
where; n is the eigenvalue of Equation 25 and assumed that n 0. Now by implementing Equation 25 into Equation 24, the differential equation governing the time portion of Green Function is derived as: dN n (T ) n . N n (T ) an ( ) . (T ) dT , T T and N n ( T ) 0
(26)
Here, the eigenfunction method is used to determine G(Z,T). Hence, G is considered as a product of two separate functions for Z and T (Sankara 2005): G (Z ,T ) M n (Z )N n (T )
n 1
Thus, the time and space portions of Green Function can be determined by solving the corresponding differential equations presented in Equations 25 and 24 respectively. 4.1 Solving the equation of space portion By assuming that eigenvalues of Equation 25 are positive, the solution of differential equation corresponding to space portion of Green Function is derived as follows
4 n 1 c1 Cos Z 2 M n (Z ) e 0.5.Z 4 n 1 c 2 Sin Z 2
(21)
The spatial boundary conditions of Mn(Z) and temporal end conditions of Nn(T) are akin to attributions of function G. Furthermore, the Dirac Delta Function of space in the right hand of Equation 12-a, is defined in terms of Mn(Z):
(Z ) an ( )M n (Z )
n 1
(22)
(27) The values of coefficient c1 and c2 and eigenvalue n are determined by imposing the boundary condition of Mn(Z), which are identical to conditions attributed to Green Function. For example, its assumed that the lower boundary of the medium (i.e. Z 0) lies on water table and the constant flow with flux value of q infiltrates into the soil column at upper boundary (i.e. Z L). The coefficients of boundary conditions belonging to this kind of main transient flow problem are presented in Equation 9. Hence, the boundary conditions of both Green Function and the space portion of function will be as follows: Boundry Condition : (a b )M n b Z 0 M n ( Z 0 ) 0 dM n Z L (Z L ) 0 dZ dM n 0 dZ (28)
By multiplying Mn(Z) in Equation 22 and integrating in the whole spatial domain of m, while considering Equation 13-a, an() is determined as: an ( )
(Z )M n (Z )dZ M n ( ) 2 2 M n (Z )dZ M n (Z )dZ
(23)
By substituting G and (Z ) from Equations 21 and 22 in Equation 12-a, the following relation is established between Mn(Z) and Nn(T):

n 1
d 2M n (Z ) dZ 2
d M n (Z ) N n (T ) dZ
d N n (T ) M n (Z ) dT n 1 an ( )M n (Z ) (T )
n 1
(24)
In order to further simplify Equation 24, it is supposed that Mn(Z) is the eigenfunction of the following differential equation: d 2M n (Z ) d M n (Z ) nM n (Z ) dZ d Z2 (25)
By imposing the first boundary condition, coefficient c1 becomes 0. Considering c2 1 and imposing the second boundary condition, n is determined from the following equation: 4 n 1 . L 4 1 0 tan n 2 (29)
1033
4.2
Solving the equation of time portion
The end condition corresponding to the differential equation of time portion is presented in Equation 25. The solution of differential equation should be continuous all over time interval and have the jump discontinuity at T (Zauderer 1983, Haberman 1983). Hence, considering these conditions, the solution of time portion of Green Function is derived as follows: N n (T ) an ( ) . e n (T ) . H ( T ) 1 T H ( T ) 0 T
Green Function can be employed in any problem associated with unsaturated soil mechanics. REFERENCES
Arampatzis, G., Tzimopoulos, C., SakellariouMakrantonaki, M. and Yannopoulos, S. 2001, Estimation of unsaturatedflow in layered soils with the finite control volume method, Irrigation and Drainage, 50: 349358. Celia, A.M., Bouloutas, E.T. and Zarba, R.L. 1990, A General Mass-Conservative Numerical Solution for the Unsaturated Flow Equation, Water Resource Research, 26(7): 14831496. Clement, T.P., William R.W., and Molz, F.J. 1994, A physically based two-dimensional, finite difference algorithm formodeling variably saturated flow, Journal of Hydrology, 161: 7190. Evans, G., Blackledge, J. and Yardley, P. 1999, Analytic Methods for Partial Differential Equations, Springer, New York. Fityus, S.G. and Smith, D.W. 2001, Solution of the unsaturated soil moisture equation using repeated transforms, International Journal For Numerical And Analytical Methods In Geomechanics.; 25: 15011524. Gardner, W.R. 1958, Some steady state solution of unsaturated moisture flow equation with application to evaporation from a water table, Soil Science, 85(4): 228232. Haberman, R. 1983, Elementary Applied Partial Differential Equations, Prentice-Hall Inc., New Jersey, USA. Ju, S.H. and Kung, K.J.S. 1997, Mass types, Element orders and Solution schemes for Richards equation, Computers & Geosciences, 23(2), 175187. Kavetski, D., Binning, P. and Sloan, S.W. 2001, Adaptive time stepping and error control in a mass conservative numerical solution of the mixed form of Richards equation, In: Advances inWater Resources, 24: 595605. Lu, N. and Likos, W. 2004, Unsaturated Soil Mechanic, John Wiley & Sons, New York. Menziani, M., Pugnaghi, S. and Vincenzi, S. 2005, Analytical solutions of the linearized Richards equation with flux boundary conditions for a half space and a finite layer, Geophysical Research Abstracts, 7, 02748, 2005. Richards, L.A. 1931, Capillary conduction of liquids through porous medium, Physics, 1: 318333 Sankara. K. 2005, Introduction to Partial Differential Equations, Prentice, Hall of India. Srivastava, R. and Yeh, T.C.J. 1991, Analytical solutions for one-dimensional, transient infiltration toward the water table in homogeneous and layered soils, Water Resources Research, 27(5): 753760. Warrick, A.W. 1975, Analytical Solution to the onedimensional linearized moisture flow equation for arbitrary input, Soil Science, 127, 7984. Zauderer, E. 1983, Partial Differential Equations of Applied Mathematics, John Wiley & Sons, New York, USA.
(30)
where H( T) is known as Heaviside Function. As a result, by substituting the Equation 29 and 30 into Equation 21, the mathematical form of Green Function is obtained as follows:
M ( Z ) M n ( ) n (T ) . H ( T ) G ( Z ,T ; , ) L n e n 1 M 2 ( Z ) dZ n

M n (Z ) e
0.5 Z
4 n 1 . Sin 2
(31) After determining the Green Function, the procedure of solution for unsaturated transient flow problem is accomplished.
CONCLUSION
This paper presented a new analytical solution for unsaturated transient flow problems, by applying the Green Identity and Green Function. In order to apply the Green function, another complementary differential operator covering the adjoining property is determined for unsaturated flow problems. Therefore, the solution of differential equation is derived as the sum of three integral terms which includes the arbitrary initial and boundary conditions. Thus the new analytical solution covers all possible boundary and initial problems, such as infiltration with constant or varied-by -time flux. For facilitating the convergence to the true solution, its sufficient to select appropriate values for upper limit of sigma. In unsaturated soil mechanics, gravitational potential plays prominent role in diffusion problems such as moisture or solute transportations. This fact causes diffusion problems to be different from typical engineering diffusion problems. Therefore, the proposed approach for applying
1034
Effect of mesh resolution on long-term water balance calculations

Murray D. Fredlund & Rob Thode R.G. Nelson
SoilVision Systems Ltd., Saskatoon, SK, Canada PDE Solutions Inc., Spokane Valley, WA, USA
ABSTRACT: The quantification of the performance of earth covers has been largely determined through the use of test plots as well as numerical modeling. Such numerical models are run with the use of complex climatological data and unsaturated soil property functions. The input information is then used by the finite element solvers to determine vertical flow rates and, ultimately, long-term percolation rates through the earth cover. Calculations are generally complicated by the fact that the infiltration into a dry soil is one of the more complex types of numerical modeling scenarios. The challenge occurs largely because of the non-linearities present in the unsaturated soil properties. These non-linearities becomes part of the numerical model which must run for 1, 10, or 50 years. This paper examines the numerical difficulties associated with these calculations and examines the impact of small numerical issues over long time periods. In particular, the role of mesh density as it is related to the accuracy of water balance calculations is examined. 1 INTRODUCTION The movement of a wetting front (or increased saturation levels) into a dry soil is a classic seepage problem that has previously received attention. Two previous models, which have previously solved this problem, are worth mentioning. The Haverkamp (1977) model involves infiltration into a 1D column of material. A series of infiltration experiments were performed by Haverkamp in the laboratory using a plexiglass column uniformly packed with sand to verify the numerical results. The model was originally solved using 1D finite elements and the 1D finite difference solution methods. Time-steps used in the analysis were varied in the original work to determine their effect on the solution. The best solution presented occurred with small time-steps (10 seconds) and a dense grid. The material properties used in Haverkamps analysis used equations defined in terms of elevation head rather than soil suction. The model was initially set up in SVFlux and then minor modifications were made to the FlexPDE finite element script file to duplicate the solution exactly. The script file presenting a precise comparison of results can be provided upon request. The results of the comparison can be seen in Figure 1. The automatic time-stepping feature in the SVFlux/FlexPDE software is used to choose ideal timesteps. It can be seen that the mesh and time-steps selected by SVFlux automatically duplicates the best results presented by the Haverkamp (1977) solution.
Numerical models are increasingly being used for the analysis of the long-term performance of earth covers. While numerical models offer an excellent way to solve the partial differential equations associated with moisture flow, there are some issues with this type of analysis that need to be addressed. In a typical analysis, climate data is collected from somewhere between 5 and 100 years and a numerical model (typically a 1D model) is run to determine the amount of precipitation that ends up as net percolation. The primary unknown in these numerical models is the Actual Evaporation (AE). AE may be calculated as a fraction of Potential Evaporation (PE) through use of the Wilson-Penman equation. It should also be noted that runoff can have an influence during highintensity storms. Significant effort has been devoted to the solution of the climatic coupling (Wilson, 1997; Shackelford, C., 2005). However, the numerical model inherently poses a problem which is fundamentally difficult to solve. The problem of infiltration of precipitation into a dry soil is a welldocumented problem case (Haverkamp, 1977). The reason for the difficulty with this type of model is that the gradients in the numerical model may be high when a precipitation event hits a soil. The soil may be dry and, therefore, has an extremely low hydraulic conductivity.
1035
-10
-20
-30
dt = 10 sec dt = 30 sec dt = 120 sec Dense Grid SVFlux
-40
-50
-60
-70 0 5 10 15 20 Depth (cm) 25 30 35 40
Figure 1. Comparison between SVFLUX Haverkamp (1977) as presented by Celia (1990).

0
and
are typically thousands of precipitation events that may be introduced into the numerical model over a period of 10, 20, 50 or 100 years. Each event may be difficult to solve as a single event. It is extremely challenging to have a numerical model which handles all applied precipitation events with satisfactory accuracy. This paper examines the use of both automatic time-stepping and mesh refinement to improve solutions for the calculation of vertical flow. A few example models are presented which illustrate the limits of numerical analysis in earth cover evaluation. These examples show the benefits of using automatic mesh refinement to extend the limits of possible solutions. 2 ANALYSIS
Pressure head (cm )
-200 Dense Grid dt = 20 sec -600 dt = 2.4 min dt = 12 min dt = 60 min -800 SVFlux
-400
-1000
-1200 0 20 40 Depth (cm) 60 80 100
Figure 2. Comparison between SVFLUX and results presented by Celia (1990).
It should be noted that the ability of a software program to handle infiltration into a dry soil is somewhat dependant on the ratio of the intensity of the applied precipitation event to the saturated hydraulic conductivity of the soil layer at the surface. To date separate studies have shown the ability of the SVFlux/FlexPDE software to handle applied precipitation events of up to 10 times the saturated coefficient of permeability (Fredlund et al., 2006). 2.1 Example no. 1
Celia (1990) performed comparisons of 1D solvers by varying the time-steps and the solution methods (finite difference or finite element). The results were considered classic solutions and are commonly used to benchmark the validity of 1D infiltration models. The solution presented by Celia used the h-based formulation of Richards equation and a Newton-Raphson iterative method. A replica of Celias model was set up using the SVFlux software. Celia presented the material properties for the model as van Genuchtens equation for the soil-water characteristic curve and as van Genuchten and Mualems equation for representing the unsaturated hydraulic conductivity curve. Since both methods are implemented in the SVFlux software the parameters used for the material could be input directly. The results of the comparison can be seen in Figure 2. As in the previous model it can be seen that the automatic mesh generation and automatic time-step refinement allow quick convergence to the correct solution. It should be noted that significant effort has previously been invested in ensuring that a single case of infiltration is properly represented in a numerical model. In todays cover models there
Pressure head (cm )
The mesh density of a numerical model is one of the primary sources of error (Yeh, 2000). In order to demonstrate this issue a simple numerical model is set up in the SVFlux software. The numerical model is a simple fixed-mesh model consisting of a vertical column of soil with unsaturated soil properties defined. A few random precipitation events are then applied to the top of the soil column and the impact of these precipitation events is then tracked in the model. The soil-water characteristic curve is first then steepened until the results of the numerical model become erroneous (Table 1). The
Table 1. Fredlund & Xing soil-water characteristic curve parameters. Fredlund & Xing ID 1 2 3 4 5 af 1.24 1.24 1.24 1.24 1.24 nf 3.83 4.5 5.5 6.6 8 mf 0.18 0.3 0.4 0.6 1.5 hr 34.9 34.9 34.9 34.9 34.9
1036
unsaturated hydraulic conductivity was estimated using the Modified Campbell method. Once the numerical model is brought to failure, the mesh density in the numerical model is increased until the model again reaches the correct answer. The results of this experiment may be seen in Figure 3. It can be seen in Figure 3 that the influence of mesh density on infiltration models is significant.
Figure 3. validity. Influence of mesh density on solution
2.2
Example no. 2
Table 2. Precipitation events applied to the numerical model. Time (days) 1 2 3 4 5 6 7 8 9 10 Precipitation (m^3/day/m^2) 0.0001 0.0005 0.001 0.005 0.01 0.05 0.1 0.5 1 5
A second experiment was then set up as a numerical model. In the numerical model the total modeling time is set to 7 days. In each of these days a progressively more intense storm is introduced to the numerical model. The particular storm events are presented in the following table (Table 2). The model is: i. first run with automatic mesh refinement turned on. Then the model is run with a default static and unrefined mesh. The static unrefined mesh yields large volume-mass errors in the calculations. The mesh density is then increased in the static-mesh scenario until a static mesh yields a theoretically correct answer. It can be seen from the scenario with the automatic mesh refinement turned on that the mesh refinement detects cases where further mesh refinement is needed. This is illustrated in Figure 4 which shows spikes in the number of nodes in the
Figure 4.
Number of nodes in the 1D model versus time.
1037
Figure 5.
Time 0 days.
Figure 6.
Time 5 days.
Figure 7.
Time 10 days.
Figure 8.
Time 20 days.
1038
model which correspond to the applications of the storm events. It should be noted that the static mesh requires a dense mesh for the entire modeling duration. This then yields excessive model run-times. A practical application of mesh refinement in 2D can be seen in infiltration into a pile of waste rock. This is illustrated in the following snapshots in time. It can be graphically seen how the mesh refinement responds to the high degree of nonlinearity encountered on the seepage front. 3 CONCLUSIONS
From the presented examples it can be seen that the use of automatic time-step refinement and automatic mesh refinement significantly enhances the ability of software to solve difficult infiltration models. This technology is especially applicable to the evaluation of earth covers. Automatic mesh refinement produces a definable improvement in the ability to handle steep unsaturated soil properties such as those of sands and gravels. It can be seen with this study that the numerical modeling time required by numerical models performing cover modeling has the potential to be reduced through the selective use of additional nodes only in times of peak intensity storms. REFERENCES
Celia, M.A. & E.T. Bouloutas, (1990). A General Mass-Conservative Numerical Solution for the Unsaturated Flow Equation. Water Resources Research, Vol. 26, No. 7, pp. 14831496, July. FlexPDE Users/Theory Manual, PDE Solutions Inc., Spokane, WA, USA.
Gitirana, G.G., Fredlund, M.D. & Fredlund, D.G. 2005, Infiltration-Runoff Boundary Conditions In Seepage Analysis, Canadian Geotechnical Conference, September 1921, Saskatoon, Canada. Gitrana, G. Jr., Fredlund, M.D. & Fredlund, D.G. 2006, Numerical Modeling of Soil-Atmosphere Interaction for Unsaturated Surfaces, Fourth International Conference on Unsaturated Soils, April 25, 2006, Carefree, AZ, USA. Haverkamp, R., Vauclin, M., Touma, J., Wierenga, P.J. & Vachaud, G. (1977). A Comparison of Numerical Simulation Models for One-Dimensional Infiltration, Soil Science Society of America Journal, Vol. 41, No. 2. Mansell, R.S., Liwang Ma, L.R. Ahuja, & Bloom, S.A. 2002, Adaptive Grid Refinement in Numerical Models for Water Flow and Chemical Transport in Soil: A Review, Vadose Zone Journal, 1:222238 Shackelford, C., 2005, Environmental Issues in Geotechnical Engineering, September 1216, Osaka, Japan. SVFlux Users/Theory/Verification Manual, 2001, Fredlund, M.D., SoilVision Systems Ltd., Saskatoon, Sask. Canada. Wilson, W. (1990). Soil Evaporative Fluxes for Geotechnical Engineering Problems. PhD Thesis, Department of Civil Engineering, University of Saskatchewan, Saskatoon, SK, Canada. Yeh, G.-T. 2000. Computational subsurface hydrology: Reactions, transport, and fate. Kluwer Academic Publ., Boston, MA.
1039
Fully coupled analysis of unsaturated porous media using a meshfree method

A. Khoshghalb & N. Khalili
The University of New South Wales, Sydney, Australia
ABSTRACT: A meshfree algorithm is proposed for fully coupled flow-deformation analysis of unsaturated porous media. After a brief discussion of the governing equations, spatial discretization of the differential equations is presented using Radial Point Interpolation Method (RPIM). Temporal discretization is achieved based on a novel three-point approximation technique with variable time step, which has second order accuracy and avoids oscillatory response. Application of the model is demonstrated through simulating experimental results from two laboratory tests reported in the literature. Good agreement is obtained between the numerical and the experimental results in both cases. 1 INTRODUCTION and the shape functions do not satisfy the Kronecker delta property (Nguyen et al. 2008). To overcome this difficulty, Liu and Gu (Liu & Gu 2001) proposed Point Interpolation Method (PIM). Wang et al. (2001) applied this technique to Biots consolidation problem. Later, Wang & Liu (2002) introduced Radial Point Interpolation Method (RPIM) to overcome a possible singularity associated with PIM (Wang et al. 2002, Wang et al. 2004, Wang et al. 2004). In this paper, a fully coupled flow and deformation model is presented for meshfree analysis of unsaturated porous media based on Radial Point Interpolation Method (RPIM). After a brief description of the governing differential equations, the meshfree representation of the model is discussed and a novel three-point approximation technique is proposed for temporal discretization of the governing equations, which has second order accuracy, is unconditionally stable and avoids oscillatory response. Application, validity and stability of the model are demonstrated using several numerical examples. 2 NOTATION AND SIGN CONVENTION
Meshfree Methods (MMs) were specifically developed to eliminate difficulties associated with conventional numerical methods due to their reliance on a mesh to construct approximation functions (Nguyen et al. 2008). This particularly the case in dealing with problems involving large deformations, explosion, fragmentation, impact, penetration, fluid-structure interaction, etc. MMs were first originated some forty years ago; however, the research effort devoted to their formulation was not considerable until early 1990s (Belytschko et al. 1996). Nayroles et al. (1992) were the first to use the Moving Least Square (MLS) in a Galerkin method called the Diffuse Element Method (DEM). Belytschko et al. (1994) refined and modified the DEM based on Galerkin weak forms. This method, often referred to as the Element-Free Galerkin Method (EFGM), has been widely applied to various engineering problems (Belytschko et al. 1996, Liu & Gu 2005, Nguyen et al. 2008, Thomas-Peter Fries & Matthies 2004). To date, MMs have been applied to a range of engineering problems including solid mechanics (Belytschko et al. 1996, Donning & Liu 1998), heat conduction (Singh et al. 2003, Sladek et al. 2004), fluid flow (Lin & Atluri 2000, Lin & Atluri 2001), etc., though their application to geotechnical engineering problems has not been fully developed. Modaressi & Aubert (1995, 1996) applied MMs to the analysis of coupled flow deformation in fluid saturated porous media based on DEM and EFGM discretization techniques. However, the treatment of essential boundary conditions in MMs which use MLS to construct shape functions (such as DEM and EFGM) is not straightforward
Compact matrix-vector notation is used throughout. Bold imprints denote vectors and matrices and over-dot represents time derivative. is the gradient operator and div() (). The identity vector is defined as [1,1,1,0,0,0]T. Sign convention of continuum mechanics is adopted throughout. Compression is taken as negative and tension as positive. Pore water pressure (pw) and pore air pressure (pa) are taken as positive in compression following the soil mechanics convention.
1041
GOVERNING EQUATIONS
a MP H P a MP Q (1 )C TU 21 22 w a a a a
(7)
Following the work of Khalili et al (2008), the differential equations governing fully coupled flow and deformation in unsaturated porous media may be expressed as 0 ) p w (1 ) p a ] F div[ D(u k k a11 p w a12 p a div rw (pw w g ) divu w k k div ra (pa a g )
a
(1) (2)
where U is the vector of nodal displacements, Pw and Pa are the vectors of the nodal pore water and air pressures respectively, R is the vecor of nodal forces, Qw and Qa are the vectors of nodal fluxes of the water and air phases respectively, and K, C, Hw, Ha and M are the global property mantrices of the system. These matrices are evaluated by assembling the following corresponding local property matrices K s s B1T DB1 dV s
V
(8-1) (8-2) (8-3) (8-4) (8-5) (8-6) (8-7)
a22 p a a21 p w (1 )divu
(3)
C s s B2T N dV s
V
with a11 cw nw a12 , a22 a21 ca na , S a12 a21 n r s
Hws Has
Vs
krw T B3 k B3 dV s w kra T B3 k B3 dV s a
(4)
Vs
where D is the drained stiffness matrix of the soil, u is the displacement vector of the solid skeleton, pa is the pore air pressure and pw is the pore water pressure. K is the intrinsic permeability of the soil, krw is the relative permeability with respect to the water phase, kra is the relative permeability with respect to the air phase, w and a are the dynamic viscosities of water and air respectively, pw and pa are the intrinsic mass densities of water and air respectively, and g is the vector of gravitational acceleration. cw and ca are the coefficients of water and air compressibility, nw and na are volumetric water and air contents respectively, n is the porosity and Sr is the saturation ratio. d(s)/ds is the incremental effective stress parameter and s pa pw is the matric suction. The incremental effective stress parameter describes the contribution of incremental suction to incremental effective stress and serves as a platform to couple flow and deformation fields. In this paper, is quantified using the approach proposed by Khalili & Khabbaz (1998) and Khalili et al. (2008.) 4 4.1 DISCRETIZED FORM OF GOVERNING EQUATIONS Spatial discretization
M s s N T N dV s
V
Qw s s N T qw dS s
S
Qa s s N T qa dS s
S
in which Vs and Ss are the domain and boundary of the support domain, respectively, N [1(x) 2(x) n(x)] is the shape function matrix, 1 is the nodal shape function, I 1,2,3, , n and n is the number of nodes in the support domain of the point of interest. Determination of I (x) and thus N will be addressed subsequently. qw and qa are the fluid fluxes across the boundary. B1, B2 and B3 are matrices for determining derivation of field variables. For a two dimensional setting, they are defined as 1 x 1 0 1 x 2 1 , x 2 1 x1
0
B1
B2
1 x1
1 x 2
,
B3
1 x 1 1 x 2
(9)
The spatial discretization of the governing Equations 1, 2 and 3 can be obtained using the Galerkin approach CP (1 )CP R (5) KU
w a
R in Equation 5 is evaluated by assembling the vector of nodal forces from all support domains. The contribution from each support domain is determined through R s s N T F dV s s N T T dS s
V S
a MP H P a MP Q C TU 11 12 w w w a w
(6)
(10)
1042
in which F is body forces and T is boundary tractions. 4.2 Temporal discretization Temporal discretization of Equations 5 to 7 is performed using the finite difference technique. A three-point discretization scheme that is unconditionally stable and has second order accuracy is adopted (Khoshghalb et al. in press). In the approach proposed, the calculations start with an initial time step of t0, and then after each time level, time step is increased by a constant growth factor of ( 1). For the time interval t t to t t, it follows that the time derivative of an arbitrary function, f, at time t t takes the form (Khoshghalb and Khalili 2009)
t t Bf t Cf t t t t Af f t 2 1 1 1 , B , C A B ( 1)
SHAPE FUNCTIONS
Equations 13 to 15 together with Equations 8 to 10 provide the necessary information for assembling the overall property matrices of a partially saturated deformable porous medium using a MM. The only pre-requisite is the determination of the nodal shape functions, I(x). A number of approaches may be adopted for this purpose including the method of Moving Least Squares (MLS), Point Interpolation Method (PIM) and Radial Point Interpolation Method (RPIM) (Liu & Gu 2005). In this work, RPIM is preferred as it creates nodal approximation functions that satisfy the Kronecker delta function property and lead to matrix equations that are well posed (Liu & Gu 2005). For a point of interest, x, the dimension of the support domain, ds, is determined as ds sdc (16)
(11-1) (11-2)
Applying Equation 11 to Equations 5 to 7, at the neighborhood of t t, and over the time interval (1 )t (i.e. from t t to t t), the overall matrix equation of the system is then expressed as EW Y (12) where matrices E, W and Y are defined in Equations 13 to 15 (N in these equations is the overall number of field nodes). Equation 12 is solved using a time marching technique based on the results of the two previous time steps. AK 2 N 2 N A C 2 N N T E A C ( Aa11M DtH w )N N N 2N T Aa21M N N A(1 )C N N U 2 N 1 W P 1 w N Pa N 1 4 N
where s is the dimensionless size parameter and dc is the average nodal spacing in the neighbourhood of the point of interest. Let n be the number of nodes within the support domain of the current point of interest. Then, RPIM constructs the shape functions for the point of interest by forcing the approximation function u(x) to pass through all the nodes in the domain of influence, using the relationship u( x ) I 1 aI RI ( x ) J 1 bJ pJ ( x )
n m
RT ( x )a pT ( x )b
(17)
where RI(x) are Radial Basis Functions (RBFs), pJ (x) are augmented polynomials, m is the number of monomial basis functions and a and b are
A(1 )C N N
Aa12 M N N ( Aa22 M DtH a ) N N 4N 4N
(13)
(14)
AR2 N 1 t t T t T t t t ( ) ( ) A B C A B a D t Ca C U C U M H P MP 11 11 w w w Y ( A B )a12 MPat Ca12 MPat t DtQw N 1 ( A B )(1 )C TU t C (1 )C TU t t ( A B )a12 MPwt Ca12 MPwt t ( A B )a22 M DtH a Pat Ca22 MPat t DtQa N1
(15)
4 N 1
1043
coefficients yet to be determined. The vectors in Equation 17 are defined as aT a1 a2 an bT b1 b2 bm RT ( x ) R1( x ) R2 ( x ) Rn ( x ) pT ( x ) p1( x ) p2 ( x ) pm ( x ) (18) (19) (20) (21)
R0 T P
P a R0 a u G G T 0 b b 0 P
P 0
(27)
in which R0 is the moment matrix of RBFs R1(r1 ) R2 (r1 ) R (r ) R (r ) 1 2 1 2 R0 ( ) R r R 2 ( rn ) 1 n Rn (r1 ) Rn (r2 ) Rn (rn ) ( n n )
(28)
At each point of interest, x, radial basis functions are calculated as RI ( x ) RI ( x1, x2 ) RI

1
P is the polynomial moment matrix

1I 2 2 2I 2
(x x ) (x x )
(22)
RI(x) is only a function of the distance between the interpolating point (the point of interest) and a node at xI. Multi-Quadratic basis (MQ) with two shape parameters, q and c is employed in this work as the radial basis function RI ( x1, x2 ) rI 2 ( c dc )
P2 ( x1 ) P 1 ( x1 ) P (x ) P (x ) 1 2 2 2 P 1 ( xn ) P 2 ( xn ) P 1 1 1 1 x11 x12 x12 x1n x21 x22 x23 x2n
2 q
,
c 0
(23)
The polynomial basis function used for augmenting RBFs has the following monomial terms pT ( x ) 1 x1 x2 x12 x1x2 x2 2 (24)
Pm ( x1 ) Pm ( x2 ) Pm ( x3 ) Pm ( xn ) ( n m )
Pm ( x1 ) Pm ( x2 ) Pm ( xn ) ( n m )
(29)
u is the vector of nodal values of the field variables uT u1 u2 u3 un (30)
Coefficients a and b in Equation 17 are determined by enforcing u(x) to pass through all n nodes in the support domain for the point of interest. As such, n lineare equations are created (one for each point in the support domain) as uK u( x1K, x2K ) u( xK ) I 1 aI RI ( xK )
n
and G is the coefficient matrix of the system of equations. Now, from Equation 27 we have a 1 u b G 0 (31)
J 1 bJ pJ ( xK ) K 1, 2, , n
m
(25)
where uk is the nodal value of the independent variable (displacement in this study) at the Kth node in the support domain. Thus, m additional equations are required to obtain a unique solution. To make the final coefficient matrix symmetric, these are chosen as
For arbitrary scattered nodes in the support domain, to render G non-singular, m must be chosen such that m n (Liu & Gu 2005). Substituting for a and b in Equation 17 from Equation 31, yields u ( x ) RT ( x )a pT ( x )b {RT ( x ) {RT ( x ) u pT ( x )}G 1 0 a pT ( x )} b (32)
I 1 aJ pJ (xI ) 0
n
J 1, 2, , m
(26)
The above system of m n equations in the matrix form can be written as
1044
On the other hand, according to the definition of shape functions we have u( x ) I 1 I ( x )u1 Nu
n
(33)
Comparing Equations 32 and 33, the shape function matrix required for the determination of the property matrices of the support domain is obtained as {N1 n or 01 m } {RT ( x ) pT ( x )}G 1 (34)
I ( x ) J 1 RJ ( x )G 1J ,I
n m
J 1 pJ ( x )G 1J n,I
(35)
where G1J, I denotes the (J, I ) entry of matrix G1.
APPLICATION
Currently there are no analytical solutions available in the literature for an unsaturated medium. Thus, the performance of the proposed formulation is investigated by comparing numerical simulations with experimental results from the literature. Fredlund and Rahardjo (Fredlund & Rahardjo 1986) performed a series of laboratory experiments to study the volume change
behavior of unsaturated soils. Two of their tests (tests 1 and 4) conducted in the triaxial apparatus are selected here for numerical modeling. Initial volume-mass properties for these tests are summarized in Table 1. The change in the stress component associated with each test is given in Table 2. The results of their experiments are presented in Figures 1 and 2. To model the experiments, the values of compressibility coefficients for each of the experiments were determined by using the total amount of deformations. The average values of the constitutive coefficient a12 were then derived for the analysis (Khalili et al. 2000). The values of permeability coefficients were selected to best fit the data. Average values of the permeability coefficients were used and assumed to be constant throughout the volume change process. In the RPIM, the MQ radial basis functions with shape parameters of q 1.03 and c 1 were used for determination of the nodal shape functions (Liu & Gu 2005). Linear polynomial basis functions were constructed assigning m 3 and the dimensionless size of the support domain was taken as s 3. Material properties were assumed homogenous and isotropic. The numerical results in terms of volume change versus elapsed time are also presented in Figures 1 and 2. In test 1, a large instantaneous volume decrease is obtained at the time when the load is applied. This value along with the final volume changes of water phase and soil structure are used in obtaining the constitutive coefficients of the numerical model. As the figures show, there is a good agreement between experimental and numerical results in both cases.
Table 1. Test no. 1 4
Initial properties of the specimens tested by Fredlund and Rahardjo. Diameter (cm) 10.00 9.83 Height (cm) 11.82 5.76 Total volume (cm3) 929.1 437.2 Water content (%) 34.3 32.1 Void ratio 1.070 0.931 Degree of saturation (%) 78.9 90.3
Table 2.
Stress change associated with each test (Fredlund and Rahardjo). Total stress () (kPa) Pore water pressure (pw) (kPa) Initial 163.8 177.3 Final 164.4 379.2 Pore air pressure ( pa) (kPa) Initial 214.4 523.0 Final 215.6 530.9
Test no. 1 4
Initial 358.7 611.4
Final 560.6 610.1
Change (kPa) 3 202.2 pw 201.9
1045
1.5
Laboratory results (Fredlund & Rahardjo, 1986) Present model
REFERENCES
Belytschko, T., Y. Krongauz, D. Organ, M. Fleming, and P. Krysl, 1996. Meshless method: an overview and recent developments. Comput. Methods Appl. Mech. Eng. (139): pp. 347. Belytschko, T., Y.Y. Lu, and L. Gu, 1994. Element-free Galerkin methods. International Journal for Numerical Methods in Engineering 37(2): 229256. Donning, B.M. and W.K. Liu 1998. Meshless methods for shear-deformable beams and plates. Computer Methods in Applied Mechanics and Engineering 152(12): 4771. Fredlund, D.G. and H. Rahardjo 1986. Unsaturated Soil Consolidation Theory and Laboratory Experimental Data, Fort Lauderdale, FL, USA, ASTM. Fredlund, D.G. and H. Rahardjo 1993. Soil Mechanics for Unsaturated Soils, Wiley-Interscience Fries T.P. and H.G. Matthies 2004. Classification and Overview of Meshfree Methods. Khalili, N., M.A. Habte, and S. Zargarbashi, 2008. A fully coupled flow deformation model for cyclic analysis of unsaturated soils including hydraulic and mechanical hystereses. Computers and Geotechnics 35(6): 872889. Khalili, N. and M.H. Khabbaz 1998. A unique relationship for for the determination of the shear strength of unsaturated soils. Geotechnique 48(5): 681687. Khalili, N., M.H. Khabbaz, and S. Valliappan, 2000. An effective stress based numerical model for hydro-mechanical analysis in unsaturated porous media. Computational Mechanics 26(2): 174184. Khoshghalb, A., N. Khalili, and A.P.S. Selvadurai, A Three-point Time Discretization Technique for Parabolic Partial Differential Equations. International Journal for Numerical and Analytical Methods in Geomechanics. under review Lin, H. and S.N. Atluri 2000. Meshless Local PetrovGalerkin (MLPG) method for convection-diffusion problems. CMESComputer Modeling in Engineering and Sciences 1(2): 4560. Lin, H. and S.N. Atluri 2001. The Meshless Local PetrovGalerkin (MLPG) method for solving incompressible Navier-Stokes equations. CMESComputer Modeling in Engineering and Sciences 2(2): 117142. Liu, G.R. 2003. Mesh Free Methods: Moving Beyond the Finite Element Method CRC Press. Liu, G.R. and Y.T. Gu 2001. A point interpolation method for two-dimensional solids. International Journal for Numerical Methods in Engineering 50(4): 937951. Liu, G.R. and Y.T. Gu 2005. An Introduction to Meshfree Methods and Their Programming, Springer. Modaressi, H. and P. Aubert 1995. Coupled FEMEFGM for consolidation modelling. Numerical Models in Geomechanics, Rotterdam, A.A. Balkema. Modaressi, H. and P. Aubert 1996. A diffuse elementfinite element technique for transient coupled analysis. International Journal for Numerical Methods in Engineering 39(22): 38093838. Nayroles, B., G. Touzot, and P. Villon, 1992. Generalizing the finite element method: Diffuse approximation and diffuse elements. Computational Mechanics 10(5): 307318.
Volume change (%)
Water phase
0.5
Solid Structure
0 10
-1
10
Elapsed time (min)
10
10
10
10
Figure 1. Comparisons between numerical results and laboratory results of Fredlund & Rahardjo (1986), Test 1.
0
Water phase
Volume change (%)
-1
Solid Structure
-2
Laboratory results (Fredlund & Rahardjo, 1986) Present model
-3 0 10
10
Elapsed time (min)
10
10
10
Figure 2. Comparisons between numerical results and laboratory results of Fredlund & Rahardjo (1986), Test 4.
CONCLUSIONS
A fully coupled flow-deformation model for unsaturated media is presented using a meshfree method. The model is formulated incrementally using the effective stress principle. Spatial discretization is performed using RPIM. Time domain is discretized using a three-point method which has second order accuracy and eliminates spurious oscillations. Two unsaturated tests performed by Fredlund and Rahardjo (1986) are simulated. Good agreement is obtained in both cases.
1046
Nguyen, V.P., T. Rabczuk, S. Bordas, and M. Duflot, 2008. Meshless methods: A review and computer implementation aspects. Mathematics and Computers in Simulation 79(3): 763813. Singh, I.V., K. Sandeep, and R. Prakash, 2003. Meshless EFG method in transient heat conduction problems. International Journal of Heat and Technology 21(2): 99105. Sladek, J., V. Sladek, and S.N. Atluri, 2004. Meshless local Petrov-Galerkin method for heat conduction problem in an anisotropic medium. CMESComputer Modeling in Engineering and Sciences 6(3): 309318. Wang, J.G. and G.R. Liu, 2002. A point interpolation meshless method based on radial basis functions. International Journal for Numerical Methods in Engineering 54(11): 16231648.
Wang, J.G., G.R. Liu, and P. Lin, 2002. Numerical analysis of Biots consolidation process by radial point interpolation method. International Journal of Solids and Structures 39(6): 15571573. Wang, J.G., G.R. Liu, and Y.G. Wu, 2001. A point interpolation method for simulating dissipation process of consolidation. Computer Methods in Applied Mechanics and Engineering 190(45): 59075922. Wang, J.G., T. Nogami, G.R. Dasari, and P.Z. Lin, 2004. A weak coupling algorithm for seabed-wave interaction analysis. Computer Methods in Applied Mechanics and Engineering 193(3638): 393556. Wang, J.G., B. Zhang, and T. Nogami, 2004. Waveinduced seabed response analysis by radial point interpolation meshless method. Ocean Engineering 31(1): 2142.
1047
Tow-dimensional fundamental solution for dynamic behaviour analysis of unsaturated soils

P. Maghoul B. Gatmiri
Universit Paris-Est, UR Navier, Ecole des Ponts ParisTech, Champs sur Marne, France Departments of Civil Engineering, University of Tehran, Tehran, Iran Universit Paris-Est, UR Navier, Ecole des Ponts ParisTech, Champs sur Marne, France (Currently in Andra, Paris, France)
D. Duhamel
Universit Paris-Est, UR Navier, Ecole des Ponts ParisTech, Champs sur Marne, France
ABSTRACT: In this paper, the coupled equations governing the dynamic behaviour of unsaturated soils are derived based on the poromechanics theory in the frame of the suction-based mathematical model presented by Gatmiri (1997) and Gatmiri et al. (1998). In this formulation, the solid skeleton displacements, water pressure and air pressure are presumed to be independent variables. After, by applying the Laplace transform and by using the Kupradze method, the two-dimensional fundamental solutions are obtained in frequency domain, for such dynamic governing differential equations. 1 INTRODUCTION In this paper first, the set of fully coupled governing differential equations of a porous medium saturated by two compressible fluids (water and air) subjected to dynamic loadings is obtained. These phenomenal formulations are presented based on the experimental observations and with respect to the poromechanics theory in the frame of the suction-based mathematical model presented by Gatmiri (1997) and Gatmiri et al. (1998). In this model, the effect of deformations on the suction distribution in the soil skeleton and the inverse effect are included in the formulation via a suction-dependent formulation of state surfaces of void ratio and degree of saturation. The linear constitutive law is assumed. The mechanical and hydraulic properties of porous media are assumed to be suction dependent. In this formulation, the solid skeleton displacements, water pressure and air pressure are presumed to be independent variables. After, the two-dimensional fundamental solutions in Laplace transformed domain are obtained by the use of the method of Kupradze for the dynamic ui pw pa governing differential equations of unsaturated porous media. Finally, the derived results are verified analytically by comparison with the previously introduced corresponding fundamental solutions in the elastodynamic limiting case.
The site effects generated by topographical features are among the sources of amplifications and de-amplifications of the seismic signals, which can be important over large frequency domains. Site response analysis of topographical structures could only be solved accurately, economically and under realistic conditions, with the aid of numerical methods such as the Boundary Element Method (BEM). The BEM is a very effective numerical tool for dynamic analysis of linear elastic bounded and unbounded media. This method is very attractive for wave propagation problems, because the discretization is done only on the boundary, yielding smaller meshes and systems of equations. Another advantage is that the BEM represents efficiently the outgoing waves through infinite domains, which is very useful when dealing with scattered waves by topographical structures. When the BEM is applied to problems with semi-infinite domains, there is no need to model the far field. In this method, during the formulation of boundary integral equations, the fundamental solutions for the governing partial differential equations should be derived first. Indeed, attempting to solve numerically the boundary value problems for unsaturated soils using BEM leads one to search for the associated fundamental solutions.
1049
GOVERNING EQUATIONS
iw w Cwa p a w ,i Sw ii Cww p
(9)
Governing differential equations consist of mass conservation equations of liquid and gaseous phases, the equilibrium equation of the skeleton associated with water and air flow equations and constitutive relation. The assumption of infinitesimal transformation and incompressibility of solid matrix is considered. 2.1 Solid skeleton
where Cww (ng1 CwnSw); Cwa Caw ng1. In this equation, wiw,a is the displacement of water or air relative to solid, Saw,a is the degree of saturation relative to water or air, C is the compressibility of water or air Cw,a d/(dpa) and g1 dSw /d(pa pw) 2.3 Mass conservation of air With the same approach presented for the water mass conservation, the mass conservation equation of the air can be written as ia,i Sa ii Cwa p w Caa p a w where Caa (ng1 CanSa); Cwa Caw ng1. 2.4 Flow equation for the water
.
The equilibrium equation and the constitutive law for the soils solid skeleton including the effect of suction are written [2]: ( ij ij pa), j pa,i fi i (1)
(10)
ij
ij pa ij kk 2 ij Fijs pa pw

(2)
where , are Lame coefficients, paw,a is the water s or air pressure, ij is the Kronecker delta and F ij is the suction modulus matrix: F s D . Dsuc
1
(3)
in which Dsuc is a vector obtained from the state surface of void ratio (e) which is a function of the independent variables of ( pa) and (pa pw). Dsuc 1 e/((1 e) (pa pw)) (4)
Based on generalized Darcys law for describing the balance of the forces acting on the liquid phase of the representative elementary volume, the water velocity in the unsaturated soil takes the following form: w/kw w g uw pw,i w (11)
in which kw denotes the water permeability in an unsaturated soil. 2.5. Flow equation for the air With the same approach presented for the water based on generalized Darcys law, the air velocity in the unsaturated soil takes the following form: a/ka a g uw pa,i a (12)
The elasticity matrix (D) can be presented by using the bulk modulus and the tangent modulus D D(K0, Et) D( pa, pa pw) (5)
where Et is tangent elastic modulus which can be evaluated as Et El Es (6)
El is the elastic modulus in absence of suction and Es ms (pa pw) (7)
in which ka is the air permeability in an unsaturated soil. 2.6 Summery of the field equations
ms being a constant, Es represents the effect of suction on the elastic modulus. K0 is the bulk modulus of an open system and evaluated from the surface state of void ratio
1 K0 1 e e / pa
By introducing (2) into (1), (11) into (9) and (12) into (10), we have
u , u , F s pw, 1 F s pa,

f 0 u , w kw u , kw pw, Cww p w Sw u Cwa pa 0 , a ka u , ka pa, Cwa p w Sa u a 0 Caa p
(13)
(8)
2.2
Mass conservation of water
(14) (15)
The conservation law for the mass of water is written:
1050
GOVERNING EQUATIONS IN LAPLACE TRANSFORMED DOMAIN
The Laplace transformation is used to eliminate the time variable of a partial differential equation. Therefore, by applying the Laplace transform with the assumption of zero initial conditions, we can rewrite compactly the transformed coupled differential equation system into the following matrix form: u i B w p
T a p fi
calculate the inverse matrix of B (B 1 Bco/det (B. For the second stage, we assume that is a scalar solution to the equation det (B) I I(x ) 0 BBco (x ) Consequently, we get co G B
(20)
0 0 0
(16)
Following Hrmanders idea, first, the determi nants of the operator B* are calculated: ) ( 2 )k k ( 2 )( 2 ) det (B w a 1 2 2 2 ( 3 )( 4 )
with the not self-adjoint operator B: ( s

2
(21)
) i j ( ) i j s1 j s 2 j
F i kw C ww s C wa s
(1 F ) i C wa s ka C aa
B
(17) s
where 1 (Sw wkws) and 2 (Sa akas) In equation (16), the partial derivative ( ),i is denoted by t and ii is the Laplacian operator. Note the operators B in (17) are not self adjoint. Therefore, for the deduction of fundamental solu tions, the adjoint operator to B has to be used. ( a
a3 s ) a2 1 a8 a12
2
2 2 2 In which the coefficients 2 1, 2, 3 and, 4 are 2 the roots, where one of its roots is the 2 1 s /, which is related to the shear wave velocity propagating through the medium. The remained three roots 2 2 2 2, 3, 4 must be determined as these which satisfy
2 3 4

2 2 2 3
s
2
a 1 Fs s
2 2

Caa s kw ka Sa 1 Fs s C aa s
3
Fs w s
2
Sw Fs s
Cww s
( sa 4 a5 s ) a9 a10 s a11 s
( sa6 a7 s )
2
B
a11 s a13 a14 s

2 kw
2 2 2 4 2 3

2 4
2 ka

(18) where a1 , a2 , a3 , a4 Sw, a5 wkw, a6 Sa, a7 aka, a8 Fs, a9 kw, a10 Cww, a11 Cwa, a12 (1Fs), a13 ka and a14 Caa. 4 FUNDAMENTAL SOLUTIONS
2 kw
Cww s

2 wa
w FsC aa 1 Fs Cwa s
2 ka
(22)
3
C
ww
C aa C kw ka
s
a FsCwa 1 Fs Cww s
2 ka
Sw FsC aa Cwa 1 Fs s
2 kw
The objective of this section is to derive the fundamental solution associated with the operator (18). Mathematically speaking a fundamental solution is a solution of the equation BG I(x y) (t) 0 where the matrix of fundamental solutions is denoted by G, the identity matrix by I and the matrix differential operator by B. In this study, because the operator type of the governing equations is an elliptical operator the explicit 2D Laplace transform domain fundamental solution can be derived by using the method of Kupradze et al. [5] or Hrmander [4]. The idea of this method is to reduce the highly complicated operator given in (18) to simple well known operators. In this method, in the Laplace transform domain, the first stage is to find the matrix of cofactors Bco to
2 3 4
CwwC aa C
2 kw ka
Sa FsCwa Cww 1 Fs s
2 wa
2 kw ka
s
2 kw ka
These three roots correspond to the three compressional waves which are affected by the degree of saturation and the spatial distribution of fluids within the medium. Secondly, by introducing the determinant, the scalar equation corresponding to (19) is given by
2 2 2 ( 1 )( 2 2 )( 3 )( 4 ) ( x ) 0
(23)
in which is an interim operator, i.e. ( 2 )kw ka (24)
1051
Equation (23) can be expressed as either of four equations (25), (26), (27) and (28):
2 ( 1 )1 ( x ) 0 2 2 2 1 ( 2 )( 3 )( 4 )
4.1 Displacement caused by a Dirac force in the solid

U S G

2
2 ss 1
(25)
K 0 ( 1r )
2 1 2 ss 2
1 2
2
( )
( (x ) 0 2 2 2 2 ( 1 )( 3 )( 4 )
2 ( 3 )3 ( x ) 0 2 2 2 3 ( 1 )( 2 )( 4 ) 2 ( 4 ) 4 ( x ) 0 2 2 2 4 ( 1 )( 2 )( 3 )
2 2 ) 2
( K )( K 2 2 2 2 2 ( 1 2 )( 1 3 )( 1
2 1 2 2 2
s
4 )
2 2
( R1 1 K1 ( 1r ) R2 1 K 0 ( 1r )) ( R1 2 K1 ( 2 r ) R2 2 K 0 ( 2 r )) ( R1 3 K1 ( 3 r ) R2 3 K 0 ( 3 r ))
2 2
(26)

( 2 K ss1 )( 2 K ss 2 ) ( )( )( 3 ) ( 3 K )( K ss 2 ) ( )( )( 1 ) ( 4 K )( K ss 2 ) ( )( )( 2 )
2 4 2 3 2 4 2 1 2 4 2 2 2 2 3 2 4 2 3 2 ss 1 2 2 2 4 2 3 2 2 2 2 2 2 1 2 2 2 ss1 2 4 2 3 2 2
(27) (28)
( R1 4 K1 ( 4 r ) R2 42 K 0 ( 4 r )) (32)
The above differential equations are of the familiar Helmholtz type. The fundamental solution of Helmholtz differential equations for an only r-dependent fully symmetric two-dimensional domain is
with R1 2xax/r3 a/r, R2 xax/r2, 2 s2/ ( 2) and

2 2 K ss 1 K ss 2
i
K 0 ( i r ) , i 1, 2, 3, 4 2
(29)
By definition of 1, 2, 3 and 4, it is deduced:

1
S (1 Fs )s Sw Fs s a ( )kw ( )ka ( kwCaa Cww ka )s w kw Fs s 2 ( )kw kw ka 2 a ka (1 Fs )s ( )ka
2 3
4
2 2
1
3 2 3 1 2 2 2 2 2 3 1 3 4 1 4 2 2 2 2 2 4 1 4 2
(30)
K ss1 K ss 2
2 2
Replacing equation (29) into (30), one obtains

K 0 ( 1r ) ( 2 2 )( 2 2 )( 2 2 ) 3 1 4 1 2 1 K 0 2r 2 2 2 2 2 2 ( 2 1 4 )( 2 3 )( 2 1 ) K 0 ( 3r ) 2 ( 2 )kw ka 2 2 2 2 2 2 ( 3 2 )( 3 1 )( 3 4 ) K0 (4r ) 2 2 2 2 2 2 ( )( )( ) 4 1 4 2 4 3
2 Cwa CwwCaa Sw FsCaa kw ka kw ka 1 S C F s w wa kw ka w kw FsCaa s 2 Sa FsCwa Cww 1 Fs s kw ka kw ka k F C s w kwCwa 1 Fs s a a s wa k k k k w a w a a kaCww 1 Fs s kw ka
4.2 Water pressure caused by a Dirac force in the solid

P wS G 3
(31) in which K0(i r) is the modified Bessel function of the second kind of order zero with the argument r | x | which denoted the distance between a load point and an observation point. Finally, we can determine the components of fundamental solution tensor by applying the matrix of cofactors B*co to the scalar function which are (Maghoul et al. [8]):
Fs 2 ( 2 ) kw
2 2 2 2 2 2 ( 2 K wS )K1 ( 2 r ) ( 4 2 )( 3 2 ) x 3 2 K K r ( ) ( ) wS 3 1 3 r ( 42 32 )( 22 32 ) 4 2 ( 4 K wS ) K1 ( 4 r ) ( 2 2 )( 2 2 ) 3 4 2 4 (33) (Cwa (1 Fs ) FsCaa )s Fs ka
where K wS
1052
4.3 Air pressure caused by a Dirac force in the solid

P aS G 4 (1 Fs ) x 2 ( 2 )ka r
4.6
Air pressure caused by a Dirac source in the water fluid

s 2 ( 2 ) kw ka
P aW G 43
2 2 ( 2 K 4 )K1 ( 2 r ) 2 2 2 2 ( 4 2 )( 3 2 ) 3 2 ( K ) K ( r ) 2 3 4 1 3 2 2 2 ( 4 3 )( 2 3 ) 4 2 ( 2 2 )( 2 2 ) ( 4 K 4 )K1 ( 4 r ) 3 4 2 4 ( FsCwa Cww (1 Fs ))s kw (1 Fs )
(34)
( ( 2 )Cwa ( w kw s Sw )(1 Fs )) 22 Cwa s 2 K0 (2r ) 2 2 2 2 ( 4 2 )( 3 2 ) ( ( 2 )Cwa ( w kw s Sw )(1 Fs ))32 Cwa s 2 K 0 ( 3r ) 2 2 2 2 ( 4 3 )( 2 3 ) ( ( 2 )Cwa ( w kw s Sw )(1 Fs )) 42 Cwa s 2 K 0 ( 4 r ) 2 2 2 2 ( 3 4 )( 2 4 ) (37)
where K 4
4.7
Displacement caused by a Dirac source in the air fluid
4.4 Displacement caused by a Dirac source in the water fluid U W (Sw w kw s )s x G 3 2 ( 2 )kw r 2 ( 2 2 2 K 3 )K1 ( 2 r ) 2 2 2 ( 4 2 )( 3 2 ) 3 2 K K r ( ) ( ) 2 3 3 1 3 2 2 2 ( 4 3 )( 2 3 ) 4 2 ( 2 2 )( 2 2 ) ( 4 K 3 )K1( 4r ) 3 4 2 4 where K 3
(Cwa (Sa a ka s )s Caa (Sw w kw s )s ) ka (Sw w kw s )
U A (Sa a ka s )s x G 4 2 ( 2 )ka r 2 2 K 4 )K1( 2r ) ( 2 2 2 2 2 ( 4 2 )( 3 2 ) 3 2 ( K ) K ( r ) 2 3 4 1 3 2 2 2 ( 4 3 )( 2 3 ) 4 2 ( 2 2 )( 2 2 ) ( 4 K 4 )K1( 4r ) 3 4 2 4 (35) with 4.8

K 4 (Cwa ( Sw w kw s ) s Cww ( Sa a ka s ) s ) kw ( Sa a ka s )
(38)
Water pressure caused by a Dirac source in the air fluid K 34 2 2 2 2 K 0 ( 2 r ) ( 4 2 )( 3 2 ) K 34 ( ) K r 2 2 2 2 0 3 )( ) ( 4 3 2 3 K 34 ( ) K r ( 2 2 )( 2 2 ) 0 4 3 4 2 4 (39)
4.5 Water pressure caused by a Dirac source in the water fluid 2 K0 (2r ) 2 2 2 ( 2 K w )( 2 w ) 2 2 2 2 ( 4 2 )( 3 2 ) K 0 ( 3r ) ( 2 K 2 )( 2 2 ) w 3 w 2 2 2 2 3 ( 4 3 )( 2 3 ) K0 (4r ) ( 2 K 2 )( 2 2 ) 4 w 4 w 2 2 2 2 ( 3 4 )( 2 4 ) (36)
2 2 with Kw w Caa s3 /( 2 )ka and
P wA G 34
s 2 ( 2 ) kw ka
P wW G 33
1 2 kw
where 2 K34 ( ( 2 )Cwa ( a ka s Sa ))Fs ) 2 Cwa s 2 4.9 Air pressure caused by a Dirac source in the air fluid
2 2 2 2 ( 2 K a )( 2 a ) K 0 ( 2 r ) 2 2 2 2 ( 4 2 )( 3 2 ) 2 2 2 2 1 ( 3 K a )( 3 a ) K r ( ) 0 3 (40) 2 2 2 2 2 ka ( 4 3 )( 2 3 ) 2 2 2 2 ( 4 K a )( 4 a ) K 0 ( 4r ) 2 2 2 2 ( 3 4 )( 2 4 )
2 2 Kw w
Sa (1 Fs )s Caa s a ka (1 Fs )s 2 ka ( 2 )ka ( 2 )ka s 2 ( 2 )
P aA G 44
1053
2 2 a with K a
Cww s3 and 2 kw
2 2 Ka a
Sw Fs s C s k F s2 ww w w s ( 2 )kw ( 2 )kw kw s 2 ( 2 )
ANALYTICAL VERIFICATION OF THE FUNDAMENTAL SOLUTIONS. LIMITING CASE: ELASTODYNAMIC
fundamental solution in Laplace transformed domain is presented by the use of the method of Hrmander or Kupradze for 2D ui pw pa formulation of unsaturated porous media. The derived Laplace transform domain fundamental solutions can be directly implemented in time domain BEM in which the convolution integral is numerically approximated by a new approach so-called Operational Quadrature Methods developed by Lubich [6,7] to model the transient behaviour of unsaturated porous media. This enables one to develop more effective numerical hybrid BE/FE methods to solve 2D nonlinear wave propagation problems in the near future. REFERENCES
[1] Chen, J. 1994a. Time domain fundamental solution to Biots complete equations of poroelasticity: Part I two-dimensional solution, Int. J. Sol. Struct. 31(10) 14471490. [2] Gatmiri, B. 1997. Analysis of fully Coupled Behaviour of Unsaturated Porous Medium under Stress, Suction and Temperature Gradient, Final report of CERMES-EDF. [3] Gatmiri, B., Delage, P. & Cerrolaza, M. 1998. UDAM: A powerful finite element software for the analysis of unsaturated porous media, Adv. Eng. Software. 29(1) 2943. [4] Hrmander, L. 1963. Linear Partial Differential Operators, Springer: Berlin. [5] Kupradze, V.D., Gegelia, T.G., Basheleishvili, M.O. & Burchuladze, T.V. 1979. Three-dimensional Problems of the Mathematical Theory of Elasticity and Thermoelasticity, North-Holland, Netherlands. [6] Lubich C. 1988a. Convolution quadrature and discretized operational calculus, I. Numerische Mathematik 52 129145. [7] Lubich, C. 1988b. Convolution quadrature and discretized operational calculus, II. Numerische Mathematik, 52 413425. [8] Maghoul, P., Gatmiri, B. & Duhamel, D. 2009. Boundary integral formulation and two-dimensional fundamental solution for dynamic behaviour analysis of unsaturated soils, Soil Dynamics and Earthquake Engineering, submitted for publication.
Having derived the fundamental solution, at this stage, it is of interest to verify the validity of these solutions in the limiting case of elastodynamic. Letting kw and ka approach infinity and w, a and Fs equal zero, the unsaturated fundamental solutions presented in this study take the form of the elastodynamic fundamental solutions (Chen 1994): G G ij ij xi x j 1 a ij b 2 2 r 2 C2 (41) (42)
G G G 0 G 3 4 3j 4j where
a K0 (s.r/C2) C2(K1(s.r/C2) C2K1(s.r/C1)/C1)/ 2 2 s.r,C1 ( 2)/and C2 /
CONCLUSION
In this paper, firstly coupled governing differential equations of a porous medium saturated by two compressible fluids (water and air) subjected to dynamic loadings are presented based on the poromechanics theory in the frame of the suctionbased mathematical model presented by Gatmiri [2] and Gatmiri et al. [3]. After that, the associated
1054
Drying shrinkage and cracking initiation in clays: Main modeling concepts and application to the ventilation of a tunnel
H. Peron, J. Eichenberger & L. Laloui
Ecole Polytechnique Fdrale de Lausanne, Laboratory of Soil Mechanics, Lausanne, Switzerland
ABSTRACT: Damage induced by desiccation in clays (i.e. drying shrinkage and cracking) is an issue to be investigated in geo-environmental engineering, especially when the clay is used as a confining barrier for industrial and nuclear waste storage. Dramatic increase of the permeability and modifications of the mechanical properties of the barrier could ensue. This study first presents a modeling framework able to address the issue of desiccation in clays. The advanced constitutive model ACMEG-S, which relies on multi-mechanism hardening plasticity is used. Conditions for mode I (opening) cracks initiation such as desiccation cracks are supposed to be met when the minor principal effective stress becomes equal to a threshold value (like in Griffith criterion). A finite element analysis of the coupled hydro-mechanical transient processes is further performed. A tunnel excavated in Opalinus clay is considered, drying caused by its ventilation is simulated. The results of the simulation show the penetration of a drying front. The resulting effective stress distribution generates a gradual plastification at the neighborhood of the excavated gallery. It is shown how stresses develop during desiccation until a cracking criterion is reached. 1 INTRODUCTION saturation (Peron et al. 2009a). This paper presents a modelling framework able to address the processes of drying and desiccation cracking in clay. It is shown how stresses can be induced during drying, and a desiccation cracking stress criterion is proposed. These processes are further studied via the finite element modelling of the ventilation of an excavated gallery in clay. The evolution of suction, degree of saturation, stresses and strains is computed. The conditions of desiccation cracking initiation are discussed. 2 CONSTITUTIVE APPROACH FOR THE MODELLING OF DESICCATION
Desiccation cracking can appear in clay barriers such as landfill liners, top covers and buffers for nuclear waste isolation and strongly affect their permeability (Albrecht & Benson 2001). In particular, desiccation cracking is an issue in deep geological nuclear waste storage engineering. It affects engineered clay barriers which are submitted to thermal and hydraulic coupled loads induced by the confined nuclear waste (Graham et al. 1997). A dried and cracked zone has also been observed in the underground gallery host clayey material due to ventilation effects (Ramambasoa et al. 2000). More generally, desiccation cracking is also likely to alter the bearing capacity and/or overall stability of any type of earthen structures, such as dams, embankments and foundations (Morris et al. 1992). Cracking affects the mechanical bulk properties of soils, especially their strength and their compressibility. Furthermore, in soil science, desiccation cracks are of interest as they affect transport of gases, moisture and nutrients to the plant roots (Hillel 1998). Evaporation of the wetting liquid (generally water) from the deformable porous medium (here the clay) induces drying shrinkage. The occurrence of desiccation cracks is made possible when the drying shrinkage is constrained, for specific conditions of suction and degree of
The fundamental mechanical feature of geomaterials considered as deformable porous media relies on the notion of effective stress. Effective stress is a combination of both externally applied stresses and the internal pressure of fluid phases, and it enables the treatment of a multiphase porous medium as a mechanically equivalent single-phase continuum. For partially saturated media, it is expressed as (Nuth and Laloui 2008a):
ij ( ij pa ij ) Sr sij
(1)
where ij is the effective stress tensor component, ij is the total stress tensor component, pa is the air
1055
pressure, and ij is the Kronecker symbol. Sr is the degree of saturation, s is the suction. The direct dependence of the stress variable on suction and saturation degree bears the following coupling: drying, which corresponds to suction increase, will cause an increase in effective stress, and therefore shrinkage of the solid matrix. The other way round, wetting is at the origin of swelling. This is one of the features of the advanced constitutive model ACMEG-S. ACMEG-S constitutive model is made of two main parts (Nuth and Laloui 2008b). The mechanical part accommodates non-linear elasticity coupled with bounding multi-dissipative plasticity (based on critical state concept). The total irreversible strain increment is induced by two coupled dissipative processes: an isotropic and a deviatoric plastic mechanism. Each of them produces plastic strain increment. The mechanical part of the model features a progressive mobilization of the plastic mechanisms. The evolution of the preconsolidation pressure pc with plastic volumetric strains and suction is as follows:
p Pc pc 0 exp ( v ) if s sE p Pc Pc 0 exp ( v ) 1 s s log ( / ) s E if s sE (2)
accounting for the increase in t as suction increases. k2 has the dimension of stress, and k1 has no dimension. Equation 3 indicates that the soil gains brittleness as the drying process advances, since mode I cracking becomes likely to occur for higher and higher current effective stresses as suction increases. It is assumed that the failure criterion given by Equation 3 does not depend on hardening variables of the deviatoric and isotropic mechanisms. If hardening of the yield surfaces occurs at constant suction, the criterion remains unchanged. Finally, the resulting mode I crack should be perpendicular to the direction of the minor effective stress. In order to apply the above concepts, a coupled hydro-mechanical analysis of the non-homogeneous transient processes is performed in the following using the finite element method. 3 FINITE ELEMENT MODELLING OF THE VENTILATION OF AN EXCAVATED GALLERY Finite element model
3.1
where pc0 is the initial preconsolidation pressure, the plastic compressibility modulus, vp the plastic volumetric strain, s a material parameter and sE is the suction air entry value (for s sE, Sr 100%). The water retention part of ACMEG-S defines the soil water retention curve, i.e. the relationship between the degree of saturation and the suction. Its mathematical formulation is based on kinematic hardening, which permits to reproduce hysteretic water retention behaviour if needed. This part of the model features direct coupling with the mechanical stress-strain part of the model (Laloui and 2009). Conditions for mode I (opening mode) cracks initiation such as desiccation cracks are supposed to be met when the minor principal effective stress becomes equal to a threshold stress value (like in Griffith criterion). Based on experimental evidence, a criterion for the evolution of this threshold (denoted t) with respect to suction is proposed by Peron (2008): t tsat k2 1 exp ( k1s k2 ) (3)
The finite element code Lagamine is used (Colin et al. 2002) in which ACMEG-S constitutive model is implemented (Nuth 2009). Lagamine performs a coupled thermo-hydro-mechanical analysis of the non-homogeneous transient processes. In isothermal conditions three primary state variables are used: gas pressure, water pressure, and displacement vector. The solid phase component is assumed to be incompressible (incompressible grains) while the water phase is (slightly) compressible. A Lagrangian formulation is adopted in the code. 3.2 Assumptions of the simulation
tsat is the value of the threshold at the saturated state (i.e. zero suction). Except if the material is cemented, the value of tsat should not greatly differ from zero. k2 and k1 are material parameters
The chosen simulation example is a circular tunnel from Mont Terri Rock Laboratory, Switzerland. The isothermal drying of initially saturated Opalinus clay (the formation encountered at Mont Terri Rock Laboratory) that could be caused by ventilation of such an excavation. Isothermal conditions are assumed, having in mind that in reality, drying is a heat-consuming process. Bi-dimensional large strain finite elements are used. Each element has four degrees of freedom: two displacements for the soil skeleton, a pore water pressure, a gas (dry air vapour) pressure. Eight nodes isoparametric elements are used. For all the calculation results presented below, gas pressure is set constant and equal to atmospheric pressure. Therefore, only liquid transport is considered (darcean flux).
1056
A gallery with a diameter equal to 2 m. is studied. A plane perpendicular to the gallery axis is simulated, in plane strain conditions. A ca. 14 m. deep zone from the gallery wall is considered (7 times the gallery radius). The dimensions of the domain have been chosen according to existing benchmarking exercise; see for instance Chavant and Fernandez (2005). The finite element mesh, the geometry and the mechanical boundary conditions that prevail in all the analysis are represented in Figure 1. Taking advantage of the symmetry, only a quarter of the domain is actually considered. In-situ anisotropic stress state and pore water pressure are applied. The vertical in-situ initial stress in the host rock is 6.0 MPa while the initial in-situ horizontal stress is 4.5 MPa. Such a condition prevails in all the analysis. In-situ pore water pressure value is set to 2.0 MPa. Initial preconsolidation pressure is equal to 20 MPa. These are values typical of the simulated zone, taken from Coll (2005). Starting from this situation and before performing the hydro-mechanical analysis of drying, an initialization stage has been performed to establish an equilibrated stress state. A convergence of the tunnel wall is computed. The ventilation of the excavated gallery basically consists in imposing one suction value on all the nodes of the gallery wall (i.e. one specified pore water pressure value), during a period of one year. The suction is gradually increased up to 100 MPa in one month, then maintained constant during 11 months. Pore water pressure remains imposed equal to 2 MPa at the top and right boundaries of the domain. Using Kelvins equation, the final
value of suction (100 MPa) corresponds to a relative humidity of 47.8% in the gallery. 3.3 Model parameters
The ACMEG-S constitutive parameters of Opalinus clay have been determined from mechanical and water retention tests results. In particular, the water retention curve of Opalinus clay has been determined upon drying, consisting of the measurement of void ratio, degree of saturation and water content at given levels of suction (Peron et al. 2007). Two different suction imposition techniques have been used: axis translation technique and vapour equilibrium technique (for a suction range comprised between 0.02 and 327 MPa, Peron et al. 2009b). The main drying curve has been determined, starting from a saturated sample (Sr 100%). On this basis and using the hysteretic behaviour as characterized by Munoz (2006), the parameters of the water retention part of ACMEG-S have been determined. The values of k1 and k2 (Equation 3) are respectively 0.5 and 8 MPa (experimental data allowing for determining these parameters with confidence are currently not available). The parameter tsat is set to zero, this can be justified by the small tensile strength of Opalinus clay (Marschall et al. 2005). The intrinsic permeability of the medium is taken constant during the whole simulation and equal to 2 1020 m2. 4 RESULTS AND DISCUSSION
4.1 Penetration of drying The evolutions of the suction and the degree of saturation along a profile perpendicular to the tunnel wall are represented after one year of drying respectively in Figure 2 and in Figure 3. After one year of drying, the penetration of a drying front appears clearly. The degree of saturation starts experiencing values significantly lower than 100%
100
Suction, s [MPa]
80 60 40 20 0 2
1 day 1 month 1 year
2.5 3 3.5 4 4.5 Distance from the gallery axis, d [m]
Figure 1. Finite element mesh used for the simulation of drying of a gallery.
Figure 2. Evolution of the suction profile along the tunnel spring line.
1057
Degree of saturation, Sr [%]
100 90 80 70 60 50 40 2 2.5 3 3.5 4

1day 1 month 1 year
4.5
Distance from the gallery axis, d [m]
Figure 3. Evolution of the degree of saturation profile along the tunnel spring line.
at about two meters from the wall, down to 40% at the tunnel wall (which corresponds to the imposed suction value, i.e. 100 MPa). 4.2 Stress and strain fields
Figure 4. Horizontal effective stress around the tunnel after one year of drying. Compressive stresses negative.
Starting from a saturated state, the product suction times degree of saturation increases due to drying. This globally induces an effective stress increase. For the considered geometry and boundary conditions, the resulting horizontal and vertical effective stress distributions are shown in Figure 4 and Figure 5, respectively. Along the tunnel wall, the value of the effective stress normal to the tunnel wall is equal to about 41 MPa (see the horizontal and vertical effective stress values respectively at the spring line, Figure 4, and crown, Figure 5), which is exactly the value of the product Sr s at this location. The corresponding net stress is therefore zero. The drying generates a gradual plastification at the neighborhood of the excavated gallery. Plastic strains are caused by the progressive mobilization of isotropic plastic mechanism, which intervenes before initial preconsolidation pressure being reached. Most of this process is made possible by the in-situ confining. At this stage, a remark should be done: the simulations highlight that there is an overconsolidation process due to drying. In ACMEG-S, this is primarily due to strain hardening that occurs as long as the material stays saturated, i.e. the suction remains below air entry value (Nuth and Laloui 2008a, Peron et al. 2009a). 4.3 Initiation of desiccation cracking, discussion
Figure 5. Vertical effective stress around the tunnel after one year of drying. Compressive stresses negative.
In the presented finite element simulations, there is no procedure for modeling the formation of a mode I discontinuity, so we discuss here the conditions for desiccation cracking initiation only. The obtained effective stress repartition around the excavation may well be close to the mode I cracking criterion given by Equation 3, due to three different phenomena.
Firstly (and ignoring the influence of the drying process), the structural effect of the excavation within the host rock induces a redistribution and a release of stresses. In certain zones, the effective stress path is therefore likely to be shifted towards smaller stresses, closer to the cracking criterion. Such a process should tend to promote cracking tangential to the tunnel walls. This aspect is not deepened here, because it requires a more sophisticated modeling of the excavation process itself and associated changes in stresses and pore water pressure than the one presented here. Secondly, the drying shrinkage of the zone near the ventilated excavation wall tends to be restrained by the still humid material located farther from the wall. This tends to shift the minor effective stress
1058
path towards smaller stresses, which promotes desiccation cracking. Note that the minor stress is tangential to the tunnel wall and takes its smaller value at the surface (the major stress there is equal to the product Sr s, see Section 4.2). At the tunnel crown, the minor effective stress is horizontal; the desiccation cracks should therefore be orientated vertically there. However the anisotropy of Opalinus clay and the existence of bedding plane are not taken into account here. Most probably, this can affect the orientation of desiccation cracking (Tsang et al. 2008). Finally, the evolution of the cracking criterion with respect to suction, given by Equation 3 is such that, for the same effective stress level, cracking becomes more likely as suction increases (the clay becomes more brittle). In Figure 6, the evolution of the minor effective stress at a point located on the wall of the tunnel (crown) and a possible evolution trend for t with respect to suction are plotted (given by Equation 3). It is emphasized that values are plotted in terms of effective stress. With the chosen criterion, during drying, the effective stress path meets the conditions for mode I cracking. The simulation shows that the fracturing may intervene rather early in the drying process, for a relatively small value of suction and a corresponding high value of degree of saturation. These results highlight the possibility of desiccation cracking as a result of coupled hydromechanical processes around underground excavated zones in clay.
CONCLUSIONS
Some key features of a constitutive model able to cope with drying shrinkage and desiccation cracking initiation were presented. Then a coupled hydromechanical analysis of the non-homogeneous transient processes was performed using the finite element method for a tunnel excavated in clay and submitted to drying from its wall. This allowed highlighting how effective stress can evolve, as a result of changes in suction and degree of saturation. The elasto-plastic processes and the conditions for mode I drying fracture occurrence were also simulated and commented. ACKNOWLEDGEMENTS This work was funded by the Swiss National Science Foundation (grant # 200021 124702, Assessment of soil damage induced by drying), and by the National Cooperative for the Disposal of Radioactive Waste (Nagra). REFERENCES
Albrecht, B.A. & Benson, C.H. 2001. Effect of desiccation on compacted natural clay. Journal of Geotechnical and Geoenvironmental Engineering 127: 6775. Chavant, C. & Fernandes, R. 2005. Evaluating the reliability of hydro-mechanical simulation: a benchmark of numerical techniques carried out by research group of MoMaS. In 2nd International Meeting Clays in Natural and Engineered Barriers for Radioactive Waste Confinement, pp. 249250. Coll, C. 2005. Endommagement des roches argileuses et permabilit induite au voisinage douvrage souterrains. Doctoral Thesis, Universit Joseph Fourier, Grenoble. Collin, F., Li, X.L., Radu, J.P. & Charlier, R. 2002. Thermo-hydro-mechanical coupling in clay barriers. Engineering Geology 64: 179193. Graham, J., Chandler, N.A., Dixon, D.A., Roach, P.J., To, T. & Wan, A.W.L. 1997. The Buffer/Container Experiment: Results, Synthesis, Issues. Atomic Energy of Canada Limited Report, AECL-11746, COG-9746-I: Chalk River, ON. Hillel, D. 1998. Environmental Soil Physics. Academic Press, San Diego. Laloui, L., Franois, B., Nuth, M., Peron, H., and Koliji, A. 2008. A thermo-hydro-mechanical stressstrain framework for modeling the performance of clay barriers in deep geological repositories for radioactive waste. In Unsaturated Soils, Advances in Geo-Engineering, Toll et al. eds., CRC Press, pp. 6380. Laloui L., M. Nuth. 2009. On the use of the generalised effective stress in the constitutive modelling of unsaturated soils. Computer and Geotechnics, Vol. 36 (12): 2023.
100
80
Suction, s [MPa]
60
40
20
A possible tensile failure criterion
Minor effective stress, ' [MPa]

3
10
12
Figure 6. Evolution of the minor effective stress at the wall of the tunnel (crown) during the simulated period, and possible tensile failure criterion with respect to suction.
1059
Marschall, P., Horseman, S. & Gimmi, T. 2005. Characterisation of gas transport properties of the Opalinus clay, a potential host rock formation for radioactive waste disposal. Oil and Gas Science and Technology, rev. IFP 60(1): 121139. Morris, P.H., Graham, J. & Williams, D.J. 1992. Cracking in drying soils. Canadian Geotechnical Journal 29: 262277. Munoz, J.J. 2006. Thermo-hydro-mechanical analysis of soft rockApplication to a large scale heating test and large scale ventilation test. Doctoral Thesis, UPC Barcelona, Spain. Nuth, M. 2009. Advanced Modeling of Unsaturated Soils: Constitutive and Hydromechanically Coupled Finite Element Analysis. Doctoral Thesis, EPFL, Switzerland. Nuth, M. & Laloui, L. 2008a. Effective stress concept in unsaturated soils: clarification and validation of a unified framework. International Journal for Numerical and Analytical Methods in Geomechanics 32: 771801. Nuth, M. & Laloui, L. 2008b. Advances in modelling hysteretic water retention curve in deformable soils. Computers and Geotechnics 35(6): 835844.
Peron, H., Hueckel, T. & Laloui, L. 2007. An improved volume measurement for determining soil water retention curve. Geotechnical Testing Journal 30(1): 18. Peron, H. 2008. Desiccation Cracking of Soils. Doctoral Thesis, EPFL, Switzerland. Peron, H., Hueckel, T., Laloui, L. & Hu, L.B. 2009a. Fundamentals of desiccation cracking of fine-grained soils: experimental characterisation and mechanisms identification. Canadian Geotechnical Journal 46(10): 11771201. Peron, H., Salager, S., Eichenberger, J., Rizzi, M. & Laloui, L. 2009b. Experimental and numerical analysis of excavation damaged zone (EDZ) along tunnels. Mont Terri Project, Technical note 200854. Ramambasoa, N., Nguyen-Minh, D. & Rejeb, A. 2000. Hydromechanical behavior of a shale - Application to the Tournemire site. In 4th NARMS, Pacific Rocks: Rock around the Rim, pp. 10671072. Tsang, C.F., Blumling, P. & Bernier, F. 2008. Coupled hydro-mechanical processes in crytalline rock and in indurated and plastic clays: A comparative discussion. Lawrence Berkeley National Laboratory: LBNL Paper LBNL-63369.
1060
Behaviour of an unsaturated soil column under loading and watering

F. Salehnia B. Gatmiri
Department of Civil Engineering, School of Engineering, University of Tehran, Tehran, Iran Center of Excellence of Infrastructure-University of Tehran, Tehran, Iran Universit de Paris-Est, Inst. Navier, CERMES, Ecole des Ponts et chausses, Paris, France
ABSTRACT: The paper presents a comparison between the behaviour response of an unsaturated soil column under loading and watering. This comparison has been done by the code -Stock (Gatmiri 19972009), that the framework of a fully coupled thermo-hydro-mechanical formulation for an unsaturated porous medium using two state variables, net stress and suction, has been incorporated into it. This formulation has shown its great capability to present all significant interaction effects among the different components in a deformable unsaturated medium under heating. The temperature-dependent state surfaces of void ratio and degree of saturation play an important role in considering the coupling effects of moisture content, deformation of skeleton and temperature. After a brief description of the formulation, behaviour of an unsaturated soil column is studied under different mechanical loadings and watering. On the whole, the main differences between the effects of these types of loadings are discussed based on the obtained results. 1 INTRODUCTION between liquid and gas, evaporation, condensation and the effects of moisture distribution on heat flow are also considered too. The first step in a theoretical development of a fully coupled thermo-hydro-mechanical model for an unsaturated soil, is choosing the adequate and independent variables, which would be able to present all significant interaction effects among the different components involved in a coupled process in a deformable unsaturated porous medium with three phases (skeleton, water and air) under heating. The model, which is used in this study has been formulated with two most widely used independent state variables: net stress and suction. It describes the water and air pore pressures distribution, and the deformation of the skeleton. In the framework of the fully coupled thermohydro-mechanical formulation, which has been presented by Gatmiri (1997), the state surfaces concept plays an important role in presenting the coupling effects for the elastic behaviour of soil. Since the medium is assumed to be deformable, the effects of temperature and moisture content on the deformation of the solid skeleton, and their inverse effects must be included, which are taken into account in this formulation, via the thermal state surfaces concept. In the following, a set of the fully coupled thermo-hydro-mechanical equations for unsaturated porous media behaviour are given,
Since the soil is continuously under the effects of temperature changes in its natural environment, a great deal of attention has been paid to the phenomenon of moisture transport due to thermal gradient. Considering the usage of nuclear energy in the world, the disposal of high-level radioactive waste in two phases or multi-phases geological formations is an example, which has a great importance in geotechnical and geo-environmental engineering under the temperature changes. Therefore, the complete coupling should be considered among thermal, hydraulic and mechanical effects in a deformable unsaturated media. In unsaturated soils, movements occur in both vapor and liquid phases. The theory of Philip & de Vries (1957) is known as a basic framework and a comprehensive theory of moisture and heat movements in an incompressible porous medium, which has been used in this study. This theory assumes that the moisture transfer in an unsaturated soil occurs in both vapor and liquid phases, under the combined influences of gravity, gradient of temperature and gradient of moisture content. In this development, in the absence of water continuity, all of the transfers are in the vapor phase and with increasing moisture content, the liquid phase transfer becomes dominant. The phase changes
1061
the void ratio and the degree of saturation state surfaces are also introduced. In this paper, the finite element code, -Stock, which has been developed by Gatmiri (19972009), and the mentioned thermo-hydro-mechanical framework has been incorporated into it, is used for the comparsion between the effects of different mechanical loadings and watering in an unsaturated porous medium. 2 THEORETICAL FORMULATION
This equation can be expressed by: q w DT T D Dw Z (2)
In the development of this formulation, two basic theories have been modified and combined in order to describe a fully coupled behaviour of an unsaturated porous medium under heating. In one hand, the nonlinear theory of isothermal behaviour of unsaturated soils under the coupled effects of net stress and suction has been extended to non-isothermal conditions. The reasoning is found on the concept of state surfaces of void ratio and degree of saturation. In the other hand, the Philip & de Vries theory of heat and moisture transfer has been modified in order to take the deformation of the skeleton into account. This fully coupled formulation has been presented in a new suction-based formulation, which is more suitable for a combination with the deformation theory of unsaturated soils. Relative to three phases of unsaturated soils and temperature, four basic sets of equations must be established and solved simultaneously. These equations are briefly presented below. More details about the formulation can be found in Gatmiri (1997), Gatmiri & Hoor (2007), Gatmiri & Salehnia (2009) and Salehnia (2010). 2.1 Moisture phase movement equations
where V vapor velocity; U water velocity; DT thermal moisture diffusivity and is equal to DTv DTw; D isothermal moisture diffusivity and is equal to Dv D w, in other words, this two terms are divided to vapor and water diffusivities. Dwz gravitational part of the equation. Concerning the moisture mass conservation law, the same concept is used in which by combining it with Equation 2, the following suction-based form for the moisture phase equation in an unsaturated soil is presented: nSr T
T S nSr P w w v n r t t t v n n 1 Sr Sr w v 1 Sr t t div( w Dwz ) div w DT T

div w Dp Pw Pg
(3)
where n porosity; Sr degree of saturation in water; T temperature; T derivative of water density with regard to the temperature in the constant water pressure; p derivative of water density with regard to the water pressure in the constant temperature; w and v respectively, water and vapor densities; Pg and Pw water and vapor pressure, respectively; and DP DPw DPv which are the new suction-based form of the vapor and water diffusivities. 2.2 Gas phase movement equations The generalized Darcys law for the motion of gas in soil is considered that it has been completed by including the effect of temperature gradient. Hence, the gas flow equation is written as: qg K g Pg Vg T g T g Pg K g Z g (4)
As it has been described, the moisture phase is consisted of the vapor and liquid phases. All of the transfers in excess of the ones, which occur exclusively in the liquid phase, are called vapor transfer. Equation 1 shows that the total moisture movement in unsaturated soil due to temperature gradient and its resulting moisture content gradient is the superposition of the flows that take place separately in each phases, vapor and liquid. Hence, the total moisture transfer governing equation as a sum of the liquid and vapor velocities can be written as the following: qvap qliq q w w w V U
DTv DTw T Dw Z Dv Dw
(1)
where Vg vector of gas velocity; qg vector of gas flow; g gas density; Kg air permeability; g specific weight of gas; and Pg gas pressure. Considering the governing mass conservation law for the gas phase in a control volume of an unsaturated porous medium, the general partial differential equation of gas movement for
1062
unsaturated cases in the final suction-based form is expressed as the following:
g 1 Sr 1 H
g n n 1 Sr 1 H t t S 1 H n g r div g Dw Z K H g g t div K g g Pg H g DTw w DTv T

K div g g H g DPw w DPV Pg g div H D D P g Pw w Pv w
The formulation of this thermal surface, depending on stress, suction and temperature has been proposed by Gatmiri et al. (1997) as:
e 1 e0

p p p p 1 m g g g w a patm b 1 c p atm exp exp[ c (T T0 )] K b (1 m )
1
(10) (5) where a; b; and c constants; and c preconsolidation stress. Through this equation, the compatibility with non-linear behaviour of soil is also ensured. 2.3.4 Thermal degree of saturation sate surface To model an unsaturated soil submitted to stress and suction needs to the description of the coupling of volumetric moisture content with temperature. Hence, the following state surface of degree of saturation, based on experimental data is expressed: Sr 1 [as bs ( pg)] [1 exp(cs(pg pw))] exp(ds(T T0)) where as; bs; cs; and ds constants. (6) 2.4 Heat flow equations Following the theory of Philip & de Vries, the total flow of latent and sensible heat in an unsaturated porous medium can be written as:
Q gradT C pw wU C pv wV C pg gVg T T0 w h fgV vVg h fg
where H Henrys constant, which corresponds to the dissolution of air in water; and Pg derivative of gas pressure regard to the temperature. 2.3 Solid skeleton behaviour Suction and net stress have been studied as two stress state variables. The equilibrium equation, the constitutive law of a non-isothermal, isotropic and non-linear case based on the isothermal equations, and the thermal state surfaces of void ratio and degree of saturation are expressed in this section. 2.3.1 Equilibrium equation The equilibrium equation can be written as: (ij ij pg),j pg,j bi 0 where bi volumetric forces. 2.3.2 Incremental constitutive law The constitutive law for the solid skeleton of an unsaturated soil, which is under suction and thermal effects, with the assumption of small deformations is as the following: d ( ij ij pg ) Dd Fd ( pg pw ) CdT F DDs1, Ds1 = s m, e 1 s 1 e ( pg pw ) C DDt1, Dt1 = t m, t 1 e 1 e T (7)
(11)
(12) where T0 an arbitrary reference temperature; hfg latent heat of vaporization; Cpw; Cpv; and Cpg water, vapor and gas heat capacities, respectively; and Fourier heat diffusion coefficient, and it can be expressed by: (1 n)s w (n )v (13)
(8)
(9)
where T temperature; e void ratio; D a nonlinear and temperature-dependent stress-deformation matrix; and mT = 1 1 0. 2.3.3 Thermal void ratio sate surface The void ratio state surface is used to account the volumetric strain for calculating the bulk modulus.
where s; w; and v soil, water and vapor heat diffusion coefficients, respectively. The energy conservation equation in a porous medium can be written as:
(14) divQ 0 t where Q heat flow; and the volumetric bulk heat content of the medium that can be expressed by Equation 15 in case CT specific heat capacity
1063
of the unsaturated mixture, which is written as Equation 16:
CT T T0 n v h fg
CT 1 n sCPs wCPw n vCPv n gCPg
(15)
Reference to Gatmiri & Arson (2008) for finding the details of the matrixes and general initial and boundary conditions is advantageous. 4 ANALYSIS AND COMPARSION
(16)
The final differential form of heat flow equation in an unsaturated porous medium is found by combining Equation 12 to 16:
CT
T t
T T0
CT t
1 S r v h fg
n t
n v h fg
Sr t
n 1 S r h fg C Pw w div C Pv w div C Pg div
v t
w h fg div DTv T DPv Pw Pg

Pw w g w

0

0
K T K h fg div K T K
g g Pg g v g Pg v g
D T D P P D Z T T D T D P P T T
Tw Tv Pv w g
The applicative example which has been studied and its results are discussed in this section is a plain strain example of an unsaturated soil column. This column includes 152 nodes with 75 unsaturated quadrilateral elements, which are in one vertical row starting from the top of the column (Fig. 1). This column is studied under two different mechanical loadings and watering. Every case has been analyzed under the consideration of a linear elastic behaviour using the code -Stock, and their main results are presented below. 4.1 Initial and boundary conditions The initial conditions are defined similarly in every studied case. Suction of 200 KPa and temperature of 20C are introduced in all of the nodes. All of the nodes are limited in the horizontal direction, and only the bottom nodes are limited in the vertical direction. Other boundary conditions are different in the loading and watering cases. In the first one with mechanical loading, just the upward nodes are under the pressure, which is applied in the first load step and its effects are studied during a year that is our analysis period. Furthermore, the soil column in this case is analyzed under two different mechanical loadings including 1.4e5 N and 7e5 N. In the second one namely watering case, there is not any mechanical loading, and the soil column is just under watering from the bottom nodes. Thus, suction in these nodes are reduced from the initial amount to zero progressively during the analysis and maintained at zero
Pg
g g
Z Z
Pg

T T0
(17)
div T 0
GLOBAL SOLUTION MATRIX FORM
Since the governing partial differential equations of the above formulation are complex and development of the analytical solutions seems to be very difficult even for the simple conditions, and probably impossible for the real and general boundary and initial conditions, the known numerical approaches such as finite difference, finite element or boundary element methods can be used for resolving the mentioned formulation. Considering the finite element technique, with application of the weighed residual method, and the weighed functions of Galerkin in terms of nodal points values of the field variables, the total spatial domain has been separated. Then, by applying the single-step integration in time, a global matrix form, which has been encoded by the second author in a practical finite element code, -Stock, based on the mentioned equations is represented as:
C Cw C R C C t K C t K C t K C C t K C t K C 0 C C t K C t K C t K 0 F u F K T K P K P t F T t F K T K P t F P F K T K P K P t F P

T g Tu TT TT Tw Tw Tg Tg wu gu wT gT wT ww gw ww gw wg gT gg gg r T0 TT 0 Tw w0 Tg g0 T w g w0 wT 0 ww w0 w g0 gT 0 gw w0 gg g0 g
(18)
Figure 1.
Geometric details of the soil column.
1064
up to a year, that obviously, it affects throughout the entire domain. 4.2 Results
In every opposed nodes, the results are the same by symmetry. Nodal vertical displacements after a year for two different mechanical loadings and watering cases are shown in Figure 2. Every case has nodal settlement, which is increased by receding from the origin. However, the settlements in the case under farther loading are more than the less one, as it was expected. That is to say, with increasing the mechanical loading, the nodal vertical displacement is increased too. As well as, in the loading case, elasticity modulus is the most effective parameter in the settlement amount, as compared to the watering case that suction modulus is dominant. Therefore depending on their effects, the amounts of settlements in the studied soil column under watering are between the amounts of them under two mentioned different mechanical loadings. For the next step, the comparsion between degree of saturation of the elements after the analysis period is noticeable. In every case, the degree of saturation is farther in the inferior elements, although in the watering case, because of the watering from the bottom of the soil column, this growth is clearer than the loadings cases. Also, there is not any notable difference between the variations of degree of saturation in the elements under two mentioned different mechanical loadings, actually, both of them do not make any observable change in the degree of saturation amounts of their elements. These results are given in Figure 3. Suction of the elements in the loading cases do not notably change. However, suction in the elements of the soil column under watering, contrary to their degree of saturation (Fig. 3), is increased by receding from the bottom of the column. The reasoning is found on the watering from the bottom nodes of the studied column. Figure 4, shows the variations of suction in the elements of the soil column under watering in a year. As it was mentioned, in every case of the mechanical loadings and watering, the nodal settlement is observed. Now, for displaying the increment of nodal vertical displacement by passing the time, the settlement of node 72 as an example is presented in Figure 5, for two different mechanical loadings and watering cases during a year, which is our analysis period. The nodal settlement in every case is increased during the time, although it is farther in the case under more loading as it was discussed in the description of Figure 2.
Figure 2. Nodal vertical displacement versus their distance from the origin, after a year.
Figure 3. Degree of saturation of the elements versus their distance from the origin, after a year.
Figure 4. Suction of the elements versus their distance from the origin, after a year.
For displaying the reduction of suction of the elements in the watering case, the variations of suction in element 30 as an example, by passing the time in a year, is shown in Figure 6. As it is clear,
1065
Figure 5.
Vertical displacement of node 72 versus time.
example for comparsion between the effects of two different mechanical loadings and watering in an unsaturated soil column during the time has been performed. As the research shows, nodal settlement in every case is increased by receding from the origin and also, during the time. In the other side, degree of saturation of the elements is farther in the inferior elements that this difference is more considerable in the watering case because of the watering from the bottom of the soil column. Contrary to the degree of saturation, suction in the watering case is increased by receding from the origin and, is reduced by passing the time because of the watering. On the whole, there is more settlement when the soil column is under farther loading and, more variations of suction and degree of saturation when this column is under watering. REFERENCES
Gatmiri, B. 1997. Analysis of fully coupled behaviour of unsaturated porous media under stress, suction and temperature gradient. Final report. Paris: CERMESEDF. Gatmiri, B. & Arson, C. 2008. -Stock, a Powerful tool of thermohydromechanical behaviour and damage modelling of unsaturated porous media. Computers and Geotechnics 35(6): 890915. Gatmiri, B. & Hoor, A. 2007. Excavation effect on the thermo-hydro-mechanical behaviour of a geological barrier. Physics and Chemistry of the Earth 32(814): 947956. Gatmiri, B. & Salehnia, F. 2009. An axisymetric elastoplastic modeling for unsaturated soils. Proc. 2nd International Conference on New Development in Soil Mechanics and Geotechnical Engineering, Nicosia, Cyprus: 456463. Gatmiri, B., Seyedi, M., Delage, P. & Fry, J.J. 1997. A new suction-based mathematical model for thermohydromechanical behaviour of unsaturated porous media. NUMOG VI, Montreal, Canada: 291296. Philip, J.R. & de Vries, D.A. 1957. Moisture movement in porous materials under temperature gradients. Trans Am Geophys Un 38: 222232. Salehnia, F. 2010. Analysis of the soil media using the theory of unsaturated soils with nonlinear behaviour. M.Sc Thesis. Tehran: University of Tehran.
Figure 6.
Suction of element 30 versus time.
suction in this element and similarly in the other elements is reduced during the time. 5 CONCLUSIONS
The theoretical framework for the analysis of fully coupled behaviour of the different phases in a deformable unsaturated porous medium, including moisture, gas and solid skeleton under heating has been cited in this paper. Using the code -Stock incorporating this THM formulation, a practical
1066
Comparison of the nonlinear elastic and elastoplastic models in THM response of unsaturated soils
F. Salehnia B. Gatmiri
Department of Civil Engineering, School of Engineering, University of Tehran, Tehran, Iran Center of Excellence of Infrastructure-University of Tehran, Tehran, Iran Universit de Paris-Est, Inst. Navier, CERMES, Ecole des Ponts et chausses, Paris, France
ABSTRACT: The paper presents a comparison between the nonlinear elastic and elastoplastic models for Thermo-Hydro-Mechanical (THM) behaviour of an unsaturated medium. A fully coupled thermo-hydro-mechanical formulation for an unsaturated soil is considered using net stress, suction as two state variables. In this formulation, all significant coupled interactions between the different components involved in an unsaturated porous medium subjected to heating have been taken into account. The coupling effects of temperature, moisture content and deformation of skeleton are included in the nonlinear elastic model, via the temperature-dependent state surfaces concept. After a brief description of the formulation, the basic principles of the elastoplastic model are described, which considers a temperaturedependent yield surface with an associated flow rule. Then, the strong ability of this model is shown by presenting the results of analyzing of an applicative example using this model, and compared with the nonlinear elastic model using the code -Stock (Gatmiri 19972009). 1 INTRODUCTION gas, evaporation, condensation, induced moisture transfer under thermal and pore pressure gradients and the effects of moisture distribution on heat flow. In this study, the theory of Philip & de Vries (1957) that is known as a basic framework and a comprehensive theory of moisture and heat movements in an incompressible porous medium, has been used. This theory assumes that the moisture transfer in an unsaturated soil occurs in both vapor and liquid phases, under the combined influences of gravity, gradients of temperature and moisture content. When deformation of the medium is significant, coupled effects of deformation, moisture content and heat, and their inverse effects must also be taken into account. These coupled effects have been included in the framework of fully coupled thermo-hydro-mechanical formulation that has been presented by Gatmiri (1997), via the thermal state surfaces concept of void ratio and degree of saturation, which have an important role in the coupled nonlinear elastic behaviour of unsaturated soils. The mentioned set of the fully coupled thermo-hydro-mechanical equations for unsaturated porous media behaviour are briefly given. Because of the complexity of the established governing partial differential equations, a numerical resolution scheme has been developed, and a finite element code, -Stock (Gatmiri 19972009), which
The first step in a theoretical development of a fully coupled thermo-hydro-mechanical model for an unsaturated soil, as it has been expressed in Gatmiri (1997), is choosing the adequate and independent variables, which would be able to present all significant interaction effects among the different components involved in a coupled process in a deformable unsaturated porous medium with three phases (skeleton, water and air) under heating. Two most widely used independent state variables are net stress and suction that the using models in this study have been formulated based on them. Also by utilizing these parameters, the water and air pore pressures distribution, and the deformation of the skeleton are described. In reality, the effects of temperature changes in the natural environment of soil are caused to pay a great deal of attention to the temperature effects in a coupled thermo-hydro-mechanical formulation in a deformable unsaturated media. The disposal of highlevel radioactive waste in two phases or multi-phases geological formations is an example which has a great importance in geotechnical and geo-environmental engineering, under the temperature changes. The important aspects which should be considered, include the phase changes between liquid and
1067
has been integrated with the mentioned thermo-hydro-mechanical framework is used for comparing the results of using the nonlinear elastic and elastoplastic models in THM response of the unsaturated soils. This comparsion that is presented in this paper, is done for evaluating the incorporated elastoplastic model into the code -Stock by the authors and displaying the main differences between the results of these models, observantly that despite the fact of simplicity of the nonlinear elastic model, considering the elastoplastic model shows the soil behaviour closer to the real one. 2 HISTORY OF THM FORMULATION
2.1.2 Constitutive law With the assumption of small deformations, the incremental constitutive law for the solid skeleton of an unsaturated soil, which is under suction and thermal effects is as the following: d(ij ij pg) Dd Fd(pg pw) CdT F DDs1, Ds1 = s m, s 1 e 1 e ( pg pw ) 1 e 1 e T (2) (3)
C DDt , Dt = t m , t
1
1
(4)
Two basic theories have been modified and combined in the developement of this formulation for describing a fully coupled behaviour of an unsaturated porous medium under heating. In one hand, the Philip & de Vries theory of heat and moisture transfer has been modified in order to take the deformation of the skeleton into account. In the other hand, the nonlinear theory of nonisothermal behaviour of unsaturated soils under the coupled effects of net stress and suction, which is based on the earlier works by the second author has been used. The reasoning is found on the concept of state surfaces of void ratio and degree of saturation. This fully coupled formulation has been presented in a new suction-based formulation, which is more suitable for a combination with the deformation theory of unsaturated soils. In the following, four basic sets of equations relative to the temperature and three different phases of unsaturated porous media are briefly presented as they can be found with more details in Gatmiri (1997), Gatmiri & Arson (2008) and Salehnia (2010). These equations must be solved simultaneously. 2.1 Solid skeleton equations
where D a non-linear and temperature-dependent stress-deformation matrix; e void ratio; T temperature; and mT [1 1 0]. 2.1.3 Thermal void ratio sate surface Thermomechanical and hydromechanical coupling coefficients and volumetric strain in order to calculate the bulk modulus, can be accounted via a void ratio state surface, which depends on stress, suction and temperature as follows:
e 1 e0
pg pg 1 m a b 1 c patm pg pw p exp[ c (T atm exp K b (1 m )
1
T0 )]
(5) As it was mentioned, net stress and suction are two using sate variables. Therefore, considering them, the equilibrium equation, the constitutive law of a non-linear, isotropic and non-isothermal case based on the isothermal equations, and the thermal state surfaces of void ratio and degree of saturation are introduced in this section. 2.1.1 Equilibrium equation For an unsaturated porous media, the equilibrium equation is written as follows: (ij ij pg), j pg, j bi 0 where bi volumetric forces. (1) where a; b; and c constants; and c preconsolidation stress. The compatibility with nonlinear behaviour of soil is also ensured in this equation. 2.1.4 Thermal degree of saturation sate surface Although the stress-strain behaviour is already coupled with temperature, for modeling an unsaturated soil submitted to stress and suction, the description of the coupling of volumetric moisture content with temperature is also needed. Hence, based on experimental data, it can be expressed by: Sr 1 [as bs( pg)] [1 exp(cs(pg pw))] exp(ds(T T0)) where as; bs; cs; and ds constants. (6)
1068
2.2 Moisture phase equations Moisture is composed of the vapor and liquid phases. All of the transfers in addition to ones which occur exclusively in the liquid phase, are called vapor transfer. The superposition of the flows that take place separately in each phases, vapor and liquid, constitute the total moisture movement in unsaturated soil due to temperature gradient and its resulting moisture content gradient. Thus, the total moisture transfer governing equation as a sum of the liquid and vapor velocities can be written as: q DT T D Dw Z w (7)
Vg
qg
g
K g Pg Pg T K g Z g T g
(11)
where Vg vector of gas velocity; qg vector of gas flow; g gas density; g specific weight of gas; Pg gas pressure; and Kg air permeability, which can be defined by: Kg c
g
g
e 1 Sr d
(12)
where c; and d constants; and g gas viscosity. The governing partial differential equation for the mass conservation of the gas phase in a control volume in an unsaturated porous medium can be written as:
where DT thermal moisture diffusivity which is equal to the sum of the thermal vapor and water diffusivities; D isothermal moisture diffusivity that is equal to the sum of the isothermal vapor and water diffusivities; and Dwz gravitational part of the equation and can be expressed by: S Sru T Dw K wz K wz 0 r 1 Sru r
b
n 1 Sr HSr div gVg t g div g HU w divV
(13)
(8)
where H Henrys constant, which corresponds to the dissolution of air in water. Substituting the equations relative to liquid, vapor and gas velocity under thermal and pressure gradients into Equation 13, the final suction-based equation of gas continuity is obtained. 2.4 Heat equations The total flow of latent and sensible heat in an unsaturated porous medium, based on the Philip & de Vries theory can be expressed by: Q gradT [C pw wU C pv wV C pg gVg ] (T T0 ) h V V h v g fg w fg (14)
where Sr degree of saturation in water; dynamic viscosity of water; and Kwz0 saturated soil water permeability. More details about these parameters can be found in Gatmiri (1997). The moisture mass conservation law, considering phase changes between water and vapor can be written as:
m div w U V t
(9)
where U water velocity; V vapor velocity; w water density; and m moisture density, which is defined by: m w (n ) nSr w n(1 Sr) (10)
where Cpw; Cpv; and Cpg water, vapor and gas heat capacities, respectively; T0 an arbitrary reference temperature; hfg latent heat of vaporization; and Fourier heat diffusion coefficient, which can be evaluated by: (1 n)s w (n )v (15)
where n porosity; and v vapor density. By combining Equations 7 and 9, the final suction-based form of the moisture phase equation in an unsaturated soil can be found. 2.3 Gas phase equations For the motion of gas in soil, the generalized Darcys law is used in which the dependence of the gas pore pressure on temperature is also considered. Therefore, the gas flow equation is written as the following:
where s; w; and v soil, water and vapor heat diffusion coefficients, respectively. The energy conservation equation in a porous medium can be written as the following:
divQ 0 t
(16)
where Q heat flow; and volumetric bulk heat content of medium, which can be expressed as: CT (TT0) (n)vhfg (17)
1069
where CT specific heat capacity of the unsaturated mixture that is defined by: CT (1 n)sCPs wCPw (n ) vCPv (n ) gCPg (18)
The combination of Equations 1418 gives the final equation of heat flow in an unsaturated porous medium. 3 SOLUTION PROCEDURE
Considering the complexity of the governing partial differential equations in this coupled formulation, it seems to be probably impossible developing the analytical solutions for the real and general boundary and initial conditions, and very difficult even for the simple conditions. Therefore, the numerical approaches such as finite difference, finite element or boundary element methods can be used to obtain approximate solutions to the presented formulation. In the code -Stock, that the second author has encoded, the governing differential equations are solved using a finite element technique. With application of the weighed residual method, and the weighed functions of Galerkin in terms of nodal points values of the field variables, the total spatial domain has been separated. Next, by applying the single-step integration in time, the following global matrix form was presented: R C C C C C C
Tu wu gu
of temperature including shrinkage of the elastic domain due to temperature increase, and dependence of the yield surface and plastic modulus on temperature. In this paper, this model is used for studying the elastoplastic behaviour of an unsaturated medium. Considering the nominative model, the basic elements of thermoplasticity for unsaturated porous media are briefly discussed below. Details of the parameters in this section can be found in Gatmiri (2000), Jenab (2000), Gatmiri & Arson (2008) and Salehnia (2010). 4.1 Yield surface Yielding can occur in a classical plasticity theory with satisfying the following criterion: F(, P, ) 0 (20)
where mean stress; p plastic strain; and hardening parameter. Extending the Barcelona model to the nonisothermal case by considering that the variation of the yield locus under temperature effects is properly represented by the change in maximum isotropic preconsolidation stress pc, which is a function of suction, accumulated thermoplastic strain and temperature, is introduced as:
tp F ij , s, v ,T q 2 M 2 p ps p0 p 0
TT
wT
C C C t K C t K C t K t K C t K C 0

T w g TT Tw Tw Tg Tg
(21) where q shear stress; M slope of critical sate line; p net mean stress; ps parameter describing the increase in cohesion with suction and temperature; and p0 relative parameter to the yield locus in a specific suction, which is depended on one related to the saturated state (pc). 4.2 Flow rule If dp indicates the increment of the plastic strain during the plastic deformation, considering an associated flow rule, one can obtain: d p F (22)
gT
t K
gT
wT
C t K C t K
gw gw gg
ww
ww
wg
gg

0 u FT 0 K TT T0 K Tw Pw 0 K Tg Pg 0 T t Fw 0 K wT T0 K ww Pw 0 Pw Fg 0 K gT T0 K gw Pw 0 K gg Pg 0 Pg

t F t F t F

Fr
T w g
(19) The details of matrixes and initial and boundary conditions can be found in Gatmiri & Arson (2008). 4 THERMOELASTOPLASTIC CONSTITUTIVE LAW
Based on the concept of Barcelona elastoplastic Basic Model (BBM), which has been suggested by Alonso et al. (1990), a thermoelastoplastic model for analyzing the behaviour response of unsaturated soils has been proposed by Gatmiri (2000), which was incorporated into the code -Stock by the authors. This model considers the main effects
where plastic multiplier. Since the infinitesimal increment of stress probably causes the plastic straining, the total strain changes is written as the sum of the elastic strain and the plastic one as: d de dp (23)
1070
The final general form of the stress-strain relationship can be obtained as the following by finding the plastic multiplier and introducing it into Equation 22:
p p d D e D p d DeT DeT DeTe dT
Des

p Des
p Dese
ds
(24)
The radial displacement throughout the entire domain and the vertical displacement of the bottom nodes are zero. Also, during the analysis, the soil column is under watering from the bottom nodes. Hence, suction in these nodes are decreased from the initial amount up to zero progressively and maintained at zero during the remnant time until 10 years, which is our analysis period. As it is expected, the whole domain is affected by it. These initial and boundary conditions are the same for nonlinear elastic and elastoplastic analysis. 5.2 Results
APPLICATION AND COMPARSION
For comparing the nonlinear elastic and elastoplastic behaviours of unsaturated soils, an applicative axisymmetric example has been studied for both of them using the code -Stock, incorporating the mentioned thermoelastoplastic model (TBBM) for analyzing the elastoplastic behaviour response and the Hyperbolic model for the nonlinear elastic behaviour response in which this nonlinear model has been modified by Gatmiri (1997) to take the effects of temperature and suction into account. This example is an unsaturated soil column, which includes 152 nodes with 75 unsaturated quadrilateral elements, as shown in Figure 1, the beginning of these elements is top of the column and they stand in one vertical row. This column is analyzed under watering using two mentioned models and besides of evaluating them, with comparing their results that are presented in this section, the basic differences of using them are discussed. 5.1 Initial and boundary conditions
5.2.1 Nonlinear elastic model Vertical displacements and water pressures in three and four different points of time are shown in Figures 23, respectively. Nodal settlement is increased by receding from the origin (Fig. 2). Also, considering watering from the bottom nodes of the soil column, the negative water pressure is farther in the upper vertical distances from the origin (Fig. 3). In reality, suction is decreased in the bottom nodes more than the upper nodes. Figures 45, respectively show the variations of the vertical displacement and water pressure during the
The initial temperature and suction in all of the nodes are assumed to be 20C and 200 KPa, respectively.
Figure 2. Vertical displacement versus distance from the origin, nonlinear model.
Figure 1.
Geometric details of the soil column.
Figure 3. Water pressure versus distance from the origin, nonlinear model.
1071
Figure 4. model.
Vertical displacement versus time, nonlinear
Figure 7. Water pressure versus distance from the origin, elastoplastic model.
Figure 5.
Water pressure versus time, nonlinear model.
Figure 8. Vertical displacement versus time, elastoplastic model.
Figure 6. Vertical displacement versus distance from the origin, elastoplastic model.
Figure 9. model.
Water pressure versus time, elastoplastic
time. As it was expected, the settlements are farther in the upper time. Furthermore, the negative water pressure is decreased by the time passing because of increasing the degree of saturation by watering. 5.2.2 Elastoplastic model Vertical displacements and water pressures in three points of time are shown in Figures 67,
respectively. Also, their variations during the time are respectively presented in Figures 89. The trend of variations of each parameter in the elastoplastic analysis is similar to the nonlinear one, which was discussed above. Although, the nodal settlements considering the elastoplastic behaviour in an unsaturated medium are obviously more than the nonlinear one (Figs. 6, 8).
1072
CONCLUSIONS
The main aim of this study was to evaluate the thermoelastoplastic model, which has been incorporated into the code -Stock by the authors, and to compare its results with the nonlinear elastic model. This study has used a fully coupled thermo-hydro-mechanical formulation that has been integrated into the code -Stock. An excellent agreement with the available analytical analyses is observed. Analyzing the unsaturated porous media using the elastoplastic model shows a more factual behaviour of unsaturated soils than the nonlinear elastic model. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430.
Gatmiri, B. 1997. Analysis of fully coupled behaviour of unsaturated porous media under stress, suction and temperature gradient. Final report. Paris: CERMES-EDF. Gatmiri, B. 2000. Thermo-hydro-Mcanique des sols saturs et non saturs dans le Code_Aster. Rapport final. Paris: CERMES-EDF-MMN-CIH. Gatmiri, B. & Arson, C. 2008. -Stock, a Powerful tool of thermohydromechanical behaviour and damage modelling of unsaturated porous media. Computers and Geotechnics 35(6): 890915. Jenab, B. 2000. Etude numrical de la modlisation thermo-lasto-plastique des sols non saturs. PHD Dissertation. Paris: Ecole Nationale des Ponts et Chausses. Philip, J.R. & de Vries, D.A. 1957. Moisture movement in porous materials under temperature gradients. Trans Am Geophys Un 38: 222232. Salehnia, F. 2010. Analysis of the soil media using the theory of unsaturated soils with nonlinear behaviour. M.Sc Thesis. Tehran: University of Tehran.
1073
Explicit stress integration with reduced drift for Barcelona Basic Model
W.T. Soowski, D. Sheng & S.W. Sloan
ABSTRACT: The paper studies the effectiveness of a drift reduction method in integrating unsaturated soil models. The drift reduction is based on the NICE (Next Increment Corrects Error) method. The NICE method can be applied to any explicit stress integration method based on Runge-Kutta method. The performance of the stress integration enhanced by the NICE scheme is demonstrated with the Barcelona Basic Model. Numerical studies in this paper involve the integration of the Barcelona Basic Model with different numbers of strain increments and different sets of model parameters and initial conditions. The improved algorithm does not require any significant computations, thus the drift reduction is virtually accomplished without significant cost. The reduction in the yield surface drift appears to be rather significant, suggesting that under normal circumstances no additional drift correction method is required. 1 INTRODUCTION Sheng et al. 2003a, b, Sheng et al. 2008c, Pedroso et al. 2008), explicit methods are generally favoured over implicit methods. However, in all the above studies, very little attention was paid to the yield surface drift. In addition, the drift correction method is likely to affect the convergence of the global iterations and this key aspect has largely been ignored. Very recently, Vrh et al. (2010) proposed a method to reduce the yield surface drift during integration of classical elastoplastic models. In this paper, this drift reduction method is incorporated into the explicit integration method by Soowski & Gallipoli (2010a, b) and is tested for its effectiveness in integrating unsaturated soil models. 1.1 Explicit stress integration with adaptive subincrementation and error control Explicit stress integration schemes usually do not integrate stresses in one step. Instead, the strain increment is divided into subincrements (substeps) which are subsequently integrated (Sloan 1987, Sloan et al. 2001, Luccioni et al. 2001). The subincrementation is usually done in conjunction with an error control mechanism. The error measure is usually based on the numerical solutions to the differential stress-strain equations of different orders of accuracy. Such explicit algorithms are referred to as algorithms with adaptive substepping and error control. Generally, these algorithms use Runge-Kutta methods of different orders of accuracy to integrate the stress-strain relations. The stress integration algorithm used in this paper is based on the algorithm recently described in Soowski & Gallipoli (2010a, b) amended with the NICE (Next Increment Corrects Error) scheme
In finite element analysis, stresses are usually integrated from given strain increments. This process is referred to as stress integration or stress point method. Stress integration, unfortunately, cannot be achieved analytically for most elastoplastic constitutive models for soils. Instead, an approximate numerical solution is required. Stress integration schemes can be broadly classified into explicit and implicit. In implicit schemes, the gradients of yield function and plastic potential are estimated at advanced stress states and the stress-strain equation is treated as a set of nonlinear equations and solved by iteration. In explicit stress schemes, the gradients are estimated at known stress states and the stress-strain equation is treated as a set of differential equations and solved incrementally. At the end of integration, the stress points may not satisfy the yield function due to numerical errors either from iteration or from incrementation. This is known as yield surface drift. It should be stressed that both implicit and explicit methods can result in yield surface drift, even though the problem is more commonly associated with the latter methods. To correct this drift, a drift correction scheme is usually applied at the end of stress integration. Integration of unsaturated soil models has been discussed by Vaunat et al. (2000), Zhang et al. (2001), Borja (2004), Hoyos & Arduino (2008), Zhang & Zhou, Tamagnini & De Gennaro (2008) using implicit methods, and by Sheng et al. (2003a, b), Sheng et al. (2004), Sheng et al. (2008b, 2008c), Sanchez et al. (2008), Soowski & Gallipoli (2010a, b) using explicit methods. Due to the possible non-convex yield surface in unsaturated soil models (Wheeler et al. 2002,
1075
by Vrh et al. (2010) which reduces drift during the integration. 1.2 Drift correction procedures
proposed by Vrh et al. (2010) and extends the NICE method to higher order explicit stress integration schemes. 1.3 Barcelona Basic Model
After integrating the given strain increment, a drift correction algorithm is usually called if the yield function |F| 0 is not satisfied to a prescribed tolerance (YT). The tolerance YT may be defined as a small constant of the usual yield function, or, in a better way, be specified as a dimensionless quantity using the normalised yield locus size (see Sheng et al. 2000, Soowski & Gallipoli 2010a). A simple method to reduce the drift is to increase the accuracy of the explicit stress integration scheme. This option, however, may be rather computationally expensive when drift occurs only occasionally. When drift correction is necessary in most increments, it usually means that the YT value is too low for the chosen accuracy of the algorithm. The standard drift correction algorithm is described by Potts & Gens (1985). This method is not always cheap computationally and may diminish the convergence of the global solution algorithm. Therefore the paper proposes an improved explicit stress integration scheme with reduced drift. The improved algorithm does not require any significant additional computations, thus the time of stress integration remains the same. The proposed scheme is built on the ideas
Table 1. Parameter Specific volume for zero suction at reference pressure Shear modulus Elastic Stiffness parameter (elastic swelling index) Slope of NCL* at zero suction Critical state line slope Reference pressure Atmospheric pressure Swelling index for suction Parameter controlling cohesion increase with suction Parameter controlling ratio of NCL* slopes at s and s 0 Parameter controlling variation of NCL* slope with suction * NCLNormal Compression Line. Barcelona Basic Model parameters. Symbol N(0) G [MPa] (0) M pc [kPa] patm [kPa] s k r [kPa]1
As an example, the Barcelona Basic Model (BBM) proposed by Alonso et al. (1990) is used to test the stress integration scheme. Other unsaturated soil models could be used (Wheeler et al. 2003, Sheng et al. 2004, 2008a), with only minor modifications to the scheme. To avoid the transition between the Modified Cam Clay model for saturated states and the BBM for unsaturated states, it is assumed that the suction remains non-negative during the integration. The transition can be dealt with by either reformulating the BBM in the effective stress space (Sheng et al. 2003a, b) or using a continuous model such as Sheng et al. (2008c). It should be noted that the transition in the BBM does not constitute any problem to the stress integration itself, but rather a discontinuity in the constitutive equation (the constitutive matrix Dep becomes undefined at the transition suction in the BBM). The BBM parameters are given in Table 1. In the displacement finite element method, the suction is solved simultaneously with the displacements, and hence the suction increment is considered known in the stress integration algorithm.
Compacted Kaolin 1.9474 3.3 0.015 0.14 0.82 43 100 0.01 1.24 0.26 0.0164
Barcelona Sandy Silt 1.89 3.0 0.01068 0.0656 1.003 1 100 0.001 0.6 0.659 0.00592
Lower Cromer Till 1.68523 7.0 0.0077 0.066 1.2 12 100 0.001 0.8 0.25 0.02
1076
STRESS INTEGRATION FOR THE BBM WITH NICE DRIFT CORRECTION
The pre-requisite for explicit stress integration is the elastoplastic tangent matrix Dep. This matrix in the improved scheme is derived using a new consistency condition. The standard consistency condition dF 0 is altered and becomes F(X0) dF 0 (2) where F(x0) indicates value of the yield function at the beginning of the substep and dF is the total differential of this function. This modified condition can be regarded as the Taylor series expansion of F around the initial point x0 with truncated higher order terms. In particular, for the BBM condition (2) reads
T F * F F 0 F d ds * dp0 s p0
It is noted that the constitutive matrices (Dep and Wep) remain unchanged, but an additional term due to F(x0) enters into the stress-strain equation. * is The hardening parameter increment dp0
* dp0 * p0
(1)
v
d p v p
* p0
v
g
* gF ( x ) ga T De d g c a T De b ds p0 0 T e p a D g d v
(6)
(3)
* where d, ds and dp0 are increments of stress, suction and hardening parameter respectively and the value of yield function F is computed at the beginning of the substep. After a standard derivation (see Soowski & Gallipoli 2010a) the obtained plastic multiplier has an additional term F(x0) in the numerator T e F ( x 0 ) a T Ded c a D b ds
Thus both the stresses and the hardening variable are modified in the scheme, yet the correction to these quantities is independent of the future substep size. As the drift correction step is independent of the strain increment, the use of the NICE scheme has no influence on the stress integration method and therefore can be incorporated into a Runge-Kutta method of any order or into a method based on Richardson extrapolation. When the NICE method is implemented in a Runge-Kutta scheme, it must be noted that F(x0) corresponds to the value of the yield function before the Runge-Kutta substep and this value is used in each stage of the Runge-Kutta method. 3 DRIFT ASSESSMENT IN THE ALGORITHM ENHANCED WITH NICE
a T Deg d
(4)
The yield surface drift is here assessed with respect to the normalised yield locus for the Barcelona Basic Model as F F/(p0 ks)2 (7)
where a
s F F 1 ,b m ,c , s 3 pat s F p* T g g m ,g d * 0 p p0 v
In the above, the superscript T denotes a transposed matrix, g is the plastic potential function, p represents the plasm is a vector [1,1,1,0,0,0]T, v tic volumetric strain and v the denotes specific volume. The stress-strain equation becomes d DegF ( x ) D e (a T D e g d ) D e g a T D e d T e 0 T e a D gd a D gd e T e e T e D g c a D b D b (a D g d ) ds a T De g d e D gF ( x ) Depd T e 0 W epds (5) a D gd
The yield locus F is independent of the current hardening parameter and suction, and thus allows for the description of the drift independently of the stress state (Soowski & Gallipoli 2010a). In the remaining part of the paper the drift in each increment is defined as the absolute value of the normalised yield locus |F |. To check the performance of the NICE, a study where three sets of model parameters and 9 initial points were considered (see Tables 1 and 2). Starting from each initial stress state, 10201 strain increments were integrated. The strain increments were chosen in the [05%] range for both shear and volumetric strain such that a 101 101 grid of evenly spaced points is created. Each strain increment was coupled with a constant suction increment. Stresses were integrated using the Modified Euler method with adaptive substepping and error control (see Sloan 1987, Soowski & Gallipoli 2010a). This study has been performed with and without initial drift correction. The initial drift is a result of inexact satisfaction of the yield function by the elastic pro-
1077
cedure and is meant to approximate the usual starting condition for a stress integration algorithm. The results were obtained for the required algorithmic accuracy of 1% with error control as proposed by Sloan (1987) and Sloan et al. (2001). For this accuracy, the value of the relative YT of 0.2% is reasonable which corresponds to a drift of 2 103. A drift not exceeding this value was used for the analysis with initial drift. The typical drift |F | when a relatively large stress increment is integrated with and without NICE is depicted in Fig. 1. It is clear that the integration enhanced with the NICE method significantly reduces the drift over the majority of the increment, though the enhanced method does not guarantee that the drift will be always lower than in the case of no drift correction. The results of the study are presented in Tables 3 and 4. The enhancement with the NICE scheme generally reduces the average drift by around 90%; the maximum values of the drift are also reduced.
Table 2. Initial stress state and preconsolidation stress. CK* 55 90.3 100 12 88 0 70 48 50 66 BSS* 54.94 828.2 800 200 820 0 650 440 500 560 LCT* 20.9 21.9 5 5 20 0 16 9 12 12
It can be seen that the NICE scheme performs well and generally no additional drift correction algorithm is required. It can also be seen that the initial drift seriously influences the future drift when no drift correction is used, but has virtually no influence on the final drift in the improved integration algorithm. Thus, one need not be worried about drift being out of control in an analysis when no additional drift correction algorithm is used. The spatial variation of the relative drift with the strain increments is given in Fig. 2 and Fig. 3. It can be clearly seen that the drift in the case of the improved algorithm does not increase and is relatively independent of the strain increments used in the integration. Finally, it has been checked how the required accuracy of computations influences the drift. To achieve that, computations for Compacted Kaolin, with a stress state corresponding to point 2, were repeated with varying accuracy. Results are presented in Table 5.
1.40E-04 1.20E-04 1.00E-04 8.00E-05
Initial value of a parameter Hardening parameter Preconsolidation stress Suction Suction increment Point 1, mean net stress Point 1, shear stress Point 2, mean net stress Point 2, shear stress Point 3, mean net stress Point 3, shear stress p* 0 [kPa] p0 [kPa] s [kPa] ds [kPa] p [kPa] q [kPa] p [kPa] q [kPa] p [kPa] q [kPa]
Relative drift
6.00E-05 4.00E-05 2.00E-05 0.00E+00 0 -2.00E-05 -4.00E-05 -6.00E-05 0.2 0.4 0.6 0.8 1
Relative step size No drift correction NICE
* CKCompacted Kaolin, BSSBarcelona Sandy Silt, LCTLower Cromer Till.
Figure 1. Comparison of drift (F) upon integration of stresses. Volumetric and shear strain increments: 5%, integration accuracy: 1%. Initial point 3, model parameters as for Barcelona Sandy Silt. Note that the last substep of the integration is very small (the leftover of strain increment).
Table 3. Comparison of average and maximum drift for Modified Euler scheme without any drift correction and with reduced drift with NICE approach. Absolute values of drift |F| used in calculations. No initial drift was allowed. No initial drift Compacted Kaolin Barcelona Sandy Silt Lower Cromer Till Point 1 2 3 1 2 3 1 2 3 Average drift [106] 2.93 2.83 2.52 70.8 51.6 37.3 21.6 17.3 11.6 Maximum drift [106] 48.5 39.5 25.0 378.2 324.0 252.7 5128 6926 2162 NICE average drift [106] 0.22 0.29 0.27 15.0 10.0 6.48 2.10 1.87 1.32 NICE maximum drift [106] 36.5 25.2 15.1 242.7 213.4 140.5 5128 6926 2162 Average improvement [%] 92.5 89.8 89.4 78.8 80.6 82.6 90.3 89.2 88.6
1078
Table 4. Comparison of average and maximum drift for Modified Euler scheme without drift correction and with reduced drift with NICE approach. Absolute values of drift |F| used in calculations. Initial drift not exceeding 0.2% was allowed. Initial drift allowed Compacted Kaolin Barcelona Sandy Silt Lower Cromer Till P1 P2 P3 P1 P2 P3 P1 P2 P3 Average drift [106] 48.6 96.1 36.5 104.5 85.9 73.8 36.3 32.7 26.7 Maximum drift [106] 229.4 501.5 219.1 412.5 352.8 280.5 5128 6926 2162 NICE average drift [106] 0.22 0.29 0.27 15.0 10.0 6.50 2.14 1.88 1.31 NICE maximum drift [106] 36.7 25.3 15.1 243.2 214.0 142.5 5128 6926 2161.8 Average improvement [%] 99.6 99.7 99.3 85.6 88.4 91.2 94.1 94.2 95.1
Figure 2. Absolute relative drift |F| when drift reduction is used. Calculations for LCT, point 1, no initial drift allowed.
Figure 3. Absolute relative drift |F| when no drift correction is used. Calculations for LCT, point 1, no initial drift allowed.
Table 5. Comparison of average and maximum drift for Modified Euler scheme without drift correction and with reduced drift by NICE approach for different integration accuracies. Absolute values of drift |F| was used in calculations. No initial drift was allowed. No initial drift Compacted Kaolin Point 2 Accuracy 1E-2 1E-3 1E-4 1E-5 Average drift [106] 2.83 0.34 0.037 0.0039 Maximum drift [106] 39.5 6.25 0.67 0.07 NICE average drift [106] 0.29 0.013 4.4E-4 1.5E-5 NICE maximum drift [106] 25.2 1.51 0.06 1.7E-3 Average improvement [%] 89.8 96.3 98.8 99.6
Table 5 shows that the efficiency of the drift reduction in the improved algorithm increases with increasing accuracy. This is expected, as the drift in the last substep is not corrected. Thus, the higher required accuracy is, the smaller the final substep
size is, and hence the smaller drift in this substep is. The data suggests that increasing the accuracy tolerance is a very efficient way of reducing drift in the improved scheme. This reduction in drift is significantly larger than in a standard algorithm.
1079
CONCLUSIONS
The paper presents an explicit stress integration algorithm enhanced with a drift reduction method. The drift reduction is obtained with the NICE scheme proposed by Vrh et al. (2010) and can be easily incorporated into any explicit stress integration scheme based on the Runge-Kutta method. The yield surface drift in the proposed scheme is reduced significantly and generally no additional drift correction algorithm is needed. Additionally, the reduction in drift is higher when the required accuracy of results is higher, which makes the increase of accuracy of an algorithm a very efficient way of reducing drift in calculations. The proposed amendment to the explicit stress integration scheme requires virtually no additional computations and thus it is highly advisable for implementation in explicit stress integration algorithms. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40(3):405430. Borja, R.I. 2004. Cam clay plasticity, Part V: A mathematical framework for threephase deformation and strain localization analysis of partially saturated porous media. Computer Methods in Applied Mechanics and Engineering, 193:53015338. Hoyos, L.R. & Arduino, P. 2008. Implicit algorithms in modeling unsaturated soil response in three-invariant stress space. International Journal of Geomechanics, 8:266273. Luccioni, L.X., Pestana, J.M. & Taylor, R.L. 2001. Finite element implementation of non-linear elastoplastic constitutive laws using local and global explicit algorithms with automatic error control. International Journal for Numerical Methods in Engineering, (50):11911212. Pedroso, D.M., Sheng, D. & Sloan, S.W. 2008. Stress update algorithm for elastoplastic models with non-convex yield surfaces. International Journal for Numerical Methods in Engineering, 76:20292062 (doi:10.1002/nme.2407). Potts, D.M. & Gens, A. 1985. A critical assessment of methods of correcting for drift from the yield surface in elasto-plastic finite elements analysis. International Journal Numerical and Analytical Methods in Geomechanics, 9:149159. Sanchez, M., Gens, A., Guimaraes, L. & Olivella, S. 2008. Implementation algorihm of a generalised plasticity model for swelling clay. Computers and Geotechnics, 35:860871. Sheng, D., Fredlund, D.G. & Gens, A. 2008a. A new modelling approach for unsaturated soils using independent stress variables. Canadian Geotechnical Journal, 45(4):511534. Sheng, D., Gens, A., Fredlund, D. & Sloan, S.W. 2008b. Unsaturated soils: From constitutive modelling to numerical algorithms. Computers and Geotechnics, 35: 810824. Sheng, D., Pedroso, D.M. & Abbo, A.J. 2008c. Stress path dependency and non-convexity of unsaturated soil models. Computational Mechanics, 42(5):685695.
Sheng, D., Sloan, S.W. & Gens, A. 2004. A constitutive model for unsaturated soils: Thermomechanical and computational aspects. Computational Mechanics, 33(6):453465. Sheng, D., Sloan, S.W., Gens, A. & Smith, D.W. 2003a. Finite element formulation and algorithms for unsaturated soils. Part I: Theory. International Journal for Numerical and Analytical Methods in Geomechanics, 27(9):745765. Sheng, D., Sloan, S.W. & Yu, H.S. 2000. Aspects of finite element implementation of critical state models. Computational Mechanics, 26:185196. Sheng, D., Smith, D.W., Sloan, S.W. & Gens, A. 2003b. Finite element formulation and algorithms for unsaturated soils. Part II: Verification and application. International Journal for Numerical and Analytical Methods in Geomechanics, 27(9):767790. Sloan, S.W. 1987. Substepping schemes for the numerical integration of elastoplastic stress-strain relations. International Journal for Numerical Methods in Engineering, 24:893911. Sloan, S.W., Abbo, A.J. & Sheng, D. 2001. Refined explicit integration of elastoplastic models with automatic error control. Engineering Computations,18(1/2): 121154. Erratum: Engineering Computations, 19(5/6): 594594, 2002. Soowski, W.T & Gallipoli, D. 2010a. Explicit stress integration with error control for the Barcelona Basic Model. Part I: Algorithms formulation. Computers and Geotechnics, 37:5967, doi:10.1016/j.compgeo.2009.07.004. Soowski, W.T. & Gallipoli, D. 2010b. Explicit stress integration with error control for the Barcelona Basic Model. Part II: Algorithms efficiency and accuracy. Computers and Geotechnics, 37:6881, doi:10.1016/j. compgeo.2009.07.003. Tamagnini, R. & De Gennaro, V. 2008. Implicit integration of extended Cam-clay model for unsaturated soils. Unsaturated Soils: Advances in Geo-Engineering Toll et al. (eds). Proceeding of the First European Converence on Unsaturated Soils. Durham, UK. Taylor & Francis, 713719. Vaunat, J., Cante, J.C., Ledesma, A. & Gens, A. 2000. A stress point algorithm for an elastoplastic model in unsaturated soils. International Journal of Plasticity, 16:121141. Vrh, M., Halilovi, M. & tok, B. 2010. Improved explicit integration in plasticity. International Journal for Numerical Methods in Engineering 81(7):910938 doi: 10.1002/nme.2737. Wheeler, S.J., Gallipoli D. & Karstunnen M. 2002. Comments on use of the Barcelona Basic Model for unsaturated soils. International Journal Numerical and Analytical Methods in Geomechanics, 26:15611571. doi: 10.1002/nag.259. Wheeler, S.J., Sharma, R.S. & Buisson, M.S.R. 2003. Coupling of hydraulic hysteresis and stressstrain behaviour in unsaturated soils. Gotechnique, 53(1):4154. Zhang, H.W., Heeres, O.M., de Borst, R. & Schrefler, B.A. 2001. Implicit integration of a generalized plasticity constitutive model for partially saturated soil. Engineering Computations, 18(1/2):314336. Zhang, H.W. & Zhou, L. 2008. Implicit integration of chemo-plastic constitutive mode for partially saturated soils. International Journal for Numerical and Analytical Methods in Geomechanics, 32:17151735.
1080
Active earth force in cohesionless unsaturated soils using bound theorems of plasticity
Samuel A. Stanier
Department of Civil Engineering, University of Sheffield, UK
Alessandro Tarantino
Department of Civil Engineering, University of Strathclyde, UK (formerly Dipartimento di Ingegneria Meccanica e Strutturale, Universit degli Studi di Trento, Italy)
ABSTRACT: Granular cohesionless soils above the water table are partially saturated but are commonly assumed to be dry in geotechnical practice. The dry soil assumption neglects the effect of suction on shear strength and soil structures are therefore over-designed. This paper presents an approach to the analysis of collapse of soil structures above the water table based on the upper and lower bound theorems of plasticity. As an example, the active earth force on a retaining wall has been calculated and compared to the value obtained from the classical dry approach. 1 INTRODUCTION press), the classical approach developed for dry and saturated soils can be easily extended to partially saturated soils. As an example, the problem of the active lateral force on a retaining wall has been assessed. Using a water retention curve modelled by a singleparameter exponential function assuming hydrostatic conditions, analytical lower and upper bound solutions have been obtained for the active earth force. Finally the active earth forces considering a non-clayey (cohesionless) soil and different wall heights and water table depths has been calculated and compared with the values obtained using the classical dry approach. 2 THEOREMS OF PLASTIC COLLAPSE APPLIED TO SOILS ABOVE WATER TABLE
Non-clayey cohesionless soils above the water table are generally assumed to be dry in routine engineering calculations. Nonetheless this is rarely the case. Soils above the water table are in fact partially saturated and have shown to exhibit significantly higher shear strength than dry soils. Practitioners and academicians find it convenient to disregard the contribution of partial saturation to shear strength as this leads to conservative design. This point of view can be questioned. Significant costs might be saved if new geostructures are designed to account for the effects of partial saturation. On the other hand, geotechnical engineers are often confronted with existing hazardous geostructures, e.g. unstable slopes. In this case, a realistic analysis of the current state of stress, including the characterisation of the partially saturated zone above the water table, is essential to assess the causes of instability and, hence, to design appropriate remedial measures. To quantify the effects of partial saturation on the stability of geostructures, methods should be developed to analyse collapse conditions in cohesionless partially saturated soils. So far, this problem has received little attention from researchers working on unsaturated soils. This paper presents an approach based on the upper and lower bound theorems of plasticity. By assuming that the shear strength of partially saturated soils is controlled by the average skeleton stress (Oberg & Salfors 1997, Boso 2005, Tarantino 2007, Jotisankasa et al. 2009, Papa & Nicotera in
The upper and lower bound theorems of plastic collapse set limits to the collapse load of a structure and can be proved for the case of perfectly plastic materials. A perfectly plastic material is one which reaches a state of non-hardening failure where the failure criterion serves as a plastic potential and so the flow rule is associated. At ultimate failure the vector plastic strain increment is normal to the failure envelope and the forces and stresses remain constant for an increment of deformation; thus, since all forces and all stresses remain constant, elastic components of strain are zero and increments of total and plastic strain are identical.
1081
In two-phase soils, the failure envelope under ultimate conditions can be defined by the following equation: ( u) tan (1)
where is the shear stress, is the normal stress, u is the pore pressure and is the effective angle of shearing resistance. Pore pressure equals the porewater pressure uw in saturated soils and the poreair pressure ua in ideally dry soils. Using the shear strength criterion given by Eq. (1), the ultimate conditions of soil structures such as retaining walls, foundations, vertical cuts, and slopes can be calculated for saturated and dry soil (Atkinson, 1981). The application of bound theorems of plasticity to soil structures above the water table requires the definition of a suitable failure criterion for partially saturated soils. For compacted (aggregated) soils, shear strength under partially saturated states can be expressed by the following equation (Tarantino and Tombolato, 2005): ( uwSre) tan ( + sSre) tan (2)
On the other hand, reconstituted and non-clayey soils are generally non-aggregated and the microstructural water ratio ewm may therefore be expected to be zero for these soils. This appears to indeed be the case as shown by Boso (2005) for a reconstituted clayey silt and Papa & Nicotera (in press) for a pyroclastic silty sand. For non-aggregated soils, shear strength can therefore be defined by the following equation: ( uwSr) tan ( + sSr) tan (4)
If Eq. (2) or Eq. (4) are used in place Eq. (1), collapse of geostructures in partially saturated soils can be analysed in a very similar fashion as for saturated and dry soils. 3 RETAINING WALL
where uw is the pore-water pressure, s is the suction (s uw), and Sre is an effective degree of saturation. This is given by: Sre ew ewm e ewm (3)
Figure 1 schematically shows a gravity wall and a cantilever wall, which are the two most common types of retaining walls. If the water table is located below the wall, pore-water pressure will be negative along the wall interface (indicated by the dashed line in Figure 1). Negative values for the lateral stress x are compatible with the shear strength criterion given by Eq. (2) and, hence, tensile lateral total stress may be established at the soil-wall interface.
x z Sa H
(a)
where e is the void ratio (volume of voids per volume of solids), ew is the water ratio (volume of water per volume of solids), and ewm is the microstructural water ratio, which separates the region of inter-aggregate porosity from the region of intra-aggregate porosity (Romero & Vaunat, 2000). Eq. (3) characterises the degree of saturation of the macro-pores, i.e. the pores between the aggregates. If the soil is idealised as a granular material where grains are made of aggregates, the degree of saturation of the macro-pores plays the same role as the overall degree of saturation in granular non-materials. The parameter ewm may be difficult to determine directly, as there might not be a clear-cut distinction between the region of inter-aggregate porosity and the region of intra-aggregate porosity. As a result, the parameter ewm may conveniently be determined by best fitting of shear strength data. The validity of Eq. (2) using Eq. (3) has been proved by Tarantino (2007) for a wide range of compacted soils (clay, sandy clay, silty sand, and gravel) and more recently by Jotisankasa et al. (2009) for another compacted silty clay.
Water table
x z Sa Hw
Water table
(b)
Figure 1. Main types of retaining walls, gravity (a) and cantilever (b).
1082
For the case of the gravity wall shown in Figure 1a, however, negative lateral stresses cannot be transmitted to the backfill soil because the concrete wall is not able to pull the soil (there is no bonding between the soil and the concrete wall). As a result, when evaluating the active lateral force on the wall, it should be assumed x 0. On the other hand, negative (tensile) lateral stresses can arise at the soil-soil interface (dotted line) for the cantilever wall shown in Figure 1b. In this case, the suction stresses across the interface allow tensile total stresses to be transmitted from the soil-concrete wall to the backfill soil. For the sake of simplicity, the paper will only examine the active lateral force Sa for the case shown in Figure 1b. In addition, the case of a nonaggregated soil for which Eq. (4) applies will be considered. 4 ANALYTICAL SOLUTION FOR ACTIVE EARTH FORCE
' L W
z Sa H
Hw
Figure 2. Kinematic mechanism to derive the upper bound solution.
horizontal components respectively, of the block displacement. Since the unit weight of the soil is given by: dry ( sat dry) Sr (9)
where dry and sat are the dry and saturated unit weight respectively, the following relation can be derived for the weight W: dry H 2 ( sat dry ) 2 W tan a H w a w Hw e a w H 1 e (a w )2
An analytical solution for the upper and lower bound active thrust can be obtained if an exponential water retention function is considered: Sr = exp(as) (5)
(10)
where a is a soil parameter. Under hydrostatic conditions suction s is given by s uw w (z Hw) (6)
By combining Eq. (8) with Eq. (10), the external work Le can therefore be calculated. The internal energy dissipation Li is given by: Li ( ) dV
V
where w is the unit weight of water, z the vertical coordinate positive downward, and Hw is the water table depth (see Figure 1). By combining Eq. (5) and Eq. (6), the degree of saturation under hydrostatic conditions can therefore be expressed as follows: Sr e a w ( z Hw ) e a w Hw e a w z 4.1 Upper bound solution To derive the upper bound active thrust, firstly a kinematically admissible mechanism needs to be considered and secondly the external work and the internal energy dissipation need to be equated. The simplest kinematism consists of a single block with planar slip surface as shown in Figure 2. The external work Le is then given by: Le Wv Sah (8) (7)
(11)
where and are the stresses normal and tangential to the slip surface respectively, and and are the strains conjugate to the stresses and . It can be demonstrated that Li is given by: Li sin sin e a w Hw sSr dz w cos H cos (a w )2
aw H H (a w Hw 1) (e a w H 1) a w e
(12)
By equating Li with Le, the following equation is obtained for the active thrust: dry H 2 2 k Sa a H a w H 1 a a H e e ) ( dry sat (a w )2 1 a w ( H H w ) e a H H w (1 ka ) (a w )2 (1 a w Hw ) e a H
w w w w w w w
where W is the weight of the block, Sa is the active thrust, and v and h are the vertical and the
(13)
1083
4.2
Lower bound solution
To derive the lower bound active thrust, firstly the state of stress in equilibrium with the applied loads needs to be considered and secondly it must be ensured that the stresses do not violate the failure criterion. If the vertical and horizontal directions are assumed to be principal directions of stress, the equilibrium stress state is given by: z z 0 x x (14)
100 90 80 70 60 50 40 30 20 10 0 0.001 Percent finer
silt
sand
gravel
4 (N=12)
b)
0.01 0.1 1 10 Particle diameter: mm 100
where z and are the vertical and horizontal directions respectively, and x0 is a constant. A lower bound value for the horizontal stress x0 is obtained by imposing the Mohr stress circle in the plane ( sSr, ) that is tangent to the failure envelope:
0 x sSr ka z sSr
Figure 3. Grain size distribution of the pyroclastic silty sand (after Papa & Nicotera, in press).
(15)
1 0.8 0.6 0.4 0.2 0 0.1 1 10 Suction (kPa) 100

From experimental data Exponential model
where ka is the active earth coefficient. The lower bound active lateral force can finally be obtained by combining Eq. (15) with Eqs. (6), (7), (9) and integrating over the height H:
Sa
dry H ( sat dry ) 2 k 1 ka w a H 2 a w H 1 a a H e ( a w ) e 2 ( a w ) 1 a w ( H H w ) e a H H a H (16) (1 a w H w )e

2
w w w w w w w
x dz
0
Figure 4. Main drying water retention curve derived from experimental data and fitting using the exponential water retention function.
The lower and upper bound solution are found to be coincident when Eq. (6) and Eq. (9) are compared; thus the exact solution for this problem has been obtained. 5 EXAMPLES OF ACTIVE EARTH FORCE IN PYROCLASTIC SILTY SAND
To show the effect of partial saturation on lateral active force in non-clayey soils, we have examined the case of pyroclastic silty sand investigated by Papa & Nicotera (in press) and Nicotera et al. (submitted). The grain distribution shows it to have no clay fraction and silt fraction no greater than 50% (Figure 3). The water retention curve was determined in the range from 0 to 20 kPa on several natural samples retrieved from a deposit about 40 km northwest of the volcano Somma-Vesuvius (Naples, Italy). The
average water retention curve is shown in Figure 4 together with the exponential water retention function given by Eq. (5) used to fit the water retention curve, by calibrating the single parameter a. Physical properties are reported in Table 1 together with the critical state friction angle determined on the basis of 48 stress-path controlled triaxial compression tests performed on undisturbed specimens. The active thrust on a cantilever wall is shown in Figure 5 as a function of water table depth. Two wall heights were considered, 3 and 6 m respectively. The most striking aspect is negative active trust for the case of partially saturated soils in contrast to the positive thrust obtained if the soil is assumed to be dry. Under partially saturated conditions, it is the backfill soil which pulls the cantilever wall and not vice versa as usually assumed
1084
Table 1. Soil parameters. s (kN/m3) 25.7 w (kN/m3) 9.8 dry (kN/m3) 8.8 sat (kN/m3) 15.2 a kPa1 0.017 36.9 36.9
Depth, z (m)
Unsaturated soil (gravity wall)
Hw= 6m
Dry soil
20 H= 3m Active thrust, Sa (kN/m)
Dry soil
4
0
(a)
-20 Partially saturated soil
Unsaturated soil (cantilever wall)
6 -20 -10 0 10 20 Active lateral pressure, x0 (kPa))
-40 4 6 8 Water table depth, H w (m) 10
Figure 6. Lateral active pressure for the case of water table 3 m depth.
40 20 0 -20 -40 -60 6
H= 6m
Dry soil
Partially saturated soil
(b)
soil is assumed to be unsaturated, vertical slope can be cut to significant depths amounting to around 2/3 of the water table depth. It is also interesting to plot lateral pressure for the case of a gravity wall, represented by a dashed line in. In this case, boundary conditions at the soil-wall interface require x0 0. If the water table is located at the base of the wall, the active thrust is about one tenth of the thrust calculated under the assumption of dry soil. 6 CONCLUSIONS
10
Figure 5. Active thrust at various water table depths for cantilever walls of 3 m (a) and 6 m (b) height.
by the classical solution, where the effect of partial saturation is neglected. It is worth noticing that negative thrust is associated with negative (tensile) horizontal stresses. It is here assumed that tensile stresses can be transmitted across the interface between the backfill soil and the cantilever wall, i.e. no cracking occurs, provided the shear failure criterion is not violated. This assumption appears to be supported by recent experimental observations (Thusyanthan et al. 2007). It is interesting to plot the horizontal stress x0 derived from the lower bound solution given by Eq. (15) as shown in Figure 6. Lateral stress is always positive in dry soil and no vertical slope can therefore be cut in ideally dry soil. However, if the
The paper has presented a simple approach to assess the active thrust in soils above the water table using the bound theorems of plasticity. By considering an exponential water retention function and under the assumption of hydrostatic conditions, an analytical solution can be derived which can be used for preliminary assessment of the effect of partial saturation on lateral active forces. Results based on a non-clayey (cohesionless) pyroclastic soil suggests that assuming the soil to be dry above the water leads to significant overestimate of the active lateral force. ACKNOWLEDGMENTS The authors wish to thank Dr. Papa and Dr. Nicotera from the University of Naples Federico II, Italy, for providing data on the pyroclastic soil.
1085
REFERENCES
Atkinson, J.H. 1981. Foundations and slopes: an introduction to applications of critical state soil mechanics. London: McGraw-Hill. Boso, M. 2005. Shear strength behaviour of a reconstituted partially saturated clayey silt. PhD dissertation, Universit degli Studi di Trento, Italy. Jotisankasa, A., Coop, M. & Ridley, A. 2009. The mechanical behaviour of an unsaturated compacted silty clay. Gotechnique 59(5): 415428. Nicotera, M.V., Papa, R. & Urciuoli G. (submitted). An experimental technique for determining the hydraulic properties of unsaturated pyroclastic soils. Oberg, A and Salfors, G. 1997. Determination of shear strength parameters of unsaturated silts and sands based on the water retention curve. Geotechnical Testing Journal, 20: 4048.
Papa, R. & Nicotera M.V. (in press). Critical state of an unsaturated pyroclastic soil. Between Theory and Practice in Unsaturated Soil Mechanics, A. Tarantino & C. Mancuso (eds). Rotterdam: Millpress. Romero, E. & Vaunat, J. 2000. Retention curves in deformable clays. In Experimental Evidence and Theoretical Approaches in Unsaturated Soils, A. Tarantino and C. Mancuso (eds), pp. 91106, Rotterdam, A.A. Balkema. Tarantino, A. 2007. A possible critical state framework for unsaturated compacted soils. Gotechnique, 57(4): 385389. Tarantino, A. & Tombolato, S. 2005. Coupling of hydraulic and mechanical behaviour in unsaturated compacted clay. Gotechnique, 55(4): 307317. Thusyanthan, N.I. Take, W.A. Madabhushi, S.P.G. Bolton, M.D. 2007. Crack initiation in clay observed in beam bending. Gotechnique, 57(7): 581594.
1086
Implicit integration of a constitutive model for partly saturated structured porous materials
R. Tamagnini, M. Mavroulidou & M.J. Gunn
Faculty of Engineering, Science and the Built Environment, London South Bank University, London, UK
ABSTRACT: The paper concerns the integration of a rate equations system for partially saturated soils that have developed intergranular structure due to their geological history or chemical treatment. The constitutive equation adopts an effective stress concept and the hardening depends on the degree of saturation Sr. the bonding exerted by the structure is modelled accounting for the dependency of the hardening law on the cementing material mass. The implicit integration is performed adopting a backward Euler scheme and splitting the problem into two stages: An elastic predictor stage and a plastic corrector updating. In the elastic predictor stage the dimension of the elastic domain can vary accounting for the dependency of the preconsolidation pressure on Sr and debonding. This mathematical feature is a remarkable difference with respect to the original scheme for the saturated unbonded material in which the preconsolidation remains unchanged in the elastic stage. The plastic corrector stage updates the effective stress using the consistency condition. In the plastic corrector stage the non-linear system is integrated solving simultaneously for the stress invariants, the plastic multiplier and the internal variable. The paper reports the validation of the proposed algorithm by some simulations of data in literature; the numerical tests include destructuration upon loading, collapse due to water content change and the effect of the structure. 1 INTRODUCTION the companion paper submitted to this conference (Tamagnini et al. 2010). 2 IMPLEMENTATION
Tamagnini proposed an extended modified Cam clay model for unsaturated soils (Tamagnini 2000 and 2004). In Tamagnini (2000) the differential constitutive equations have been integrated according to Borja & Lee (1990) with an implicit return-mapping algorithm that follows the work of Simo and Hughes (1998). Lies equation (Simo & Hughes 1998) was introduced to integrate the double dependency of the hardening on both the volumetric plastic strains and the rate of the degree of saturation. In this paper the integration of the constitutive equation is addressed in a different way. The differential equation system is solved with respect to the four unknowns simultaneously. The unknowns are respectively the two invariants of the effective stress tensor, the preconsolidation pressure and plastic multiplier. The integration is performed adopting the fully implicit NewtonRaphson technique. The numerical algorithm is developed in the framework of a research on lime treated unsaturated soils and the free energy of the chemical bonding is accounted for in the hardening. The model formulation is reported in
The modified Cam clay is integrated enhancing the return mapping scheme proposed Simo & Hughes (1998). The improvement is based on two points: the modification of the elastic trial step that describes the variation of the effective Bishops stress and the effects of mechanical bonding and its degradation. At time tn [0;T ] of the time domain the following state variables (i.e. respectively: the total strain ijn, the plastic strain pijn the degree of saturation Srn and the amount of bonding mass mn) are known:
p ij n ; ij n ; Srn ; mn
(1)
As the WRC is the evolution law for degree of saturation there is another evolution law controlling the amount of bonding mass. The initial stress n and the elastic strain tensor are known
1087
(in the following the superscript indicating the effective stress will be omitted). The integration algorithm updates these variables at the time tn1 [0;T ]. The unknowns are defined as:
p ij Dijkl kl kl p f ij ij c pc vp, Sr , m p
(2)
In Equation (2) n is respectively the Bishop stress in unsaturated conditions and the Terzaghi effective stress in the fully saturated conditions but its rate includes the rate of the Gibbs potential, ij (u) (see Tamagnini et al. 2010), where u the vector of the displacement rate that is computed during the linearisation of the global iterative problem; the plastic multiplier; Sr the rate of the degree of saturation; and m the rate of the bonding solid cementing the granular media. The changes in the degree of saturation are computed with an evolution law controlling the energy of the interfaces and the fluids pressure (the water retention curve). The rate of the preconsolidation pressure pc controls the dimension of the elastic domain, that is the closure of the following domain in the effective stress space: E ;Sr ,m : ij ; h S R n f ( ij ; pc ) 0
condition imposed by equation (4), if the trial stress belongs to the elastic domain the stress is updated and the stiffness matrix is the elastic tangent matrix. Conversely if the trial elastic stress drives the current stress outside or on the plastic yield surface, the trial step is followed by the plastic corrector stage. The key issue for the extended hardening rule with dependency on the degree of saturation is the variation of the boundary of the elastic domain during the elastic predictor stage. This variation is driven by the increment Sr n1 and m n1 . This feature of the algorithm implies that unloading due to saturation or destructuration could be compatible trial p with the condition f n 1 ; pc v n ; Sr n 1, mn 1 0 triggering a plastic corrector stage that models the plastic compression (collapse). At time tn, in the equilibrium condition at the general iteration k, the increments:
k ij ; Sr k n 1 n 1
(7)
are known, and the total strain tensor can model the variation of mass of the bonding material (see Tamagnini et al. 2010). Then, the trial state variables can be written as:
ij trial ij D e kl k n 1 n ijkl n n 1 trial k k k pc n1 pcn exp bSr n1 exp v v s trial trial trial trial f pn f 1 ; qn 1 ; pc n1 n1
(3)

(8)
where h the vector of the internal variables. System (2) has to satisfy the Kuhn-Tucker condition: 0 f ( ; p ) 0 f ( ij ; pc ) 0 ij c and the initial condition: (4)
b is the sensitivity parameter for changes in the degree of saturation and is the sensitivity parameter for the destructuration The hardening rule is integrated in closed form and split adopting Lies formula (see Simo & Hughes 1998); the trial invariants are: 1 trial trial pn 1 ii n1 3
trial qn 1
p ij ; ij ; pc
tn
p ij n ; ij ; pc n n
(5) (6)
3 ij 2
trial n 1
(9)
ij
tn
p D e ijkl kl n kl n
At this time the check for plasticity is applied: if :

trial trial trial f pn 1 ; qn 1 ; pc n1 0
For the solution the backward Euler scheme is applied to the system of equations. The problem is split into two parts: the first part is a trial elastic step that is called elastic predictor in which the deformation is totally elastic. In this stage the dimension of the elastic domain of equation (3) changes due to the changes of the degree of saturation and the bonding mass consumption. This is a remarkable difference with respect to the original return mapping for saturated and unbonded soils (see for example Borja & Lee 1990 or Rouainia & Muir Wood 2000). After the check of the discrete
k then : ()trial n 1 ()n 1 EXIT else : Plastic corrector
The plastic corrector uses the trial stress and trial hardening as the initial condition and through a return mapping computes the stress and plastic strains. From the discrete Khun-Tucker condition: 0 f ( n1 ; pc n1 ) 0 n 1 (10)
The aim of the plastic corrector is the definition of the plastic multiplier at time tn1 and the
1088
subsequent computation of the stress through the following equations:

pk ij k ij trial D e kl n 1 n 1 ijkl n k pk trial pc n1 pc n exp( n kl ) f k f ( p kl ; q k ; p k ) n 1 n 1 c n1 n1
NUMERICAL TESTS ON IDEAL FULLY SATURATED SOILS
(11)
The non-linear equation system can be rewritten as:

k 2qn k trial k 1 qn 1 qn 1 3 n n 1 2 M k k k pc k pc n trial exp n 1 2 pn 1 pc n1 n 1 2 q fn 1 n 1 pn 1( pn 1 pc n1 ) 0 M k trial k k k pn 1 pn 1 K n n 1 2 pn 1 pc n1
(12)
The first tests are run to assess the integration algorithm; they are particular stress paths in which the closed form solution is known; namely an isotropic compression and an undrained shear test. The tests are performed adopting the following parameters: 0.12 0.01 M 1.3 0.3. Figure 1 shows the comparison between the closed form and the numerical integration for the drained isotropic compression The numerical algorithm is also tested on an undrained shear tests for normally consolidated and over consolidated condition. The comparison between the closed form solution and the numerical algorithm is reported in Figure 2. The accuracy for different number of steps for the normal consolidated shear (2, 10 and 100 steps respectively) is reported in Figure 3.
1.62
specific volume : v
This is a non-linear system of four equations with four unknowns: pn1, qn1, pcn1 and n1 . Note that Srn1 and ij are known at the start of the integration and the dependency of the stress (through the preconsolidation pressure) is accounted for in the elastic trial predictor stage. This feature is important in the consideration that follows. The system can be rewritten as: p p K n 2 p p k q k q trial 3 k 2 qn 1 n n 1 n 1 n 1 M2 A( x ) 0 (13) k trial k k k exp p p 2 p p c n1 c n n 1 n 1 c n1 q 2 n 1 pn 1 ( pn 1 pc n1 ) M
k n 1 trial n 1 k n 1 k n 1 k c n 1
1.6
1.58
1.56
1.54 Numerical integration Cloased form
1.52
1.5
1.48 0 400 800 1200 1600
mean effective stress: p' (KPa)
Figure 1.
500
Numerical test: isotropic compression.
in which:
k k k k xT pn 1, qn 1, pc n1 , n 1

1
(14)
400
Applying Newton-Raphson to system (13) we obtain:

deviator stress : q
A( x )i
A( x ) A( x ) dx 0 dx Ri x i x i
300 Closed form Numerical normal consolidated Numerical over consolidated 200
(15)
where Ri the residual. The problem is solved when the residual is smaller than a prescribed tolerance value. if: A(xi dxi) TOL then: xn1 xi dxi EXIT else: Perform another iteration The coefficients of the Jacobian are reported in the Appendix.
100
0 0 100 200 300 400 500
mean effective stress : p'
Figure 2.
Numerical test: undrained shear.
1089
400
deviator stress : q (KPa)
300
1.6
200
100 steps 20 steps 2 steps
specific volume : v
1.2
100
0.8
Exp. data (structured) Exp. data (remoulded)
0 0 100 200 300 400 500
mean effective stress : p' (KPa)
Figure 3.
Accuracy for different number of steps.
0.4 10 100 1000 10000
vertical stress : v'

0 0 100 200 300 400
Figure 5. Destructuration in oedometric condition, model results and experimental data after Callisto et al. (2002).
2
100
volumetric strains : dv (%)
4
80
60
Model (remoulded) Exp. data remoulded Model (structured) Exp. data (structured)
40
10
Exp. data Model
Figure 4. Simulation of the experimental data of the Pisa clay after Callisto et al. (2002).
20 0 5 10 15 20 25
deviatoric strain : s
NUMERICAL SIMULATION OF FULLY SATURATED NATURAL STRUCTURED MATERIAL: THE PISA CLAY
Figure 6. Destructuration in deviator compression, model prediction and data after Callisto et al. (2002).
The ongoing research provides the laboratory investigation of the lime treated London clay in fully saturated and partially saturated condition. At present the full set of data is not complete and the model formulation is validated against data found in the literature. The natural Pisa clay has been tested and the results are shown for example in Callisto et al. (2002). These results are simulated below, in order to validate the destructuration predicted by the suggested model. Figure 4 shows the difference between the natural soils and
the remoulded one. The figure reports also the model prediction in which the typical decrease of elastoplastic stiffness for the structured material is recorded. The results of the destructuration can be clearly seen in the simulation presented in Figure 5, in which debonding occurs in oedometric conditions. The destructuration in undrained shear testing is reported in Figure 6 (stress-strain behaviour) and Figure 7 (stress path). The adopted parameters
1090
100
CONCLUSIONS
80
60
40
20
Exp. data Model
0 0 20 40 60 80 100
The paper presented a new integration for the extended modified Cam Clay with an enhanced hardening law. The integration is performed with the implicit Newton-Raphson technique, modifying the trial stage of the return mapping. This implies that in destructuration, for example, the yielding surface moves during elastic trial loading triggering an over amount of plastic strains that simulates plastic relaxation and collapse, The integration was applied in the numerical simulation of destructuration in fully saturated structured clay and in the collapse upon wetting of a silty soil. The integration has been obtained in the framework of a research on lime-stabilised clays and it will be implemented in FE codes for the study of this type of soil. 7 APPENDIX
Figure 7. Stress path during destructuration, model prediction and data after Callisto et al. (2002).
The derivative of A(x) with respect to x can be expressed as a matrix: A( x ) Jlk x with the following components K J11 1 2 J12 J21 J23 J32 J 44 0.0 K J13 (15)
0.8
0.7
void ratio : e
0.6
0.5
0.4 1
mean net stress : v-ua
10
100
1000
J14 K (2 p pc ) J22 1 6 M2 J24 6 q M2
Figure 8. (2009).
Collapse on wetting after Jotisankasa et al.
are: 0.24 0.064 M 0.984 0.25 4.0 (In these preliminary tests, the effects of the specific volume on the new hardening are not taken into account). From the figures it can be seen that although it has a simple formulation, the model shows the ability to simulate this type of physical behaviour in fully saturated conditions. 5 NUMERICAL TEST ON COMPACTED PARTLY SATURATED SOILS
p exp J31 2 cn (2 p pc ) p exp J33 1 cn (2 p pc ) J34 pc n (2 p pc )exp (2 p pc ) J24 2 p pc J24 2q M2
The algorithm is validated in the simulations of compacted silty clay (Jotisankasa et al. 2009) during collapse on wetting. Figure 8 reports the simulation of the tests.
J34 p
1091
ACKNOWLEDGEMENTS This study is part of a larger research programme on the hydro-mechanical properties of lime-treated clays, funded by the UK Engineering and Physical Sciences Research Council (EPSRC) through grant EP/E037305/1. REFERENCES
Borja, R.I. & Lee, S.R. 1990. Cam-Clay plasticity, Part I: Implicit integration of elasto-plastic constitutive relations, Comp. Meth. Appl. Mech. Eng. 78: 4972. Callisto, L., Gajo, A. & Muir Wood, D. 2002. Simulation of triaxial and true triaxial tests on natural and reconstituted Pisa clay, Gotechnique 52 (9): 649666. Jotisankasa, A., Coop, M. & Ridley, A. 2009. The mechanical behaviour of an unsaturated silty clay. Gotechnique 59 (5): 415428.
Rouainia, M. & Muir Wood, D. 2000. An implicit algorithm for finite strain Cam clay elasto-plastic model, Mech. Cohes.-Frict. Mater. 5: 469489. Simo, J.C. & Hughes, T.J.R. 1998. Computational Inelasticity (Interdisciplinary Applied Mathematics) SpringerVerlag Berlin and Heidelberg GmbH & Co. K. Tamagnini, R. 2000. Modellazione dei terreni non saturi e implementazione agli elementi finiti. MSc Thesis, University of Rome La Sapienza (in Italian). Tamagnini, R. 2004. An extended Cam-clay model for unsaturated soils with hydraulic hysteresis. Gotechnique 54 (3): 22322. Tamagnini, R., Mavroulidou, M. & Gunn, M.J. 2010. A constitutive model for cemented unsaturated soils and weak rocks. In: 5th International Conference on Unsaturated Soils UNSAT2010, Barcelona, 68 September 2010.
1092
Numerical simulation of an unsaturated slope failure during an earthquake

R. Uzuoka, K. Maruyama, T. Mori & M. Kazama T. Unno
Department of Civil and Environmental Engineering, Tohoku University, Sendai, Japan Institute of Technology, Penta-Ocean Construction Co., Ltd., Nasu, Japan
N. Sento
Department of Civil Engineering, College of Engineering, Nihon University, Koriyama, Japan
ABSTRACT: A dynamic three-phase coupled analysis is validated through numerical simulations of seismic behavior of a damaged fill slope during an earthquake. The equations governing the dynamic deformation of unsaturated soil are derived here based on porous media theory and constitutive models. The skeleton stressstrain constitutive relation is a simplified elasto-plastic constitutive model based on the non-associated flow rule and nonlinear kinematic hardening rule. The soil water characteristic model is a hysteresis model to reproduce the wetting process during cyclic loading. An actual fill slope which was partially collapsed and flowed during the 2003 earthquake is examined. The physical and mechanical properties are investigated with in-situ and laboratory tests. The numerical results showed the reduction of skeleton stress in the unsaturated fill as well as liquefaction in the saturated fill during the earthquake. 1 INTRODUCTION with the dynamic three-phase coupled method proposed by authors (Uzuoka et al. 2009). The artificial fill slope is located on old valley at Tsukidate town in Miyagi prefecture. The fill slope was partially collapsed and flowed during the 2003 earthquake (Uzuoka et al. 2005). The fill material is a volcanic sandy soil. The physical and mechanical properties are investigated with in-situ and laboratory tests. The material parameters of the constitutive models are determined through the calibration of the laboratory tests. The numerical method is validated through the simulations of ground water flow at steady state and the seismic responses of the fill slope during the 2003 earthquake. The seismic behavior of unsaturated and saturated fill and the failure mechanism are discussed through the numerical simulations.
Fill slopes have often been damaged in past earthquakes. In most cases, a fill slope consists of unsaturated and saturated soil; therefore the seismic analysis should consider the dynamic behavior not only of saturated soil but of unsaturated soil. Most past seismic analyses, however, focused on the seismic behavior of saturated ground such as liquefaction. Liquefaction analyses of saturated ground based on Biots porous media theory (Biot 1962) have been studied since 1970s (e.g. Zienkiewicz & Shiomi 1984). Coupled analyses of unsaturated soil response have been presented in (Meroi & Schrefler 1995) using the degree of saturation as a prime variable, without determining the pore air pressure explicitly. However, the compressibility of the pore air plays an important role in the dynamic response of unsaturated soil (Okamura & Soga 2006, Unno et al. 2008), and hence should be considered in the formulation of boundary-value problems. Recently pore air pressure has been treated as a primary variable in (Schrefler & Scotta 2001, Ravichandran & Muraleetharan 2009); however, their numerical applications were not validated for an actual case. In this study numerical simulations of seismic behavior of an actual fill slope are performed
2 2.1
NUMERICAL METHOD Balance and constitutive equations
Firstly the basic equations (Uzuoka et al. 2009) are derived based on porous media theory (de Boer 2000, Schrefler 2002).
1093
Mass balance equation for phase ( s, w, a) is D div v 0 Dt (1)
Ds sw D s pc D s ( p a pw ) c c Dt Dt Dt
(7)
where D /Dt is the material time derivative with respect to phase, is the partial density of phase, v is the velocity vector of phase. The mass exchange among three phases is ignored here. The linear momentum balance equation of phase is
where c is the specific water capacity. The specific water capacity is calculated from the Soil Water Characteristic Curve (SWCC). The SWCC is assumed as
w se 1 exp(alg pc blg )
D v ^ a div b p Dt

s
w ( ss
w w sr )se
w sr
clg
, (8)
(2)
where a is the acceleration vector of phase, is the Cauchy stress tensor of phase, b is the body force vector, ^ p is the interaction vector of phase against other phases. Constitutive equations are the followings. The partial Cauchy stress of each phase is derived (e.g. Borja 2006) as s (1 n )( s w pw s a pa )I w ns w pw I, a ns a pa I (3)
where sws is the saturated (maximum) degree of saturation, swr is the residual (minimum) degree of saturation and swe is the effective water saturation. The relationship between swe and suction pc is assumed as a logistic function with the material parameters alg, blg and clg. The logistic SWCC is continuous function at pc 0; therefore the convergence in the iterative scheme can be achieved. The above SWCC is modified to fit a scanning curve during undrained cyclic shear in later. The permeability coefficient of water and air are assumed to be dependent on the effective water saturation as
w k w k k ws ksw ( se ) , k as ksa (1 se )
where is the skeleton stress tensor (e.g. Gallipoli et al. 2003), pw is the pore water pressure and pa is the pore air pressure. These pressures are defined as positive in compression. n is the porosity, sw is the degree of water saturation and sa is the degree of air saturation. The interaction vector for each phase is assumed as
^ ^ ^ p p p s w a
(9)
^ p grad n p
n R g s n v ks
(4)
where g is the gravity acceleration, n is the volume fractions, R is the real density, ks is the permeability coefficient of phase (w or a). The compressibility of pore water under an isothermal condition is assumed as D s wR wR D s pw w Dt Dt K (5)
where kws is the saturated (maximum) coefficient of water permeability, kas is the dry (maximum) coefficient of air permeability, k and k are the material parameters. Combining the balance and constitutive equations, we derive the governing equations which include the momentum balance equations of the overall three-phase material and the mass and momentum balance equations (continuity equations) of the pore water and air with the following assumptions. 1) The soil particle is incompressible, 2) the mass exchange among phases is neglected, 3) The material time derivative of relative velocities and advection terms of pore fluids to the soil skeleton are neglected, 4) an isothermal condition are assumed. The momentum balance equations of the overall three-phase material is derived as
a s div ( s w pw s a pa )I b
(10)
where Kw is the bulk modulus of pore water. The compressibility of pore air under an isothermal condition assumed as 1 D s pa D s aR R Dt Dt (6)
where is the overall density of three-phase material. The mass and momentum balance equations of the pore water and air are derived as
s w s a ns w wR wR D p wR D p n c n c Dt Dt Kw w wR s s div v k ws div grad pw wR b wRa s 0 g
where is the absolute temperature, R is the specific gas constant of air. The constitutive relation between water saturation and suction is assumed as
(11)
1094
s a s w ns a aR D p aR D p R n c Dt n c Dt s a aR div v s as k div grad pa aR b aRa s g
. 0 (12)
time derivative of displacement of soil skeleton .. as, pore water pressure p w and pore air pressure ..a p . The weak forms are linearized and solved by Newton-Raphson method iteratively at each time step. The linearized forms of the weak forms are derived as
s D w s [ a s ] D w s [ p ] D w s [ p ] w( k ) a .. ..w D w w [ a s ] D w w [ p ] D w w [ p ] w(w k) ..w .. a
This simplified formulation is called u-pw-pa formulation. Although the governing equations are derived in the regime of finite strain, we assume infinitesimal strain in the following study for simplicity. 2.2 Constitutive equation for skeleton stress
D w a [ a s ] D w a [ p ] D w a [ p ] w(ak ) (17) where w , w and w are the weak forms of the equations (10)(12) respectively, Dws[as] is directional derivative of ws with respect to as, the ws(k) is the residual at the iteration step of (k). The iteration is continued until the norm of the residual vectors becomes less than the convergence tolerance of 1.0 107. In the finite element formulation, Galerkin method and isoparametric 8-node elements are used. The soil skeleton displacement and the fluid pressures are approximated at 8 nodes and 4 nodes respectively to avoid a volumetric locking.
s w a
..w
.. a
A simplified elasto-plastic constitutive equation for skeleton stress is used here. Assuming that plastic deformation occurs only when the deviatoric stress ratio changes, the yield function is assumed as f k s / p k 0 (13)
where p is the mean skeleton stress, s is the deviatoric stress tensor, k is the material parameter which defines the elastic region. is the kinematic hardening parameter (back stress) and its nonlinear evolution rule (Armstrong & Frederick 1966) is assumed as . .p .p . a be p d , d ep
.
2.4 Implicit stress integration Implicit stress integration and consistent tangent modulus at infinitesimal strain (e.g. Simo & Taylor 1985) are used to achieve the convergence of global iteration of (17). The return mapping algorithm in stress space is used as g , r2 n , r3 f r1 ( tr ) c e
(14)
where a, b are the material parameters, e p is the plastic deviatoric strain rate tensor. With non-associated flow rule, the plastic potential function is assumed as g M m ln p / p a 0 (15)
(18)
where Mm is the material parameter which defines the critical state ratio, pa is p when 0. Finally stress-dependent elastic module are assumed as K e K * p G e G * p (16)
where Ke is the elastic bulk modulus, Ge is the elastic shear modulus, K* and G* are the dimensionless elastic module respectively. 2.3 Finite element formulation and time integration
where (tr) is the trial skeleton stress for a given strain increment at the global iteration step, is the plastic multiplier, ce is the elastic tensor, n is at the previous time step, is the increment of . Until the norm of left-handed residual vector r of (18) becomes less than the convergence tolerance of 1.0 1010, the nonlinear equations are solved iteratively by Newton-Rapshon method with respect to , and . The skeleton stress and back stress for a given strain increment at the global iteration step are obtained by local iteration of (18) at each stress integration point. The consistent tangent modulus is obtained as c ep e (tr ) (19)
Weak forms of the equations (10)(12) are implemented in a finite element formulation. Newmark implicit scheme is used for time integration. The primary variables are the second-order material
where cep is the elasto-plastic tensor, e(tr) is the trial elastic strain (given strain). The differentiation
1095
of (19) is carried out at each converged stress derived from the local iteration of (18).
NUMERICAL DATA FOR SIMULATION
Numerical simulations of seismic behavior of a damaged fill slope are performed with the numerical method. The fill slope is located on old valley at Tsukidate town in Miyagi prefecture. The fill slope was partially collapsed and flowed during the 2003 earthquake (Uzuoka et al. 2005). The physical and mechanical properties are investigated with in-situ and laboratory tests. 3.1 Material parameters
Figure 1.
The fill material is a volcanic sandy soil. The physical and mechanical properties are investigated with in-situ and laboratory tests. The material parameters of the constitutive models for skeleton stress and SWCC are determined through the calibration of the undrained cyclic triaxial tests with unsaturated soil. The specimen was made of silty sand obtained from the in-situ fill. The initial dry density of the specimen was about 1.21.3 g/cm3 which roughly agreed with the dry density of the in-situ fill. The degree of water saturation was from about 40% to 79% by controlling air pressure during the isotropic consolidation process. The pore water pressure was almost zero after the consolidation and the pore air pressure increased with the decrease in water saturation. The net stress was about 20 kPa for all specimens and the mean skeleton stress varied with the initial suction dependent on initial water saturation. The cyclic shear was applied to the specimen under undrained air and water conditions. The input axial strain was the sinusoidal wave with multi step amplitudes whose single amplitudes were 0.2, 0.4, 0.8, 1.2, 1.6, and 2.0 with every ten cycles. The frequency of the sinusoidal wave was 0.005 Hz. This loading rate is slow enough to achieve an equilibrium condition between air and water pressure in a specimen. Figure 1 (a) shows SWCC during the undrained cyclic triaxial tests and main drying wetting curves obtained from the water retention test. The measured SWCC are not on the main drying and wetting curves; therefore the scanning curves should be used to reproduce the SWCC during undrained cyclic shear. The logistic SWCC function of (8) is modified to reproduce the scanning curves with simple scaling method in Figure 1 (b).
w se
w w 1 se clg 1 se c 0 0 blg ) 1 exp( alg pa 1 w w 1 se1 1 se1 c p0 pc c c c (20) ( pc p0 ) p0 pa 1 c c c p0 p0 p2
where pc0 and swe0 are the initial suction and effective water saturation respectively, pc1 ( pc0) and swe1 are the suction and effective water saturation on main wetting curve respectively and pc2 is the suction on the scanning curve at swe 1. The material parameters of this modified logistic SWCC model are calibrated to reproduce the measured SWCC during undrained shear in Figure 1 (a). In the triaxial test simulations, the finite element formulation presented in the previous section is not used. Assuming that the variables in the specimen are homogeneous, only the local equilibrium is considered. Table 1 shows the calibrated material parameters of the constitutive model. Figure 2 shows the time histories of pore water pressure, pore air pressure, mean skeleton stress (positive in compression), deviator stress and void ratio from tests and simulations in the case with the initial water saturation of 78.9%. In the test results (denoted Test in the figures), the pore water and air pressure increase, while the mean skeleton stress decreases during cyclic undrained shear. In this case, the mean skeleton stress attains almost zero, which means that the specimen liquefies completely and the suction also becomes to zero. In the simulated results (denoted Model in the figures), the model well reproduces the overall tendency of the test results although the fluctuations of the mean skeleton stress and deviator stress in early time are overestimated. Therefore, in this case with relatively high initial water saturation, the simplified constitutive equation of soil skeleton is applicable to predict pore water and air responses of unsaturated soil in the framework of three-phase porous media
1096
Table 1. Material parameters of the fill. Elasto-plastic model parameters Dimensionless shear modulus, G* Dimensionless bulk modulus, K* Nonlinear hardening parameter, a Nonlinear hardening parameter, b Critical state stress ratio, Mm Yield function parameter, k SWCC parameters Maximum degree of saturation, sws Minimum degree of saturation, swr Main drying curve, alg, blg, clg Main wetting curve, alg, blg, clg Scaling parameter, pc2 (kPa) Physical parameters of water and air Bulk modulus of water, Kw (kPa) Real density of air, aR (t/m3) Gas parameter, 1/(R (s2/m2) 1100 800 1470 1.47 0.98 0.09 0.99 0.255 1.45, 16, 0.037 3.0, 5.0, 0.25 4.0 2.0 10 1.23 103 1.25 105
6
Figure 3. Finite element model of cross section of the fill. Table 2. Material parameters for seismic analysis. Fill n sR (t/m3) wR (t/m3) kws (m/s) kas (m/s) k/k (kPa) (kPa) 0.48 2.48 1.0 1.1 104 1.0 106 3.0/0.33 0.0048 Silt 0.43 2.68 1.0 1.0 109 1.0 1010 181812 45453 0.0048 Tuff 0.43 2.74 1.0 1.0 1011 1.0 1012 437290 437290 0.0019 SS/FT 0.43 2.74 1.0 1.0 1011 1.0 1012 1125000 1125000 0.00096
Figure 4. Time history of input acceleration (Fukumoto et al 2007). Figure 2. Test and simulation with initial water saturation of 78.9%.
theory. The modifications of the constitutive equations are necessary for more precise reproduction. 3.2 Finite element model and boundary conditions
Figure 3 shows the cross section of the in-situ fill slope. Assuming plane strain condition, the cross section is used for finite element modeling. The fill is elasto-plastic material while the silt, tuff and sand stone/fine tuff are linear elastic material with Lame coefficients and . Kelvin type viscosity is assumed in all materials and its viscous coefficient is proportional to the elastic modulus with a multiplier coefficient . Table 1 and 2 show the
material parameters for all layers. The fill elements near both edges are elastic because the behavior is out of scope. Before the seismic analysis, a static self-weight analysis is performed in order to determine the initial stress and moisture conditions. The soil displacement at the bottom boundary is fixed in all directions and the lateral boundaries are vertical rollers. The bottom and right lateral boundaries are impermeable and a left part of the surface on the fill is permeable with zero water pressure. The total water head along the right lateral boundary is fixed because there was no heavy rain before the earthquake. In the seismic analysis, the input acceleration is applied through dampers at the bottom. The lateral boundaries are also supported with dampers
1097
which are connected to the free field motions. The free field motions are separately calculated by one dimensional seismic response analyses. This acceleration history is a calculated wave by using empirical Greens function method using two after shocks comparing several seismic intensities surrounding Tsukidate town (Fukumoto et al. 2007). The coefficients in Newmark implicit time integration are 0.6 and 0.3025. The time increment is 0.002 seconds and the time duration is 20 seconds. 4 NUMERICAL RESULTS OF SIMULATION
Figure 5 shows the initial distributions of pore water pressure, pore air pressure, suction and water saturation in the elasto-plastic fill obtained by the static self-weight analysis. The air entry value of the fill is about 10 kPa; therefore the boundary between saturated and unsaturated portions is located slightly above the ground water table. The ground water table roughly agrees with the measured one after the earthquake (Uzuoka et al. 2005). Figure 6 shows the distributions of pore water pressure, pore air pressure, suction and water saturation in the elasto-plastic fill after the earthquake. Both pore water and air pressures increase in the saturated and unsaturated fill, thus the suction decrease in the unsaturated fill. However, the water saturation does not largely increase because the scanning SWCC is used during the seismic analysis. Figure 7 shows the distribution of Skeleton Stress
Figure 6. Distributions after the earthquake (The deformation scale is ten times as the model scale.).
Figure 7. Distribution of skeleton stress reduction ratio after the earthquake (The deformation scale is ten times as the model scale.).
Reduction Ratio (SSRR) after the earthquake. The SSRR is defined as 1 p/p0 where p0 is the initial value of mean skeleton stress p. The SSRR in the saturated fill including the capillary zone reaches almost one at some elements; liquefaction occurs in some saturated parts. The SSRR in the unsaturated fill above the ground water table increases, which reduces the stiffness and strength of unsaturated fill. These results will cause the failure of fill during the earthquake. However, the relationship between the portion with large SSRR and the observed failure surface (dash line in Fig. 7) is not obvious. We need further investigations to reproduce the observed failure configuration. 5 CONCLUSIONS
Figure 5.
Initial distributions before the earthquake.
A dynamic three-phase coupled analysis was validated through numerical simulations of seismic behavior of a partially collapsed fill slope during the 2003 earthquake. The equations governing the dynamic deformation of unsaturated soil were
1098
derived here based on porous media theory and constitutive models. The numerical results showed the reduction of skeleton stress in the unsaturated fill as well as liquefaction in the saturated fill during the earthquake. These results will cause the failure of fill during the earthquake. REFERENCES
Armstrong, P.J. & Frederick, C.O. 1966. A mathematical representation of the multiaxial Bauschinger effect. C.E.G.B. Report RD/B/N731, Berkeley Nuclear Laboratories, Berkeley, UK. Biot, M.A. 1962. Mechanics of deformation and acoustic propagation in porous media. J. of Applied Physics 33: 14821492. de Boer, R. 2000. Contemporary progress in porous media theory. Applied Mechanics Reviews 53(12): 323369. Borja, R.I. 2006. On the mechanical energy and effective stress in saturated and unsaturated porous continua. Int. J. of Solids and Structures 43(6): 17641786. Fukumoto, S., Unno, T., Sento, N., Uzuoka, R. & Kazama, M. 2007. Estimation of Strong Ground Motions at Tsukidate Landslide Site during the 2003 Sanriku-Minami Earthquake. J. of Japan Association of Earthquake Engineering 7(2): 160179 (in Japanese). Gallipoli, D., Gens, A., Sharma, R. & Vaunat, J. 2003. An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour. Geotechnique 53(1): 123135. Meroi, E.A. & Schrefler, B.A. 1995. Large strain static and dynamic semisaturated soil behavior. Int. J. for Numerical and Analytical Methods in Geomechanics 19(8): 1106.
Okamura, M. & Soga, Y. 2006. Effects of pore fluid compressibility on liquefaction resistance of partially saturated sand. Soils and Foundations 46(5): 93104. Ravichandran, N. & Muraleetharan, K.K. 2009. Dynamics of unsaturated soils using various finite element formulations. Int. J. for Numerical and Analytical Methods in Geomechanics 33: 611631. Schrefler, B.A. & Scotta, R. 2001. A fully coupled dynamic model for two-phase fluid flow in deformable porous media. Computer Methods in Applied Mechanics and Engineering 190: 32233246. Schrefler, B.A. 2002. Mechanics and thermodynamics of saturated/unsaturated porous materials and quantitative solutions. Applied Mechanics Reviews 55(4): 351388. Simo, J.C. & Taylor, R.L. 1985. Consistent tangent operators for rate-independent elastoplasticity. Computer Methods in Applied Mechanics and Engineering 48: 101118. Unno, T., Kazama, M., Uzuoka, R. & Sento, N. 2008. Liquefaction of unsaturated sand considering the pore air pressure and volume compressibility of the soil particle skeleton. Soils and Foundations 48(1): 8799. Uzuoka, R., Sento, N., Kazama, M. & Unno, T. 2005. Landslides during the earthquake on May 26 and July 26, 2003 in Miyagi, Japan. Soils and Foundations 45(4): 149163. Uzuoka, R., Mori, T., Chiba, T., Kamiya, K. & Kazama, M. 2009. Numerical prediction of seepage and seismic behavior of unsaturated fill slope, in Proc. of the International Symposium on Prediction and Simulation Methods for Geohazard Mitigation, Kyoto, 159165. Zienkiewicz, O.C. & Shiomi, T. 1984. Dynamic behavior of saturated porous media: The generalized Biot formulation and its numerical solution. Int. J. for Numerical and Analytical Method in Geomechanics 8: 7196.
1099
Engineering applications and case histories

Soil-atmosphere interaction
Influences of the soil-atmosphere interface interactions on the soil water content and temperature profiles
K.V. Bicalho & G.P.D. Vivacqua Y.-J. Cui
Federal University of Espirito Santo, Vitoria, Espirito Santo, Brazil ENPC CERMES, UR Navier, Universit Paris Est, France
ABSTRACT: This paper presents the results of a series of numerical analyses aiming to predict temperature, water content, and pore-water pressure changes due to climatic effects in a soil profile during a long term period by considering the soil atmosphere interface interactions. A soil-atmosphere interface model is used to calculate the evaporation rate and heat flux on the soil surface; the water transport equations (liquidDarcys law and vaporFicks law) coupled to heat flow equation (de Vries 1963) are solved to determine the profiles of soil temperature, water content or pore water pressure. A sensitivity analysis was also carried out to study how the variations of the soil albedo value (i.e., the ratio of reflected to incident solar radiation), the saturated hydraulic conductivity and the Initial Temperature Profiles (ITP) can influence water content, suction and temperature changes. Our analysis uses field measurements in Mormoiron, France, carried out from 2004 to 2005. The results show that the active zone (where the suction profile is influenced by seasonal environmental changes) is generally about 1.5 m deep in the investigated region. The results also suggest that the ITP can affect the temperature profiles but its influence on the suction and water content profiles is very small. During the cold season, precise albedo values are not very important nor very sensitive in influencing the water balance. 1 INTRODUCTION of water from a wet soil (Choudhury et al. 1986 or Xu and Qiu 1997). The model was validated with several data sets and able to satisfactorily predict the behaviour and volumetric water content profiles for non-cohesive and cohesive soils by Gao (2006). This paper discusses the ability of the method to predict evaporative fluxes over extended periods of drying between rainfall events. The predicted soil temperature profiles are compared with direct measurements at a site in Mormoiron in France. It is used an extensive data collection covering both atmospheric and soil data at the site. In order to model the changes in soil temperature and water content (or suction) profiles due to climatic effects during a given period, it is necessary to determine the soil albedo value (i.e., the ratio of reflected to incident solar radiation). Albedo is a key parameter which controls surface energy exchange. It is a function of several surface parameters including soil color, water content, roughness and vegetation cover. Since the soil albedo value is not known, the sensitivity of predicted temperature changes due to ground-atmospheric interactions to the variations of the soil albedo is investigated. It is also discussed the influence of the initial soil temperature
It is widely recognized the importance of understanding and predicting soil water content (or suction) and temperature profiles in the unsaturated region. The spatial variation of soil water content (suction) in a drying soil sample is mainly dependent on the local environmental conditions, initial water content and temperature, hydromechanical properties of the soil, and on the boundary conditions at the soil-atmosphere interface. The objective of this study is to simulate in situ water content, suction and temperature changes due to climatic effects in a soil profile during a given period by considering the soil atmosphere interface interactions. It is applied the principle of mass and energy conservation to describe onedimensional water (liquid and vapour) and heat flow in unsaturated soil and the surface energy balance approach to evaluate the evaporation fluxes from a soil surface. A one dimensional explicit finite difference program developed by Gao (2006) is used for doing the numerical simulations. The program models the coupled water flow and heat flow in unsaturated soil and uses an energy estimation method for determining the evaporation rate
1103
profile and saturated hydraulic conductivity on the predicted soil temperature and water content (or suction) profiles. 2 SOIL-ATMOSPHERE INTERFACE MODEL
Aerodynamic resistances ra is calculated as: ra ra 0 1 (1 Ri (Ts Ta )) (4a)
The model computes the evaporation rate from soil by solving a coupled water (liquid and vapour) transport equations (Darcys law and Ficks law), heat flow equation (de Vries 1963) analysis, and the surface energy balance used for defining a reliable boundary setting method for extended periods of evaporation simultaneously. The energy balance equation expressing the net radiation flux for the surface is (Blight 1997): Rn LE H G (1)
where 0.75 in unstable condition (Ts Ta), and 2 in stable condition (Ts Ta), Ri is the Richardson Number and ra0 is the aerodynamic resistances derived from a logarithmic wind profile: za d ln z0 k 2ua
2
Ri
5g ( za d ) , ra 0 2 ua Ta
(4b)
where LE is the latent energy transfer (positive for evaporation and negative for condensation), H is the sensitive heat flux for the air (positive when energy is used to warm the air and negative when the air loses energy due to cooling), and G is the ground heat transfer (positive when energy is transferred to the subsurface and negative when energy is transferred to the atmosphere). The energy estimation method proposed by Choudhury et al. (1986) or Xu and Qiu (1997) is used for determining the evaporation rate of water from a wet soil. In this approach, H and LE are calculated from the turbulent exchange equations by: H
where z0 is the roughness length parameters for momentum (wind) and sensible heat transport, Za is the measurement height for wind speed ua and relative humidity d is the displacement height, and d 0 for bare soil, k is a constant equal to 0.41. Details of the used method are discussed in Gao (2006). 2.1 Constitutive functions
To solve the governing equation the suctionvolumetric water content and suction-unsaturated hydraulic conductivity relationships must be known. The relationships are (Juarez-Badillo 1992):
w
ws r
r s 1 1 ws w1 r s1 ks k s 1 s 1 kw1 s1
r
(5)
aC pa (Ts Ta )
ra Lv Mw ( pvz 0 pvza ) RTra
(2) (3)
kw
(6)
LE
where a is the air density, Cpa is the specific heat of air equal to 1.013 103 J/(kg . K), Ts is the soil surface temperature, Ta is the air temperature at reference height Za, T is the average temperature and T(Ta Ts)/2, pvz0 is the vapor pressure at the soil surface, pvza is vapor pressures of in the air at reference height Za, ra is the aerodynamic resistances for the sensible and latent heat fluxes, Lv is the latent heat of vaporization, Mw is the molecular mass of water equal to 0.018 kg/mol, R is the gas constant equal to 8.314 J/(mol . K). These equations take into both account both the soil-atmosphere resistance (ra), depending on the wind velocity (ua), the soilatmosphere temperature (Ts Ta) or vapor pressure (pvz0 pvza) gradients. The soil heat flux G is then calculated from the energy balance equation.
where ws is the saturated volumetric water content, r is the residual volumetric water content, w1 is the value of water content corresponding to suction s1, and is the parameter that controls the shape of the s-w curve, ks is the water permeability at s 0, and kw1 is the hydraulic conductivity corresponding to suction s1. The thermal conductivity of soil, , is (de Vries 1963):
fs s s fww w fa a a fs s fww fa a

air
(7)
a dry
water
vapor
where the thermal conductivity of solid, s (k)q (k*)1-q (Johansen 1975), q % quartz, for q 0,
1104
Suction (kPa)
s 2.0 and q 100%, s 7.7, the thermal conductivity of water w (0.57 W/mC), dry-air (0.025 W/ mC), water-vapor (0.608 w W/mC), fs, fw and fa are the weight coefficient for solid, water and air respectively and fw 1.0, fs 1 s 1 w fa
1
1,00E+06 1,00E+05
Juarez-Badillo (1992) - Z > 3.45 m Juarez-Badillo (1992) - Z < 3.45 m
1,00E+04 1,00E+03 1,00E+02 1,00E+01 1,00E+00 1,00E-01 0,00 10,00 20,00 30,00 40,00 50,00 60,00
(8)
Suction (kPa)
1 3 a gi 1 1 3 i 1 w
1
(9)
1,00E+06 1,00E+05 1,00E+04 1,00E+03 1,00E+02 1,00E+01 1,00E+00 1,00E-01 0,00E+00 1,00E-12 2,00E-12 3,00E-12 4,00E-12 5,00E-12 6,00E-12 Hydraulic conductivity (m/s) Juarez-Badillo (1992) - Z < 3.45 m Juarez-Badillo (1992) - Z > 3.45 m
where gi are called shape factors (g1 g2 g3 1) (Gao 2006): for w 0.121 g1 g2 0.333 0.105 (w 0.121) 0.105 0.236 0.121
for w 0.121 g1 g2 3 0.105 0.015 w 0.015 0.121
Figure 1. The adopted suction-volumetric water content and suction-unsaturated hydraulic conductivity relationships.
NUMERICAL SIMULATIONS
A one-dimensional computer program developed by Gao (2006) using an explicit finite difference method was used to predict water content, porewater pressure and temperature profiles considering the coupled water flow and heat flow and the interaction between the ground and atmosphere. The coupled water flow and heat flow equations are solved assuming that the soil skeleton is rigid. The input data are ws 0.49 0.4, r 0.08, w1 0.24, s1 700 200 kPa, and 1.1, ksat 1.2 1011 2.4 1010 m/s) and kw1 1.2 1014 m/s, s1 40 kPa, " 1.25 (Figure 1). The thermal coefficients Cw (4.15 106 J/m3C), Cs (2.24 106 J/m3C), and q 50% and s 3.92. The climatic data measured at the Mormoiron from December 2003 to December 2005 (i.e., solar radiation (0.05 to 0.35 kW.m2), energy, precipitation, runoff, wind speed (2 to 14 m/s), air temperature (0 to 25C), and air humidity) were used in the numerical simulations (Figure 2). The air relative humidity varies between 30 and 100%, but it does not necessary follow the precipitation pattern. The air relative humidity depends not only on precipitation, but also on air temperature and wind speed (Cui & Zornberg 2008). The analyses performed employed the same values of the depth of the analysis (ZMAX 5.25 m), the constant spacing (z 0.0050.05 m), the time step (t 0.5 s),
the runtime, TMAX (s), bottom volumetric water content boundary (b 0.30), bottom temperature boundary (Tb 14C), and initial volumetric water content profile (Figure 4). The atmospheric water balance from 1964 to 2005 at the site Mormoiron has shown that evapotranspiration often exceeds precipitation. Figure 3 shows the atmospheric water balance from December (D) 2003 and December (D) 2004 and January (J) to December 2005. A water deficit is observed in most of time throughout the studied years except for a brief period in December 2003, October 2004, April and October 2005. Thus, the investigated years correspond to drier conditions where the recharge of the water table did not take place. Since the Initial Temperature Profile (ITP) is not known, we investigate the sensitivity of predicted volumetric water content (or pore-water pressure) and temperature profiles to the variations of the ITP (Figure 4). The dependence of the soil temperature on the soil water retention curve was not considered in the numerical simulations. The insensitivity of the results (Cases B and D) presented in Figure 5 is visible. The volumetric water content and pore water pressures at depth 1.5 m are almost constant. Figure 6 presents the predicted temperature profiles from January to July 2004 considering cases B and D. As can be seen from this comparison, the value of the ITP can affect the temperature profiles. The influence is more accentuated for the near surface layers, where more extreme variations in temperature occur. The results show that the temperatures increase
1105
4 3.5
3 2.5 2 1.5 1 0.5 0 0 100 200 300 400 500 600 700 800 900
Precipitation / Evaporation (mm/month)
250 200 150 100 50 0 D J F M A M J J A S
Solar radiation (kW/m2 day)
Precipitation (mm/month) Evaporation (mm/month)
Time (days)
Month (2003/2004)
80
Rainfall (mm)
40 20 0 0 200 400 600 800 1000
Precipitation / Evaporation (mm/month)
60
250 200 150 100 50 0 J F M A M J J A S O
Precipitation (mm/month) Evaporation (mm/month)
Time (days)
30
Month (2005)
Air Tem perature (C)
20
10 0
Figure 3. Atmospheric water balance for the investigated region in France from December (D) 2003 to December (D) 2004 and January (J) to December (D) 2005.
0 200 400 600 800 1000
-10
Time (days)
0 5
Volumetric water content (%) 10 15 20 25 30 35 0 1 Depth (m) 2 3 4 5
Air relative humidity (%)
120 100 80 60 40 20 0
200
400
600
800
1000
Time (days)
Temperature (degree)
15
0 0 5 10 15 20
Wind speed (m/s)
10
1 2
Depth (m)
Case B 3 4 5 6 Case D
0 0 200 400 600 800 1000
Time (days)
Figure 2. Distribution of the daily Solar Radiation (10 kW/m2), Rainfall (mm), Air Temperature (C), Air Relative Humidity (%), and Wind speed (m/s).
Figure 4. Initial volumetric water content and temperature profiles.
with the depth during the cold season (January to March 2004) and decrease with the depth during the warm season (April to July 2004). The results of the simulations using different saturated hydraulic conductivities (case B ksat 1.2 1011 2.4 1010 m/s; case B4 ksat 1.2 109 2.4 108 m/s) are presented in Figures 7 and 8. The insensitivity of the results is visible. Comparison of the soil temperature profiles for Cases B and B4 reveals the same insensitivity of the results to the considered changes in ksat. The years of 2004 and 2005 correspond to drier
conditions (high suction values) and changes of ksat of this magnitude have not significant effect on the considered suction-unsaturated hydraulic conductivity relationship (see Figure 1). The results presented in Figures 7 and 8 show that the active zone (where the suction profile is influenced by seasonal environmental changes) is generally about 1.5 m deep in the investigated region. In order to investigate the effects of the thickness of the upper layer (Z) in the two layers homogeneous soil column, Figure 9 shows the results of the numerical simulations of volumetric water
1106
Figure 5. Influence of initial temperature profiles on the volumetric water content profiles.
Temperature ( C)
0 0 0,5 1 5 10 15 20
o
Figure 8.
Influence of the ksat on the suction profiles.
Depth (m)
01/2004-Case D 1,5 2 2,5 3 3,5 02/2004-Case D 03/2004-Case D 01/2004-Case B 02/2004-Case B 03/2004-Case B
Temperature ( C)
10 0 0,5 1 12 14 16 18 20 22 24 04/2004-Case D 05/2004-Case D 06/2004-Case D 07/2004-Case D 04/2004-Case B 05/2004-Case B 06/2004-Case B 07/2004-Case B
Depth (m)
1,5 2 2,5 3 3,5
Figure 9.
Influence of Z on the water content profiles.
Figure 6. Influence of initial temperature profiles on the temperature profiles (2004).
Figure 7. profiles.
Influence of the ksat on the water content
content profiles using different thicknesses of the upper layer, i.e., Z 3.45 m (Case B) and Z 0.5 m (Case B5). The results suggest that the differences between Cases B and B5 are very small. The changes in soil profiles during a given period depend on the ratio of reflected to incident solar
radiation (i.e., the soil albedo value). It is a function of several surface parameters including soil color, water content, roughness and vegetation cover, usually being lower for wet and rough conditions. The albedo value ranges from 0 to 1. The value of 0 refers to a blackbody, a theoretical media that absorbs 100% of the incident radiation. Albedo ranging from 0.10.2 refers to dark-colored, rough soil surfaces, while the values around 0.40.5 represent smooth, light-colored soil surfaces. The value of 1 refers to an ideal reflector surface (an absolute white surface) in which all the energy falling on the surface is reflected. The sensitivity of predicted temperature changes to the variations of the soil albedo is investigated in Figure 10, where case B (soil albedo 0.15) and case C (soil albedo 0.05). The insensitivity of the results (Cases B and C) during the cold season (January to March 2005) is shown in Figure 10. Small changes in the soil temperature values (increase) due to the variation (decrease) of the soil albedo (Case C) are observed during the warm season (April to August). Figure 11 presents the comparison of predicted and measured changes in soil temperature at three different depths (0.5 m, 1.5 m, and 2.5 m) during 2005 in Mormoiron, France. The results suggest that in the near the surface layers the simulations were less satisfactory due to probably vegetation
1107
Temperature (oC) 0 0 0,5 1 Depth (m) 1,5 2 2,5 3 3,5 5 10 15 20 25 30 0 1 /2 0 0 5 - C ase B 0 2 /2 0 0 5 - C ase B 0 3 /2 0 0 5 - C ase B 0 4 /2 0 0 5 - C ase B 0 5 /2 0 0 5 - C ase B 0 6 /2 0 0 5 - C ase B 0 7 /2 0 0 5 - C ase B 0 8 /2 0 0 5 - C ase B 0 1 /2 0 0 5 - C ase C 0 2 /2 0 0 5 - C ase C 0 3 /2 0 0 5 - C ase C 0 4 /2 0 0 5 - C ase C 0 5 /2 0 0 5 - C ase C 0 6 /2 0 0 5 - C ase C 0 7 /2 0 0 5 - C ase C 0 8 /2 0 0 5 - C ase C
Figure 10. Influence of the soil albedo values on the temperature profiles (2005).
0 0 0,5 D e p th (m ) 1 1,5 2 2,5 3 3,5
10
Temperature (oC) 15
20
0 1 /2 0 0 5 -30 M easured 25 0 2 /2 0 0 5 - M easured 0 3 /2 0 0 5 - M easured 0 4 /2 0 0 5 - M easured 0 5 /2 0 0 5 - M easured 0 6 /2 0 0 5 - M easured 0 7 /2 0 0 5 - M easured 0 8 /2 0 0 5 - M easured 0 1 /2 0 0 5 - C ase C 0 2 /2 0 0 5 - C ase C 0 3 /2 0 0 5 - C ase C 0 4 /2 0 0 5 - C ase C 0 5 /2 0 0 5 - C ase C 0 6 /2 0 0 5 - C ase C 0 7 /2 0 0 5 - C ase C 0 8 /2 0 0 5 - C ase C
values) and changes of ksat of this magnitude have not significant effect on the considered suctionunsaturated hydraulic conductivity relationship (see Figure 1). The results also show that the active zone (where the suction profile is influenced by seasonal environmental changes) is generally about 1.5 m deep in the investigated region. During the cold season, precise albedo values are not very important nor very sensitive in influencing the water balance. The comparison of predicted and measured changes in temperature profiles suggest that in near the surface layers the simulations are less satisfactory due to probably vegetation effects or other mechanical phenomena (i.e., soil cracking). ACKNOWLEDGEMENT The first author is grateful for sponsorship by the Brazilian government agency CAPES and CNPq. This work has been performed within the French ANR-RGCU project: ARGIC. The meteorological data was provided by Mto France and the soil data by BRGM. REFERENCES
Blight, G.E. 1997. Interactions between the atmosphere and the earth. Geotechnique, 47 (4):715767. Choudhury, B.J., Reginato, R.J. & Idso, S.B. 1986. An analysis of infrared temperature observations over wheat and calculation of latent heat flux. Agric. For. Meteorol., 37:7588. Cui, Y.J., Lu Y.F., Delage P. & Riffard M. 2005. Field simulation of in site water content and temperature changes due to ground-atmospheric interactions. Geotechnique, 55 (7):557567. Cui, Y.J. & Zornberg, J. 2008.Water Balance and Evapotranspiration Monitoring in Geotechnical and Geoenvironmental Engineering. Geotech Geol Eng DOI 10.1007/s10706-008-9198-z. De Vries, D.A. 1963. Thermal properties of soils. In physics of plant environment (ed. W.R. van Wijk), pp. 210235. Amsterdam: north-Holland. Gao Y.B. 2006. From Meterological data to the predication of embankment stability. Technical report, CERMES-ENPC, Paris, France. Johansen, O. 1975. Thermal conductivity of soils. Ph.D. thesis, Trondheim, Norway. (CRREL Draft Translation 637, 1977). ADA 044002. Juarez-Badillo, E. 1992. Leyes naturals en geomecanica. Volumen Raul J. Marsal 31. Mexico: SMMS. Xu Qin & Qiu Chong-Jian 1997. A variational method for computing surface heat fluxes from ARM surface energy and radiation balance system. J. of App. Meteorology, 36 (1):311.
Figure 11. Comparison of predicted and measured changes in temperature profiles during 2005.
effects or other mechanical phenomena (i.e., soil cracking). A sensitivity analysis of temperature and water content profiles to the changes to the variations of other unknown parameters (i.e., soil water content that depend on the soil temperature) should also be investigated. Cui et al. (2005) proposed to consider the superficial zone independently, using different values of the soil parameters. 4 CONCLUSIONS
The numerical analyses carried out to investigate the sensitivity of water content, pore-water pressure and temperature changes to the variations of the Initial Temperature Profiles (ITP) suggest that the value of the ITP can affect the temperature profiles and the influence of the considered changes on the pore water pressure and volumetric water content profiles is very small. The numerical analyses reveal the insensitivity of water content, pore-water pressure and temperature changes to the variations of the saturated hydraulic conductivity ksat. The years of 2004 and 2005 correspond to drier conditions (high suction
1108
An experimental root barrier to protect the foundations of a building from the desiccating effects of tree roots
G.E. Blight
ABSTRACT: An experimental root barrier to isolate the shallow foundation of a house from the desiccating effect of the roots of a nearby tree is described and its performance is reported. Four years after installation, the barrier and a similar one installed 23 years earlier, were exhumed to study the growth of roots near the barrier. The paper describes what was found. 1 INTRODUCTION
Little is known about the effectiveness of root barriers in preventing root activity beneath house foundations. They are usually fairly expensive to install. As with underpinning, there is uncertainty about how extensive a root barrier should be, both laterally and in depth. If the trees are older than the building, a barrier that cuts off the root system beneath a house will cause swelling and heave as if the tree were entirely removed. Inserting root barriers close to trees can be dangerous. If sufficient of the root system is severed, the tree may lose lateral stability and fall. Even if rapid instability is not caused, the tree may slowly die and become unstable at a later date.
Damage to buildings on shallow foundations by desiccation arising from invading tree roots is a long-known and well-recognized problem, but is difficult to deal with (e.g. U.K. Building Research Establishment, 1985, Biddle, 2001). An earlier paper (Blight, 2006) described the performance of an experimental root barrier that, in 2006, had been in place between a large deciduous tree and the wall of a house for three years. In the winter of 2007 (July), it was decided to fell the tree and complete the experiment by observing soil moisture contents and their seasonal changes with the root barrier still in place, but without the tree, and finally, to exhume the barrier and investigate the root development on either side of it. This paper should be regarded as a continuation of the 2006 paper, but to provide continuity, the earlier paper will very briefly be summarized. 2 EARLIER OPINIONS AND INSTALLATION OF ROOT BARRIERS
A vertical root barrier interposed between the tree or trees and the building would seem to be the obvious way of protecting a shallow foundation from the effects of localized shrinkage of clays caused by extraction of water by invading tree roots. However this technique is so seldom described, as to be almost unknown in the literature of expansive/ shrinkable clays. For example, root barriers are not mentioned in the 1983 Geotechnique Symposium in Print on the Interaction Between Vegetation and Structures. The British Building Research Establishment (BRE) Digest 298 (1985) mentions root barriers as a possible remedy, but in a way which is not likely to encourage trying root barriers in practice, as the following quote will show:
More recently, Biddle (2001) has made a very similar statement to the 1985 BRE statement quoted above in which he discourages the use of root barriers. In the same symposium volume, Radevsky (2001) stated that some local authorities (in Australia) believe in root barriers to prevent (insurance) losses and have been researching into appropriate specifications for the size, composition and optimum location of such barriers. In a third paper, Pengelly and Addison (2001) describe the remediation of a building that had been damaged by the roots of a line of oak trees situated 2.4 m from one of the exterior walls, by means of a root barrier. But there was no follow-up study of the barriers efficacy. 3 THE EXPERIMENTAL SITE
The experimental site is at Clarens, which is situated in the Free State Province of South Africa at latitude 2831 south and longitude 2826 south and longitude east at an elevation of 2000 m above sea level. Clarens is in the foothills of the Maluti mountain range, and is ringed by sandstone cliffs of the
1109
Clarens formation, which in turn are capped by strata of basaltic Stormberg lavas. The test site is situated on a gently sloping plain of colluvium consisting of silty clay deriving from the weathering of the sandstones and lavas. The colluvium is underlain by horizontally bedded sandstone. The soil is moderately expansive/shrinking, and contains closely spaced slickensides and extensive microfissuring between slickensides. The sand, silt and clay contents are close to constant with depth, and vary as follows: sand: 2025%, silt: 4753%, clay: 2233%. Index data for the Clarens soil are as follows: Liquid Limit: 46%, Plasticity Index: 18%, Linear Shrinkage: 9%. The soil properties are almost constant with depth and the transition from colluvium to unweathered sandstone occurs very suddenly over a depth of 50100 mm. At the location of the house the soil depth is only 500600 mm. The water table lies within the sandstone at 57 m below the surface. The climate at the site is continental, with warm, moist summers (daily temperatures between 10 and 25C) and cool dry winters (10 to 15C. Being in the Southern Hemisphere, the summer solstice occurs on December 21 and the winter solstice on June 21. 4 A SAMPLING OF EARLIER WATER CONTENT MEASUREMENTS RELATED TO THE CLARENS ROOT BARRIER
shown by the relative positions of the 25% water content contour on either side of the barrier. Figure 1c shows the water stored in the soil (i.e. the integration of the water content over the soil depth) on either side of the barrier, with the storage under the tree always less than that to the left of the barrier. 5 THE EFFECT OF FELLING THE TREE
The tree was felled at the start of July 2007, after 2 months of the dry season. The results of subsequent sampling for soil water content are shown in Figure 2. The first sampling for water content was in mid-August when there still appeared to be the possibility of a discontinuity in water content contours at the root barrier. However, the stored water profile appeared continuous across the barrier. By early December 2007, the water content contours and stored water profiles both looked continuous, as was the case 9 months later, in late July 2008. Hence there can be little doubt that the differentially applied desiccating effect of the tree roots caused the discontinuities visible in Figure 1, and also that the root barrier prevented the invasion of roots from desiccating the soil near the house foundation. 6 EXAMINATION OF THE EXHUMED ROOT BARRIER
Figure 1a shows the plan layout of the root barrier experiment at Clarens, including the wall of the house, the root barrier and the tree. The barrier was installed in July 2003 and was of very simple construction. A trench 500 mm deep was excavated, severing all the roots it crossed. The trench was lined with corrugated glass-fibre reinforced acetate sheeting, with the corrugations running vertically, and each 450 mm long sheet overlapped by two corrugations to form a barrier that extended from 500 mm to 50 mm below surface. The trench was backfilled with the soil excavated from it, the backfill being compacted, as well as possible, in layers placed at 100 mm depth intervals. Figure 1b shows water content contours measured on three occasions by sampling at the positions shown, with a 50 mm diameter hand auger. The discontinuity of the water content contours between the left and right of the root barrier shows how desiccation by the tree roots was confined to the volume of soil to the right of the root barrier. The wet season at Clarens starts in October and ends in April. The effects of the rain are shown by, for example, the changing elevation of the 25% water content contour on either side of the barrier, and the effects of the desiccation by the tree are
The root barrier was exposed at three points along its length on 30 July 2008, and one of these exposures is shown by Figure 3(a), which should be viewed so that the number (5 m) on the measuring tape is the right way up. At this point, a major root about 120 mm in diameter had been severed to install the root barrier. The corrugated root barrier, with a lap joint crosses the field of view diagonally in the centre of the photo. The part of the root closest to the camera was dead and had rotted. The cambium layer of the root attached to the tree had healed (the light ring around the root behind the tape) and fine fibrous roots were sprouting from the bark of the root beyond the wound. The fine roots protruding from the sides of the inspection trench are all grass roots. In July 1985 the writer had installed a similar root barrier to isolate the roots of a newly planted grape vine from the walls of the house. This barrier is composed of overlapping plane sheets of asbestos cement. The barrier had never been inspected in the 23 years since its installation and there had been no indication of undue desiccation of the soil between the barrier and the
1110
Figure 1. (a) Plan layout of experimental root barrier. The effect of the root barrier on: (b) water contents in the soil and (c) water storage in the soil.
house. In the meantime, the vine had grown very large, with twin 150 mm diameter main stems. It was decided to open up and inspect this barrier as well, and part of the exposed barrier is shown in Figure 3(b), taken from the house side of
the barrier. Here it was found that although no large diameter roots had directly crossed the barrier, one 15 mm diameter root had entered through the overlap between asbestos cement sheets (marked by the darker coloured pen to the
1111
Figure 2.
The effect of felling the tree on (a) water content contours on 3 dates and (b) water stored in the soil.
right) and had continued growing between the barrier and the wall of the trench on the house side. Other large roots had entered the overlaps on the vine side, had passed along the house side of the barrier and then back to the vine side at the next overlap (marked by the lighter coloured pen to the left. Whatever side they were,
it was obviously easier to grow along the side of the trench than to penetrate the undisturbed soil either toward the house or on the vine side. A few very small diameter (1 to 2 mm) roots had penetrated the barrier and grown toward the house (e.g. the small root visible crossing the large root near the dark coloured pen).
1112
barrier. The barrier need not be anything elaborate or expensive, and because tree roots are concentrated in the upper half to one metre of soil (see, e.g. Dobson & Moffat, 1995) they do not have to be very deep to be effective. The measurements show that a root barrier can cause the soil water regimes on either side of the barrier to be completely different. The examination of the 22 year old root barrier showed, in addition, that the trench in which the barrier is installed causes roots to grow preferentially along the barrier and inhibits any tendency for roots to penetrate the barrier and invade undisturbed soil on the other side. REFERENCES
Biddle, P.G. 2001. Tree root damage to buildings. In Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE Geotechnical Special Publication 115, Reston, USA, ASCE: 123. Blight, G.E. 2006. Interaction between trees and buildings on shrinkable soilsa detailed field study. In Miller, G.A., Zapata, E.C., Houston, S.L. & Fredlund, D.G. eds, 1: 281292. Unsaturated Soils 2006, ASCE Geotechnical Special Publication 147, Reston, USA, ASCE, 281292. Building Research Establishment 1985. The Influence of Trees on House Foundations in Clay Soils. Digest 298. Watford, U.K. Building Research Establishment. Dobson, M.C. & Moffat, A.J. 1995. A re-evaluation of objections to tree planting on containment landfills. Waste Management and Research, 13(6): 579600. Pengelly, A. & Addison, M. 2001. In-situ modification of active clays for shallow foundation remediation. In Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE Geotechnical Special Publication 115, Reston, USA, ASCE 192214. Radevsky, R. 2001. Expansive clay problemshow are they dealt with outside the U.S.? In Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE Geotechnical Special Publication 115, Reston, USA, ASCE: 172191.
Figure 3.
(a) 2003 root barrier exposed in 2008.
Figure 3.
(b) 1985 root barrier exposed in 2008.
CONCLUSIONS
The experiment on the root barrier installed in July 2003 showed that the effects of desiccation by invading tree roots on shallow foundations can successfully be prevented by means of a root
1113
Soil heat and evaporation from soil

G.E. Blight
ABSTRACT: Because of difficulties in applying the solar energy balance method of measuring evaporation from soil surfaces, an investigation into the method has been made. The main difficulties are: Comparisons of evaporation estimated by energy balance with corresponding soil water balances have shown that the energy balance method overestimates evaporation. The method assumes that solar energy that heats the soil detracts from energy available to produce evaporation. The energy balance equation does not predict nocturnal evaporation or inefficiencies in converting energy into evaporation. The paper briefly addresses these difficulties. 1 INTRODUCTION wavelengths between 0.4 and 0.7 m.) This shortwave radiation causes day time heating of Earths surface (near-surface air, vegetation and soil). To preserve the overall radiation balance of Earth, most of the heat (Ro) is re-radiated at predominant long wavelengths of 720 m (Sellers, 1965). The net incoming solar radiation at ground level, Rn is absorbed by the circular projected area of Earth, during daylight hours. The outwardly radiated heat Ro can leave Earths surface throughout the day and night, and is therefore radiated from the entire surface of the globe. If the radius of Earth is r, for a balance between incoming and outgoing radiation, on average and ideally, Rn . r2 Ro4r2 or Ro Rn/4. During daylight hours the net short-wave energy gain exceeds the net long-wave loss, causing an energy surplus at the surface. This is reversed in the late afternoon and at night when the long-wave radiation dissipates the surplus energy (Munn, 1966). According to Penman (1963) and others, the daily quantity of evaporation can be determined by measuring the components of the solar energy balance equation: LE Rn WE (G H P) (1) where the terms are usually expressed in energy units of kJ/m2, the time-integral over a day of W/m2, and LE is the latent heat consumed in evaporation, Rn is the net incoming vertical short wave solar energy. WE is the wind energy (which will not be considered here). G is the energy expended in heating the near-surface soil (the soil heat); H is the energy consumed in heating the near-surface air (the sensible heat), and P is the energy used in plant photosynthesis (usually less than 2% of Rn).
The solar energy balance method and the equations based on it were pioneered by Bowen (1926), Penman (1963) and Monteith (1980) and represent the generally accepted way of measuring evaporation from a soil surface. In using the energy balance method, a number of difficulties and short-comings have become apparent which led to an investigation into the method being started. Preliminary results of the investigation have been published recently (Blight, 2009) and this paper briefly summarizes these and reports the results of additional research. Although this research was undertaken with municipal solid waste landfilling as the primary area of concern, the principles and conclusions of the work can equally be applied in other areas of unsaturated soil mechanics, such as mine tailings storage and the drying of water purification and sewage sludges. 2 MEASURING EVAPORATION BY SOLAR ENERGY BALANCE
At noon at any point on Earths surface, the solar power penetrating the outer limit of the atmosphere at noon is a fraction, depending on the latitude, of 830 W/m2. In passing through the atmosphere, the incoming radiation will be reduced, and about 50% will reach the ground surface, where 520% of the 50% will be reflected back. The remaining 4047.5% constitutes Rn, the net incoming radiation or solar power. Incoming solar radiation has a predominant wavelength spectrum of 0.32 m (visible light has
1115
LE can be used to calculate the quantity of daily evaporation by dividing the daily total LE by the latent heat for vaporisation in kJ/kg, that is evaporation E LE/ in kg/m2, the equivalent of mm of water. For water, varies almost linearly with temperature (e.g. Calder, 1990) from 2477 kJ/kg at 10C to 2417 kJ/kg at 35C, that is by 2.4 kJ/kg per degC. In terms of equation (1), G is a heat sink that subtracts from LE and therefore reduces evaporation. One of the problems associated with equation (1) is that it includes only the net incoming short-wave-length radiation Rn, but omits the outgoing long-wave-length heat radiation Ro. Because the rate of outgoing heat energy varies during the day and night and can be affected by local weather conditions, short-term fluctuations arise that should be represented by a storage term S, which may be positive or negative (Oke, 1978). With these additions, equation (1) becomes LE (Rn Ro) WE (G H P) S (1a)
In equations (1) and (1a), Rn is measured and G and H are calculated from measured temperatures and specific heats. The method of evaluating each term has, over time, become fixed by specifying a method of evaluation. Thus Rn is evaluated by measuring and time-integrating the vertical component of the incoming solar power. There is some contradiction in the established literature on the terms in the solar energy balance. Penman (1963) does not include soil heat in his much used equation for predicting evaporation, although he does include sensible heat H. Monteith and Unsworth (1990), however, regard G as important both as a sink for Rn during the day, and as a source for LE when the soil begins to cool in the afternoon. The accuracy of equation (1) is difficult to check because of the difficulty of establishing a reference value against which the result of applying the method can be compared. However, recent comparisons (Blight & Fourie, 2005, Blight, 2005) have shown that equation (1) over-estimates actual evaporation. Some of the difficulties associated with equation (1) are illustrated by Figure 1. Because it appeared simpler to study evaporation from a single component system such as water, before progressing to studies on the two or three component system that is soil, the first of the experiments recorded in this paper studied evaporation of water. A 500 mm diameter by 100 mm deep stainless steel pan was let into the surface of a grassed area so that its rim was level with the soil surface. The pan was filled with water and left overnight to reach temperature equilibrium with its surroundings. In the accepted conventional way, the vertical
components of the incoming and reflected solar power (in W/m2) were measured throughout the day, as well as water temperatures. Thereafter the experiment was duplicated with the pan filled with moist sand. An impervious vessel was used so that the depth of soil from which water was evaporating could be precisely defined. The results of the measurements are summarized in Figure 1. In the upper diagram, the parabola-like curves represent the vertical component of the net incoming solar power Rn(V) (incoming minus reflected) in W/m2. The ogive-like curves represent the time-integrated values of Rn(V), i.e. Rn (V), in kJ/m2. The soil heat G and its counterpart, the water heat WH are also shown in Figure 1. Sunrise was at 07.00 and sunset at 17.00. The lower diagram shows the measured water and soil temperatures, relative to the temperature at sunrise, throughout the day. The first difficulty that becomes apparent is that the maximum water heat WH of 8380 kJ/m2 calculated from the maximum water temperature was considerably (77%) in excess of the net cumulative vertical solar energy for the day of 4740 kJ/m2. Obviously, more energy was being absorbed by the water than was represented by Rn(V). When the maximum (Rmax), vertical (Rv) and reflected vertical components of the solar power were measured separately, it was found that Rmax may be 3 times the net value Rn(V), which, in conjunction with the independently measured WH and G values in Figure 1, must mean that not only the vertical component of solar energy is absorbed at a soil or water surface, but also oblique components. A similar result was observed for the soil heat G where the measured maximum soil heat of 5410 kJ/m2 was 3% in excess of the corresponding Rn(V). The second difficulty relates to the values of G. In this set of measurements, G exceeded Rn(V) from 09.00 until 16.00. If G is interpreted as a loss and as subtractive from Rn(V), equation (1) would predict negligible evaporation for the day, whereas the actual evaporation measured by means of the water contents of cores of sand taken from the pan (see caption to Figure 1) averaged 1.6 mm/day. The energy back-calculated from the measured evaporation also exceeded the measured Rn(V) for the water. The third difficulty is that according to equation (1) if Rn 0, and there is no wind, no evaporation should occur. In other words, on calm nights, no evaporation should take place between sunset and sunrise, whereas common qualitative experience is that some evaporation does occur at night. Because evaporation, either from water or sand, could not be measured very accurately in the 500 mm diameter pan tests, in exploring the above difficulties, it was decided to change to smallerscale tests in which more accurate measurement of
1116
Figure 1. Results of twin experiments made on stainless steelpan set level with ground surface. Pan filled firstly with water, secondly with sand. Measured evaporation (averaged over 5 consecutive days): Water: 3.2 mm/day, Sand: 1.6 mm/day.
evaporation rates and better control of boundary conditions is possible. 3 EVAPORATION AFTER SUNSET
A series of tests was carried out using water or sand contained in plastic pots. The pots are 195 mm in diameter with a cross-sectional area of 1/33 m2 and have a capacity of about 3.5 . The sides and bottoms of the pots were insulated with 4 layers of plastic bubble wrap. Evaporation was measured by weighing to 0.5 g. The specimens consisted either of water or lightly tamped fine sand. Temperatures of the sand were measured to 0.5C by means of two thermocouples taped to a wooden dowel and located on the vertical axis of the pot at the quarterpoints of depth of the sand. Water temperatures
were measured (with stirring) by a thermocouple probe. For each test, specimens were exposed outside after sunset (to allow equilibration) and measurements of temperature and mass were started at sunrise next morning (07.00) and continued past sunset at 17.00 until 22.00. Final measurements were made at sunrise next day. In all cases, the tests were made during the dry season (May to October) under cloudless skies. The results of a typical trio of tests are shown in Figure 2. Figure 2a shows that at sunset the temperature of the water or soil was about 2/3 of the maximum recorded value. Thus if soil or water heat plays a role in evaporation, there was considerable potential for evaporation to continue after sunset. Figure 2b shows that evaporation continued to occur until late into the night. The fact that considerably less evaporation occurred after 17.00
1117
Figure 2. Test to determine the occurrence of evaporation during night hours. Specimens were insulated with four layers of bubble-wrap plastic.
1118
than between 07.00 and 17.00 is because the heat converted to evaporation before sunset appears as evaporation in Figure 2b, but not as an increase in temperature in Figure 2a. After 17.00, as there was no wind, the only driver of evaporation according to equation (1a) was the stored heat which, regarding H as negligible, and G or WH as sources rather than sinks, represented S. Figure 1c shows the heat energy stored in the water or sand throughout the day and expressed as potential evaporation plotted against the actual cumulative evaporation. Because the lines were initially steeper than 1:1, the rate of increase of G, G, exceeded twice the rate of actual evaporation, E. As G approached equality with E, the slope of the lines flattened and after about 1.5 mm of evaporation, G 0. In other words, if the G versus E curve is horizontal the incoming solar energy is just enough to support the evaporation and any heat losses. After sunset, which corresponded with about 3 mm of evaporation from the water and the almost saturated sand, the steepness of the lines shows that G or WH were decreasing more than the increase E, i.e. the conversion of G or WH into evaporation was less than 100% efficient. For the water and wet sand, it was only about 22% efficient, probably because of inefficient conversion as well as losses through the inadequate insulation of the sides and bottoms of the pots. For the partly saturated sand, the efficiency of conversion was only 9%. 4 CONCLUSIONS
solar energy, for much of the day. This suggests that no evaporation will take place during this time (Figure 1). Observation shows that this is not so. During sunlight hours solar energy absorbed by soil or water is simultaneously partly converted into evaporation and partly stored as water or soil heat. Evaporation continues after sunset by conversion of stored heat until the temperature of the water or soil equilibrates with the ambient air temperature. Conversion of soil heat into evaporation becomes less efficient as the soil desaturates. REFERENCES
Blight, G.E. (2009). Solar heating of the soil and evaporation from a soil surface. Geotechnique Vol. 50, No. 4, 355463. Blight, G.E. (2005). Consequences of raising the height of a landfill in a water-deficient climate. Waste Management Vol. 25, 10211036. Blight, G.E. & Fourie, A.B. (2005). Experimental landfill caps for semi-arid and arid climates. Waste Management Research Vol. 23, 113125. Bowen, I.S. (1926). The ratio of heat losses by conduction and by evaporation from any water surface. Physics Review 27, No. 6, 779787. Calder, I.R. (1990). Evaporation in the uplands. Chichester: Wiley. Monteith, J.L. (1980). The development and extension of Penmans evaporation formula. In Applications of soil physics (ed. D. Hillel), 265275. New York: Academic Press. Monteith, J. & Unsworth, M. (1990). Principles of environmental physics. New York: Routledge. Munn, R.E. (1966). Descriptive micrometeorology. New York: Academic Press. Oke, T.R. (1978). Boundary layer climates. London: Methuen. Penman, H.L. (1963). Vegetation and hydrology, Technical Communication No. 53. Harpenden: Commonwealth Bureau of Soils, Commonwealth Agricultural Bureaux. Sellers, W.D. (1965). Physical climatology. Chicago: University of Chicago Press.
Most current methods of measuring or calculating evaporation from soil or water surfaces are based on the solar energy balance in which the major input energy is the net solar energy and the major output is heat to provide latent heat of vaporization, LE. It is assumed that solar energy consumed by heating the soil (G) is an energy loss and does not contribute to LE. The soil heat G may be as large as, or larger than the measured incoming vertical component of the
1119
A prototype soil column to calibrate numerical models accounting for soilatmosphere interaction
M. Caruso & C. Jommi
Structural Engineering Department, Politecnico di Milano, Italy
ABSTRACT: Many numerical models have been proposed to analyse water infiltration and evaporation across the topsoil. In most cases, in these analyses, soil-atmosphere interaction is accounted for by imposing inward and outward fluxes at the boundary of the discretised mesh, based on some hydrologic estimation. However, in practical applications, fluxes, which govern the water mass balance, are a required output of the analysis more than an input of the model. Fluxes can be obtained in a reliable way from the numerical analysis if meteorological conditions are properly translated into forcing conditions at the soil-atmosphere interface. The final aim of the experimental work presented here is to find a correct way to assign boundary conditions in numerical analyses, starting from the soil properties and the meteorological data. A prototype column, 120 cm high, was designed and equipped with moisture sensors and with a meteorological station, reproducing a monitoring setup which can be adopted in real scale applications. The design and some relevant data from a two month monitoring time period are presented and discussed. 1 INTRODUCTION air-pressure condition (Milly 1988, Hillel 1998). A more accurate description of the physical processes may be obtained by analysing a non-isothermal problem accounting for liquid and vapour transport (Philip & de Vries 1957, Milly 1984). Regardless the approach adopted, in most cases the infiltration problem and the evapotranspiration problem are tackled separately, following the many different theoretical or empirical formulations developed to analyse the inflow of water into the soil or the outflow from the soil surface. The different proposals to estimate rainfall infiltration provide a time-dependent rate of infiltration. The rate at which water enters the soil is expressed as a function of some model parameters which try to describe, in a summarised way, the different factors that are supposed to affect the effective infiltration rate: soil type, retention properties, initial water content, surface conditions vegetated or notrainfall characteristics, among others (Mishra et al. 2003). The Penmans (1948) proposal to relate potential evaporation with vapour pressure gradient between the soil surface and the ambient atmosphere seems to be the reference for the majority of literature proposals when analysing water outflow from the soil surface. Nonetheless, effective evaporation rate may fall below the potential evaporation rate when the water content decreases (Hillel 1998) or soil suction increases (Wilson et al. 1997, Gowing et al. 2006). The approaches adopting an infiltration model for rainy intervals, and a separate evaporation one for the non-rainy periods, in the end provide inward and
The time evolution of soil moisture distribution in the upper soil layers and the water balance, resulting from the transport processes taking place in the unsaturated zone, are a key factor in many applications in geotechnical engineering, groundwater hydrology, soil science and environmental protection. It is well known that soil moisture distribution is a controlling factor in the hydrological and mechanical processes responsible for slope instability and shallow landslides (e.g. Xue & Gavin 2008). Infiltration and evaporation influence shallow water-table fluctuations, groundwater recharge (Gowing et al. 2006), and contaminant transport towards the groundwater. Distribution and movement of water in the unsaturated zone near the soil surface is governed by the interaction with the atmosphere (Milly 1996). Infiltration is governed by the water made available at the ground surface due to rain intensity and its time distribution. Evaporation is governed by cyclic heating and cooling of the unsaturated zone, which may create vertical gradients in temperature, and differences between the soil and the atmosphere vapour pressures. Nonetheless, the transport and the retention characteristics of the upper soil layers also play a relevant role, which is seldom taken into account properly when hydrological models are adopted. In the study of the time evolution of moisture distribution by means of numerical models, two classes of approaches are adopted. Richards (1931) model is still the most widely accepted framework for flow analysis under isothermal and constant
1121
outward fluxes as forcing boundary conditions. They can reproduce quite well the water content evolution in the upper soil layers, if properly trained on some experimental data. Nonetheless, they may loose their predictive capabilities, when changes should intervene in the climatic conditions or in the soil properties. For this reason, a unique coupled thermo-hydraulic two-phase model, with forcing boundary conditions conveniently derived directly from meteorological data, may be more efficient in the analysis of cyclic sequences of rain and evaporation events over a long time period (e.g. Saito et al. 2006). To determine the best way in which the meteorological data may be translated into forcing boundary conditions, and to verify the predictive capabilities of the numerical model, accurate controlled data are mandatory (Angelaki et al. 2004). A prototype soil column apparatus was designed to this aim (Freeze & Banner 1970, Yang et al. 2004), equipped with a complete meteorological station, and with capacitive soil water sensors, which can be easily installed in situ when a real scale monitoring programme is planned. 2 2.1 EXPERIMENTAL SETUP The soil column apparatus
access tube top cap
acrylic top plate 7.7 threaded holes 20 probe soil porous stone acrylic base plate
section 1 section 2 section 3
section 4 20
section 5
probes PVC access tube iron support and stiffener O-rings
O-ring
expandable bung
10
20
30
40
50 cm
Figure 1. Schematic layout of the prototype column apparatus.
A schematic layout of the soil column apparatus is shown in Figure 1. Its components are all made of non-conductive materials (mainly acrylic), to avoid electrical interference with capacitive probes. The soil container is a 1200 mm high acrylic transparent tube, with a 5 mm thick wall and an internal diameter of 290 mm. The diameter was chosen to ensure that the sensors response were not influenced by the wall of the acrylic tube. Previous experimental data (and see also Paltineanu & Starr 1999, Schwank et al. 2006) had shown that the capacitive probes measurements are affected by the water content in a cylindrical volume of soil of about 160 mm diameter and 150 mm height, centred on the probe. The column is fastened at its bottom into an acrylic base plate, by insertion in a 5 mm wide annular hole sealed with a couple of o-rings, to avoid evaporation or outflow of liquid water. The base plate internally allows for the allocation of a porous stone, eventually connected to an external water level regulation device and pressure measurement system. At the centre of the base plate a 25 mm depth circular hole was designed, to ensure correct centred positioning of the capacitance probes PVC access tube. An expandable bung inserted at bottom of the PVC tube and an outside o-ring prevents the access tube from wetting. At its top, the access tube is sealed by a watertight cap, equipped with a cable gland, which can be easily opened for probes inspection. On the lateral walls, 1/8 BSPP holes, sealed
with threaded plugs, were designed to allow for the possible installation of lateral measuring devices. The prototype column was mounted in a dedicated iron caddy, designed to avoid relative displacements and possible damage during mounting and transportation of the apparatus. Four angular vertical stiffeners increase the resistance against lateral stress during tamping operations. The column and the caddy are fixed one to the other by four passing screws, inserted into a holed top plate perfectly fitting the acrylic tube. 2.2 The soil water measuring system The soil water measurement system is based on Sentek EnviroSMART capacitive probes, connected to a 16 bit SDI-12 Sentek A/D converter board, which manages up to 16 capacitors. Basically, the instrument provides the resonance frequency of an inductive-capacitive circuit, in which the soil participates as capacitive element, depending on its water content. The sensor accuracy is 0.06% of the readout resonance frequency (Sentek, personal communication). Sentek boards are designed to return a scaled frequency SF, in the range bounded by the resonance frequency in water and in air (Paltineanu & Starr 1999). A small-scale chamber had been previously designed to provide the reference data in water and in air, and to verify the calibration of the sensors inserted in a soil representative volume. Preliminary tests had supported the choice of a standard calibration relationship (Sentek 2001)
1122
10
10
in deriving the volumetric water content VW/ VTOT from the scaled frequency SF, provided the soil water was not excessively conductive. 2.3 The meteorological station
The meteorological station components, produced by LSI-Lastem, consist of a rain gauge, a thermo hygrometer with an anti-radiant shield, a combined wind speed-direction sensor and a global radiometer. The rain gauge has a resolution of 0.2 mm, with accuracy of 0.2 mm/min in the range between 0 and 1 mm/min, and of 1% in the range 110 mm/min. The thermo hygrometer has accuracy of 0.1C for temperature, and of 1.5% for humidity. The combined wind sensor has a speed resolution of 0.05 m/s and a resolution of 4 for direction, a speed accuracy of 0.1 m/s 1% and a direction accuracy of 1% of the respective readout values. The global radiometer has an uncertainty of 10% over a day measurement. 2.4 The data acquisition system
then hand compacted, as uniformly as possible, in layers of 10 cm thickness, following the undercompaction procedure proposed by Ladd (1978). The as compacted mean void ratio was e0 0.550, with porosity n0 0.35 and initial degree of saturation Sr0 0.50. The meteorological station was installed at the same height of the soil-atmosphere interface, thirty centimetres far from the soil column. Before the beginning of the monitoring period, an initial set-up time of two weeks was spent to verify if the probes installation depth was correct to monitor significant water content changes. The final choice for the probes depth is indicated in Figure 1. The prototype column was set up at the Structural Engineering Department of the Politecnico di Milano, and left at environmental atmospheric conditions, starting from the end of October 2009. 3.2 Monitoring data
A new data acquisition system was implemented for the prototype apparatus. Communication of the Sentek board to a PC was established by means of a SDI-12 to RS 232 interface. A low-noise power supply device was also designed. The system is managed by dedicated software, developed in the LabVIEW environment. The software allows for registration and real time visualisation of the whole set of measurements. The weather station is managed with a dedicated LSI-Lastem 16 bit data logger system for environmental applications, with an integrated d-c power supply. Setting up and collection of the data are made possible by a RS232 interface with a proprietary protocol and software. 3 EXPERIMENTAL RESULTS
3.1 Soil samples A sandy silt, whose main properties are reported in Table 1, was initially air dried and sieved with a 2 mm sieve. The soil was mixed with tap water, left to homogenise in a humid chamber for 24 hours, and
Table 1. % sand 26.3 Soil properties (AASHTO standard). % silt 70.4 % clay 3.3 Specific gravity 2.72 Liquid limit 0.33 Plastic limit 0.20
The data collected in the first two months, from 20/10/2009 to 20/12/2009 are reported in Figure 2. Soil water content was registered every 5 minutes, while the weather parameters were collected every 20 minutes. In the figure, the probes data are already processed to give the volumetric water content time evolution at the various depths. Inspection of the data of the first 15 days shows little oscillations of the probe measurements, which appear to be perfectly synchronous, except for a very small delay, with temperature daily oscillations, reflecting the temperature effect on the soilwater measurement system. It is worth noting that these oscillations tend to disappear after the second rainy event, when the temperature decreases below 10C. All probes are placed at the same distance from the outer wall, and hence they should be all influenced by the external temperature in a similar way. In the same period, after the first rainy event, the soil water content decreases at an almost constant rate in the first 30 cm of soil, and slightly increases at depths below 40 cm from the soil surface. The data collected allow for distinguishing clearly the response of the soil to the forcing meteorological conditions. The progression of infiltration fronts following rainy events may be appreciated by the sudden increase of the water content, with a time lag obviously increasing from the top to the bottom of the soil column. Equally, the evaporation rate tends to decrease with depth, and can hardly be appreciated below the first 50 cm depth. Immediately after rain stops, in the upper portions of the soil, the water content starts to decrease at a rate which seems to be independent from the actual temperature, relative humidity or wind velocity. The datum seems to suggest that
1123
0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 20/10/09 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 20/10/09 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 20/10/09 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 20/10/09 0.35 0.30 0.25 0.20 0.15 0.10 0.05 0.00 20/10/09 16
(cm3/cm3)
3.3
probe 1 - z = 7.7 cm 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09
Water mass balance
probe 2 - z = 17.7 cm 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09
probe 3 - z = 27.7 cm 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09
probe 4 - z = 47.7 cm 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09
probe 5 - z = 57.7 cm 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09 250 200 150 100 50 0 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09 200 160 120 80 40 0 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09 4 3 2 1 0 30/10/09 09/11/09 19/11/09 date 29/11/09 09/12/09 19/12/09 Cumulated rain (mm)
rain (mm/h)
12 8 4 0 20/10/09 20 10 0 -10 20/10/09 100 80 60 40 20 0 20/10/09
To provide deeper insight into the soil water mass balance, three rainy events are analysed in more detail. The first one has a cumulated rain of 43.8 mm in about 11 hours (Fig. 3), the second one 36 mm over a 14 hour period (Fig. 4) and the last one 47.6 mm over 22 hours (Fig. 5). In each figure, the probe readings and the rainfall event description (intensity and cumulated values) are reported, together with diagrams which help in the analysis of water balance. To perform the balance, the column was subdivided into five sections, starting from the top soil-atmosphere interface (Fig. 1). The volumetric water content measured by each probe was assumed to be representative of the whole section. The incremental and the cumulative mass variations were then calculated starting from the top section and adding progressively the underlying layers. The data referring to the first event (Fig. 3) clearly show the advancing front starting from the first section and reaching the third section after few hours. Sections 4 and 5, below the first 30 cm, are not affected directly by the rain event. The amount of water which can be stored in the upper layers correspond to the difference between the water content at the beginning of rain and the water content corresponding to saturation. Once saturation is reached, the probe reading remains constant until rain stops. Constant readings in the upper soil layers at increasing water content at greater depths denote an infiltration process continuing downwards. The cumulative mass variations in the first sections show that all the rain infiltrates into the soil, due to the initial low water content.
0.35 0.30
(cm3/cm3)
(cm3/cm3)
(cm3/cm3)
(cm3/cm3)
(cm3/cm3)
relative humidity (%)
wind velocity (m/s)
Solar radiation (W/m2)
temperature (C)
probe 2
0.25 0.20 0.15 0.10 0.05 probe 1
probe 3
probe 5
probe 4
Figure 2.
Experimental data: 20/10/200920/12/2009.
0.00 21/10/09 16 rain (mm/h) 12 8 4 0 21/10/09 80

Section 1 Section 1+2 rain
22/10/09 date
23/10/09 60 45 30 15
the dominating process just after a rainy event is downward infiltration, at least for a short time interval. Afterwards, the water content variation tends to stabilise to a constant decreasing rate, which is interpreted as the effect of evaporation towards the upper surface. Only the first 50 cm topsoil seem to contribute significantly to the evaporation process, although a slight decrease of the water content at a depth of 57.7 cm begins to be appreciated when the actual saturation degree is close to one. Obviously, this must be considered only a preliminary and partial conclusion, as the data presented refer to a cold season, with pretty low solar radiation and temperature.
22/10/09 date
0 23/10/09 3
Section 1+2+3 Section 1+2 Section 1
mw (g)
40
Section 1+2+3
1.5
-40 21/10/09
22/10/09 date
23/10/09
Figure 3.
Rainfall event 22/10/2009.
1124
cumulative mass variation (kg)
cumulated (mm)
0.35
(cm3/cm3)
probe 1 probe 2
0.30
probe 3 probe 4 probe 5
0.25
0.20 02/11/09 8 rain (mm/h) 6 4 2 0 02/11/09 80

Section 1 Section 1+2
03/11/09 date
04/11/09 cumulated (mm) 40 30 20 10
03/11/09 date
rain Section 1+2+3 Section 1 Section 1+2
0 04/11/09 3
mw (g)
40
Section 1+2+3
1.5
cumulative cumulated mass mass variation variation (kg) (g)
a clear consequence of the role played by the initial soil water distribution. The influence of the initial water content distribution is even more evident in the third rainy event (Fig. 5), characterised by a cumulated rain similar to the first one, albeit distributed over a longer time period. The infiltration front reaches quickly the last monitored section, as the upper layers were close to saturation already at the beginning of the rainfall. Nonetheless, only one third of the total amount of water available from rainfall infiltrates into the soil, as the peak in the cumulative water mass shows, just at the end of the rainfall. After this rainy event, water ponding was clearly evident at the top surface. 3.4 Soil response to snowfall
-40 02/11/09 03/11/09 date 04/11/09
Figure 4.
0.35 (cm3/cm3)
Rainfall event 02/11/2009.

probe 4
0.30 probe 5 0.25 probe 3 probe 2 probe 1
0.20 29/11/09 16 rain (mm/h) 12 8 4 0 29/11/09 80

Section 1 Section 1+2
30/11/09
date
01/12/09
02/12/09 60 45 30 15
30/11/09
date
01/12/09
0 02/12/09 4
On the 18th of December 2009, a snowfall started at about 6.00 pm (Fig. 6). The temperature fell below zero and remained below zero for more than two days. A visual inspection of the soil column revealed the formation of a thin layer of ice and snow on the top of the soil column, which progressively increased with snow accumulation up to 20 cm. After few hours from the beginning of the snowfall, the first probe reading started to decrease, followed in turn after some delay by the other probes. All the probes readings went on decreasing monotonically for the following two days. The data show that the water content in the upper part of the column began to increase only at the beginning of a rainfall in the following days. Water content decrease in the upper part was rather surprising, although significant infiltration
0.35 0.30
(cm3/cm3)
Section 1+2+3+4+5 rain
mw (g)
40
Section 1+2+3
Section 1+2+3
0
Section 1+2 Section 1
cumulative mass variation (kg) cumulated mass variation (g)
cumulated (mm)
0.25 0.20 0.15 0.10 15/12/09
probe 5
probe 4 probe 3 probe 2 probe 1
-40 29/11/09
30/11/09
date
01/12/09
02/12/09
temperature (C)
8 4 0 -4 -8 15/12/09 17/12/09
humidity
Infiltration following the second rainy event analysed (02/11/09, Fig. 4) reaches more rapidly the third section and starts to affect also the lower section 4. The amount of water stored into the soil layers is still measured by the water content variation in the first three sections. In fact, when the fourth probe reading starts to increase (at the beginning of the day after the rain), the water content is already decreasing in the upper layers, denoting that a downward redistribution process is taking place. The amount of water that infiltrates into the soil is less than that available from rainfall, even if the cumulated rain is lower than in the previous event and its duration is longer, which is
80 70 60 temperature 50 40
19/12/09 21/12/09 date
23/12/09
Cumulated mass variation (kg)
2000
Section 1
0 -2000 -4000 -6000 15/12/09
Section 1+2
Section 1+2+3+4+5 Section 1+2+3
17/12/09
19/12/09 21/12/09 date

start snowing
23/12/09
Figure 6. Response of the soil column after a snowfall.
1125
relative humidity (%)
Figure 5.
Rainfall event 29/1130/12/2009.
17/12/09
19/12/09 21/12/09 date
23/12/09 90
was not expected to take place, due to the thin ice layer on the top of the soil. Decreasing temperature and almost constant relative humidity, recorded at the soil-atmosphere boundary in these days, do not seem to justify an increasing rate of evaporation. The observed soil water content evolution may be better interpreted as the effect of water redistribution towards deeper layers driven by gravity, moving the lower part of the column towards saturation, while the upwards flux towards the atmosphere is temporarily reduced by the presence of the top ice layer. The interpretation is somehow confirmed by visual inspection, revealing the formation of a water table at the bottom of the column, which, for the whole duration of the monitoring period, was kept water tight. 4 FINAL REMARKS
Francesco Avanzi and Nicola Paccanelli for their contribution to the experimental investigation. REFERENCES
Angelaki, A., Sakellariou-Makrantonaki, M. & Tzimopoulos, C. 2004. Laboratory experiments and estimation of cumulative infiltration and sorptivity. Water, Air & Soil Pollution: Focus 4(45): 241251. Freeze, R.A. & Banner, J. 1970. The Mechanism of Natural Ground-Water Recharge and Discharge 2. Laboratory Column Experiments and Field Measurements. Water Resour. Res. 6(1): 138155. Gowing, J.W., Konukcu, F. & Rose, D.A. 2006. Evaporative flux from a shallow watertable: the influence of a vapour-liquid phase transition. J. of Hydrology 321(14): 7789. Hillel, D. 1998. Environmental Soil Physics, London: Academic Press. Ladd, R.S. 1978. Preparing test specimens using undercompaction. Geotechnical Testing Journal 1 (1): 1623. Mein, R.G. & Larson, C.L. 1973. Modeling infiltration during a steady rain. Wat. Resour. Res. 9 (2): 384394. Milly, P.C.D. 1984. A simulation analysis of thermal effects on evaporation from soil. Water Resour. Res. 20(8): 10871098. Milly, P.C.D. 1988. Advances in modeling of water in unsaturated zone. Transp. in Porous Media 3: 491514. Milly, P.C.D. 1996. Effects on thermal vapour diffusion on seasonal dynamics of water in the unsaturated zone. Wat. Resour. Res. 32 (3): 509518. Mishra, S.K., Tyagi, V. & Singh, V.P. 2003. Comparison of infiltration models. Hydrol. Proc. 17: 26292652. Paltineanu, I.C. & Starr, J.L. 1997. Real-time soil water dynamics using multisensor capacitance probes: laboratory calibration. Soil Sci. Soc. of Am. J. 61: 15761585. Penman, H.L. 1948. Natural evaporation from open water, bare soil and grass. Proc. Roy. Soc. London A 194: 120145. Philip, J.R. & de Vries, D.A. 1957. Moisture movement in porous materials under temperature gradients. EOS Trans. AGU 38: 222232. Richards, L.A. 1931. Capillary conduction of liquids through porous medium. Physics 1: 318333. Saito, H., Simunek, J. & Mohanty, B.P. 2006. Numerical analysis of coupled water, vapor, and heat transport in the vadose zone. Vadose Zone Jour. 5: 784800. Schwank, M., Green, T.R., Matzler, C., Hansruedi, B. & Fluhler, H. 2006. Laboratory characterization of a commercial capacitance sensor for estimating permittivity and inferring soil water content. Vadose Zone Jour. 5: 10481064. Sentek 2001. Calibration of Sentek Pty. Ltd. Soil Moisture Sensors. Sentek, Stepney S.A., Australia. Wilson, G.W., Fredlund, D.G. & Barbour, S.L. 1997. The effect of soil suction on evaporative fluxes from soil surface Can. Geotech. J. 34 (2): 145155. Yang, H., Rahardjo, H., Wibawa, B. & Leong, E.-C. 2004. A soil column apparatus for laboratory infiltration study. Geot. Testing J. 27(4): 347355. Xue, J. & Gavin, C. 2008. Effect of rainfall intensity on infiltration into partly saturated slopes. Geotech. Geol. Eng. 26: 199209.
The prototype soil column apparatus, developed for the analysis of soil-atmosphere interaction, allows for tracking rain infiltration, evaporation outflow and moisture movements inside the soil. Soil water balance may be provided by careful post processing of the soil water content data. The soil column was instrumented with capacitive probes only, in view of the field applications, where the installation of other measuring instruments would be more difficult or less significant. Although suction probes would provide a more complete picture of the hydraulic state of the soil, they were not installed, as their response time is much slower than that of water content probes, whenever installed in soils finer than sands. To infer suction evolution, it is worth passing through the water retention curves, which can be usefully determined in the laboratory on representative soil samples. From the results of the first two months monitoring period, the water saturation distribution at the beginning of rainfall events clearly emerges as the key factor in water infiltration, given the soil properties and the forcing boundary conditions. Internal redistribution takes place continuously and affects increasing depths as the average water content increases. A snowfall event seems to reduce temporarily upward evaporation, promoting a faster downward redistribution of the amount of water already stored in the soil, driven by gravity. Subsequent infiltration due to snow melting can take place rather easily. ACKNOWLEDGEMENTS The authors wish to thank Dr. Cristina Rulli (DIIAR, Politecnico di Milano) for lending the meteorological station, Angelo Lunghi (LSILastem, s.r.l) for the helpful support in its setup,
1126
Use of satellite observations for estimating the degree of saturation of superficial soils
L. Montrasio, R. Valentino & C. Quintavalla
University of Parma, Italy
ABSTRACT: For the development of alert systems for soil slip occurrence, it is important to evaluate the degree of saturation of shallow soils (Sr) over wide areas. Taking into account the possibility to estimate spatial and temporal variation of soil moisture using remote sensing techniques, a possible correlation between Sr and the daily output of a sequential data assimilation system called ACHAB (Assimilation Code for HeAt and moisture Balance) has been studied. ACHAB is based on integrated use of remotely sensed Land Surface Temperature (LST) and common data on meteorological forcing such as air temperature, wind-speed and incident solar radiation. The aim of this study is to understand if it is possible to use ACHAB output (a daily value of evaporative fraction for the whole Italian territory) to define the parameter Sr that could be introduced in a simplified model for the description of soil slip triggering mechanisms on territorial scale. 1 INTRODUCTION tive) and outflowing (negative) water mass from the soil-atmosphere interface. The water balance can be estimated using experimental techniques or numerical modeling. The spatial and temporal scale of the problem to be analyzed usually suggests the most appropriate choice between the two approaches. It is generally accepted that areas of limited extension can be effectively monitored by experimental devices for a limited period of time. On the other hand, the estimation of the overall water balance over wide areas and with reference to long periods of time is usually tackled by numerical modeling of infiltration and evapo-transpiration processes (Caruso 2007). The choice of the numerical model implies the definition of some soil properties (water retention curve and permeability function) and some contour conditions resulting from climate characteristics, use and cover typology of the soil and groundwater aquifer depth (Scanlon et al. 2002). In the SLIP model, while the geometric features of the slope and the physical characteristics of the soil can be reasonably considered unchangeable for a certain slope, the degree of saturation of the soil has to be evaluated changing in consequence of weather conditions and rainfalls. The shear strength parameters are, in turn, strongly influenced by the degree of saturation of the soil. Currently in the SLIP model Sr is considered varying with seasonal trend, according to rainfall conditions. Given the possibility to determine spatial and temporal variation of soil moisture using remote sensing techniques, it was possible to correlate the degree of saturation of the soil to the daily
Among different types of landslides, those involving small sections of superficial soils, often called soil slips, are particularly dangerous because of rapid formation, the difficulty in prediction and high density in distribution on a susceptible territory. The economic and social impacts of this kind of events have recently lead to the development of a simplified and physically based stability model, called SLIP, able to foresee the occurrence of soil slips (Montrasio 2000, Montrasio & Valentino 2008). The model has been applied on local scale to some rainfall-induced shallow landslides in the Italian territory (Montrasio & Valentino 2007, 2008, Montrasio et al. 2009) and has been implemented in a platform for a real-time territory control (Montrasio & Valentino 2009, Valentino et al. 2009). The model defines the safety factor of potentially unstable slopes taking into account the geometric characteristics of the slope, the geotechnical properties and the shear strength parameters of involved soils. Moreover, this method enables a direct correlation between the safety factor and rainfall intensities through a simplified water-flow model. For the improvement of the SLIP model and the development of alert systems for soil slip occurrence, it results important to evaluate and introduce the degree of saturation of shallow soils (Sr) as a variable for the definition of the safety factor. The variations in soil water content in the upper soil layers are traditionally analyzed through a water balance equation. This balance is the algebraic sum, over a significant period, of inflowing (posi-
1127
output of a sequential data assimilation system called ACHAB (Assimilation Code for HeAt and moisture Balance). This system is based on the integrated use of remotely sensed Land Surface Temperature (LST), from satellite acquisition on a spatial resolution of 3 km2, and common data on meteorological forcing such as air temperature, wind-speed and incident solar radiation (Boni et al. 2001, Caparrini et al. 2004). The model has been tested in two different versions: dual-source and API (Antecedent Precipitation Index). The output considered in this analysis is the Evaporative Fraction (EF), which represents the ratio between the energy consumed for evapo-transpiration and the net available energy and is related to the wetting history and dry-down in shallow soils. This study attempts to correlate an energy balance parameter, such as EF, to an indicator of soil moisture, such as the degree of saturation through data comparison. The aim of the study reported in this paper is to understand if it is possible to use ACHAB output (a daily output produced for the whole Italian territory) to define the parameter Sr that has to be introduced in the model for triggering mechanisms of soil slips on a territorial scale. 2 THE ACHAB MODEL
Figure 1. Example of evaporative fraction map for Italian territory.
gives only one result, the evaporative fraction is estimated using the API as a constrain. Figure 1 shows an example of evaporative fraction map for the Italian territory, obtained as the output of ACHAB-API model. 2.1 Surface energy balance and evaporative fraction
The ACHAB model is a sequential data assimilation system, for the estimation of hydrological components related to land surface. The assimilation scheme allows the simultaneous collection of determinant parameters of land surface water and energy balance (turbulent transfer coefficient for heat fluxes, evaporative fraction, indices of soil moisture) with a very limited requirement of ancillary data and empirical assumptions. In addition to the system-state observations (Land Surface TemperatureLST), the assimilation system requires common data on meteorological forcing such as air temperature, wind-speed and incident solar radiation (Caparrini et al. 2006). Two different versions of the model has been developed and tested: 1. ACHAB-2S: different contributions of soil and vegetation to the radiometric temperature are explicitly taken into account through dual-source formulation based on satellite vegetation indices; 2. ACHAB-API: the API equation is introduced into the assimilation scheme in order to model the soil moisture dynamics in a simplified way. For the dual-source version of the model, which separates energy balances at the surface and gives two different results, the output parameter of evaporative fraction is evaluated for bare soils (EFs) and vegetated areas (EFv). For the API version, which
The Evaporative Fraction (EF) is defined as the ratio between the latent heat flux and the difference between net radiation and ground flux or, equivalently, the ratio between the latent heat flux and the sum of sensible and latent heat fluxes (Sini 2005): EF
ET ET Rn G ET H
(1)
The latent heat flux (ET) is the energy, per time and surface units, exchanged between the earths surface and the atmosphere (ET is evapo-transpiration). The latent heat is the energy that bonds water molecules in liquid phase and is released during a change of state from liquid into vapor. The energy released during the passage from liquid to vapor state does not generate a temperature increase but represents the potential energy of water vapor molecules. H is the sensible heat flux, i.e., the energy which can be evaluated by measuring air temperature with a thermometer. Since EF can be calculated over large areas using satellite imagery, it is a suitable indicator for the description of soil moisture conditions on a regional scale, while the traditional methods of
1128
study of soil moisture storage in the unsaturated zone (experimental techniques and numerical modeling) present many problems. 3 COMPARISON BETWEEN EF VALUES AND SOIL MOISTURE DATA
In order to determine a possible correlation between the EF obtained from the ACHAB model and the effective soil moisture, EF and soil moisture data on a limited area where compared. The considered study area is in San Pietro Capofiume (Bologna), located in the Emilia Romagna Region. EF data, related to this area, have been evaluated from national scale maps produced every day by the ACHAB model for the Italian territory. Using these data, an analysis of the temporal trend of the EF, for the period between March 2005 and January 2006, was carried out. The analysis revealed that ACHAB-2S does not capture the seasonal variation in soil moisture, as the EF values remain on high even during summer (Fig. 2). This behavior is especially highlighted in EF data for vegetation (Fig. 3). By analyzing the EF data from the API version of the model a larger variability in the obtained values (probably related to rainfall intensities) is observed, but even in this case a seasonal trend cannot be seen (Fig. 4). The soil moisture data used for the comparison, relative to the study area of San Pietro Capofiume,
Figure 4. EF data from ACHAB_API over San Pietro Capofiume site.
are obtained from field measurements of water content in the unsaturated zone carried out by the Regional Agency for Environmental Protection of Emilia Romagna Region. The soil moisture was monitored with Time Domain Reflectometry (TDR) devices, which were installed at seven different depths in the soil between 10 and 180 cm, corresponding to pedological profile layers. The experimental measurements consisted of daily volumetric water content () at each depth and refer to a period of almost three years, between September 2004 and April 2007. Over the same period rainfall and temperature data are also available. The degree of saturation Sr for each depth has been then calculated on the basis of the measured volumetric water content. It is worth remember that the degree of saturation defines the water volume percentage (Vw) in the volume of void space in the soil (Vv) and can range from 0, in completely dry soil, to 1 in saturated soil. For the exact evaluation of soil saturation the parameter of effective saturation (Se) should be used. Effective saturation is usually calculated on the basis of volumetric water content by using the well known equation: Se
r s r
(2)
Figure 2. EF data for bare soil over San Pietro Capofiume site.
where is the volumetric water content, r is the residual water content, which represents the adsorbed water, while s is the saturation water content, which represents the maximum volumetric water content. In the present work r is assumed equal to zero. Considering the negligible effect of such approximation, the degree of saturation (Sr) is considered equal to the effective saturation (Se): Sr Se
s
(3)
Figure 3. EF data for vegetated soil over San Pietro Capofiume site.
The degree of saturation Sr is calculated through Equation 3 for each depth of the sample site. Moreover, an empirical correlation has been introduced to obtain, from the EF data, the
1129
related values of degree of saturation (Ahmad & Bastiaanssen 2003, Scott et al. 2003), in order to compare them with the measured values of Sr. The relationship is the following:
EF a exp b s
the comparisons in Figures 56. Comparisons between estimated and measured values of Sr at greater depths in the soil have also been considered, but they are not reported here since there was not a good correlation between the data. 4 INTRODUCTION OF ACHAB OUTPUT IN THE SLIP MODEL
(4)
where a and b are calibration parameters. In this study the parameter a is set to 1, as reported in Scott et al. (2003), for normalized soil moisture, while b assumes a value equal to 0.94, according to a calibration procedure carried out over the available data. In Figures 56 one can notice that the degree of saturation, which is estimated respectively from EFs and EF_API, has not the seasonal trend that characterizes experimental data. The greatest differences between the estimated values of Sr and measured values of Sr occur in June and July. Furthermore, the estimated degree of saturation has shown a behavior that can be correlated only with the trend of the soil moisture in the shallow soil layer (at a depth of 0.25 m from the ground level), as one can observe from
Figure 5. Comparison between estimated Sr (from EFs data) and measured Sr at depths of 10 and 25 cm for San Pietro Capofiume site.
In order to evaluate whether the estimated values of the degree of saturation of the soil, derived from ACHAB system, could be used in the model for characterizing the triggering mechanism of shallow landslide (Montrasio 2000, Montrasio & Valentino 2008), those data have been introduced in the SLIP model. The SLIP model has been previously applied to 45 sites of the Italian Apennines in the Province of Reggio Emilia, where many rainfall-induced shallow landslides occurred in April 2005. In this model the degree of saturation currently takes into account the seasonal trend and assumes typical values that can range from 0.6, during the summer, to 0.9 in winter, within the limits of the superficial layers of the soil (depths of 0.71 m from the ground level) and especially in northern Italy regions. This trend is confirmed by many experimental observations (Quintavalla, 2006), even by those realized in the San Pietro Capofiume site (Fig. 7). The SLIP model, with the simplified trend of the input parameter Sr, allows to obtain the behavior of the safety factor as a function of time and captures both the instability situation, i.e. Fs 1 in the expected date, and the stability conditions (Fs 1) in the remainder of the time. As an example, in Figure 8 is shown the result of the SLIP model applied to the site of Baiso (Emilia Romagna Apennine, Northern Italy), where a soil slip occurred on 10 April 2005. The values of Sr with seasonal trend in the SLIP model have been replaced by estimated values
Figure 6. Comparison between estimated Sr (from EF_ API data) and measured Sr at depths of 10 and 25 cm for San Pietro Capofiume site.
Figure 7. Comparison between experimental observations of Sr at different soil depth in the San Pietro Capofiume site and Sr considered changing with seasonal trend.
1130
Figure 8. Behavior of the safety factor as a function of time in the site of Baiso, with the simplified trend of Sr .
Figure 10. Trend of the safety factor as a function of time in the site of Canossa. Is highlighted the introduction of Sr values obtained from EFs data (a) and from EF_API data (b).
Figure 9. Trend of the safety factor as a function of time, in the site of Baiso. Is highlighted the introduction of Sr values obtained from EFs data (a) and from EF_API data (b).
of Sr, obtained from the EF data produced by ACHAB. The values of Sr have been estimated for the examined time interval over two representative sites where the SLIP model has been applied: Baiso and Canossa. Through this modification, it is possible to observe that the trend of the safety factor for given test sites is significantly influenced by the change in values of the degree of saturation: in particular Fs reaches values related to instability conditions (Fs 1) even in periods that are not associated with real landslide phenomena. This seems to be due to the fact that the values obtained by ACHAB model do not capture a seasonal trend. Therefore, as it can be noticed by the graphs in Figure 9, in the Baiso site, after the introduction of Sr values derived from soil evaporative fraction data (Fig. 9a) and Sr values derived from EF_API
data (Fig. 9b), the safety factor reaches values lower than one not only in April, when the historical event of soil slip happened, but also in October. A possible explanation of this fact is that, in the month of October, Sr is indeed relatively high as a consequence of rainfalls and is not affected by the strong drying of the soil occurred during the summer, like it was previously supposed. Even analyzing the graphs for the study area of Canossa (Fig. 10a,b), it is clearly visible that the safety factor detects instability situations not only in April 2005, but even in November and December of the same year. In these months, the values of the estimated degree of saturation are higher than those hypothesized for the seasonal trend shown in Figure 7.
After the analyses carried out in the previous section, it is possible to state that the ACHAB model needs some improvements to make its outputs applicable to the SLIP model. The main problems detected by the comparisons (lack of seasonal trend in the EF values, differences between estimated and measured or modeled values of Sr) are probably related to the following facts: the ACHAB model, due to its estimation of energy balance at the surface, calculates values
1131
of EF correlated to the soil moisture of the shallow layer of the soil, up to near 0.25 m from ground level; EF is high during the summer, when the surface heating is at its peak. This happens because at this time seasonal precipitations moisten the surface and cooling shift toward the more efficient latent heat mechanism. So the sensible heat decreases while EF value increases; the ACHAB model provides EF estimations with a spatial resolution of 3 km2, and with no correlation to the soil type, while the comparisons have been made with punctual field measurements and with modeled data dependent on soil porosity. An important feature about the trend of the degree of saturation has emerged by analyzing the available data: the seasonal trend considered in the SLIP model catches properly the soil saturation because in the model the rainfalls (basic variable to evaluate soil moisture) are already taken into account as input parameters while the variable land surface temperature (that makes Sr vary seasonally) is not considered as an input parameter of the model and is introduced by means of the trend of the degree of saturation.
gratitude to Dr. R. Rudari and Dr. F. Tomei for their cooperation. REFERENCES
Ahmad, M. & Bastiaanssen, W. 2003. Retrieving soil moisture storage in the unsaturated zone using satellite imagery and bi-annual phreatic surface fluctuations. Irrigation and Drainage Systems, 17: 141161. Boni, G., Castelli, F. & Entekabi, D. 2001. Sampling strategies and assimilation of ground temperature for the estimation of surface energy balance components. IEEE Transactions on Geoscience and Remote Sensing, 39:165172. Caparrini, F., Castelli, F. & Entekabi, D. 2004. Variational estimation of soil and vegetation turbulent transfer and heat flux parameters from sequences of multisensory imagery. Water Resources Research, 40: W12515. Caparrini, F., Castelli, F., Entekabi, D. & Campo, L. 2006. Sequential assimilation of land temperature for monitoring surface components of the hydrologic cycle. EUMETSAT Meteorological Satellite Conference, Helsinki, June 2006. Caruso, M. 2007. Una metodologia teorica, sperimentale e numerica per la previsione del bilancio idrico nei terreni superficiali non saturi. PhD Thesis, University of Parma. Montrasio, L. 2000. Stability analysis of soil slip. Proc. of Int. Conf. Risk 2000, C.A. Brebbia (eds), Wit Press, Southampton. Montrasio, L. & Valentino, R. 2007. Experimental analysis and modeling of shallow landslides. Landslides, 4, 33: 291296. Montrasio, L. & Valentino, R. 2008. A model for triggering mechanisms of shallow landslide. Natural Hazards and Earth System Sciences, 8: 11491159. Montrasio, L. & Valentino, R. 2009. X-SLIP: un modello per linnesco di frane indotte da piogge, fisicamente basato, implementato su piattaforma Experience. Annual Meeting of Italian Geotechnical Researchers, IARG 2009, Roma, Italy, 911 September 2009. Montrasio, L., Valentino, R. & Losi, G.L. 2009, Rainfallinduced shallow landslides: a model for the triggering mechanism of some case studies in Northern Italy. Landslides, 6: 241251. Quintavalla, C. 2006. Valutazione del grado di saturazione in terreni superficiali per lanalisi di stabilit di pendii in terra. Degree Thesis, University of Parma. Scanlon, B.R., Healy, R.W. & Cook, P.G. 2002. Choosing appropriate techniques for quantifying groundwater recharge. Hydrogeology Journal, 10: 1839. Scott, C., Bastiaanssen, W. & Ahmad, M. 2003, Mapping root zone soil moisture using remotely sensed optical imagery. Journal of Irrigation and Drainage Engineering, ASCE, 129 (5): 326335. Sini, F. 2005, Satellite data assimilation for the estimation of surface energy fluxes. PhD Thesis, University of Basilicata. Valentino, R., Montrasio, L. & Losi G.L. 2009, Modellazione dellinnesco di frane superficiali indotte da piogge: dalla scala locale a quella regionale, Annual Meeting of Italian Geotechnical Researchers, IARG 2009, Roma, Italy, 911 September 2009.
CONCLUSIONS
Due to the need of rapid and reliable techniques for the estimation of the degree of saturation of soils, which are involved in rainfall-induced landslides, a new method of soil moisture detection using a remote sensing technique (the ACHAB system) has been analyzed, in order to evaluate the possibility of inserting the calculated values in the model for characterizing the triggering mechanism of shallow landslide (the SLIP model). To enable the use of ACHAB processing for a more accurate estimation of the degree of saturation, the ACHAB model should be improved to provide an estimation with a seasonal trend, in order to obtain a higher variability in EF values. Developing these changes, the ACHAB model results would be useful to determine the Sr parameter and to take advantage of it in the SLIP model, especially in its implementation on a platform for the real-time territory control. ACKNOWLEDGMENTS EF data were provided by the CIMA Foundation of Savona, while experimental data of the degree of saturation were provided by ARPASIM Emilia Romagna (Regional Agency for Environmental ProtectionHydro-Meteorological Service). The authors would like to express their
1132
Study of stratification resistant to salt damage

S. Nomura, K. Kawai, S. Kanazawa & A. Iizuka
S. Ohno
Kajima Corporation, Tokyo, Japan
P. Thirapong
Tokyo Institute of Technology, Tokyo, Japan
S. Tachibana
Saitama University, Saitama, Japan
ABSTRACT: Desertification due to salt damage has recently become a serious problem throughout the world. Salt damage is a phenomenon where salt in the groundwater gets concentrated at the ground surface and kills vegetation. In this study salt damage due to disequilibrium of water balance was simulated with unsaturated soil/water coupled simulation code, as applied the material transport equation. Evaporation and precipitation cycles were applied to a model of ground that contained salty groundwater in its deep region. This allowed the successful reproduction of the phenomenon of salt accumulation on the ground surface. Additionally, identical climatic conditions were applied to model grounds consisting of two types of soil, different in terms of their soil-water retention characteristics. It was found that two methods of overlay and insertion of a layer of coarse-grade material, known as mulching, are effective in the prevention of salt accumulation. 1 INTRODUCTION model for unsaturated soil proposed by Karube & Kawai (2001) by applying the method introduced by Kawai et al. (2007). With the formulated model, however, it is difficult to identify the required soil parameters, and, as a result of sensitive dependence of the model on parameters, the numerical results could be unstable. In this study, we simulate salt accumulation using the constitutive model developed by Ohno et al. (2007). This model enables an easy identification of soil parameters and reproduction of test results. Furthermore, strategies for preparing a suitable ground, prepared from
Today, the issue of desertification has become a serious environmental problem of global proportion. Deserts are expanding at a rate of 60,000 km2 per year. One major cause of desertification is salt damage. Salt accumulation occurring near the ground surface, due to man-made or natural causes, degrades the land and inhibits plant growth. In the northeastern region of Thailand, traditional sun-drying salt manufacture is actively conducted using high-salinity groundwater taken from deep underground. In recent years, however, salt has spontaneously risen up to the ground surface during long-time precipitation and evaporation cycles, seriously affecting agricultural crops. Figure 1 is a photo of a river bed in northeastern Thailand. Salt deposits are present in the white areas on the ground. Fujikawa et al. (2006) incorporated a material transport equation into the conventional framework of unsaturated soil/water coupled finite element analysis and formulated an advectiondispersion problem taking ground deformation into consideration. They successfully expressed salt accumulation through alternate wetting and drying cycles. Fujikawa et al. formulated a constitutive
Figure 1. Naturally occurring salt damage.
1133
different soil materials and resistant to salt damage, are proposed. 2 FORMULATION OF UNSATURATED SOIL/WATER COUPLED PROBLEM CONSIDERING MATERIAL TRANSPORT
Using the effective stress, the yield function is expressed as follows: f , , vp MD ln
p q D vp 0 psat p
(4)
Salt damage becomes a problem in arid lowprecipitation regions where the ground is often unsaturated. Simulation of such areas requires unsaturated soil mechanics. The mathematical model used in this paper is described below. 2.1 Constitutive model for unsaturated soil
1 3 p : 1, q s : s , s p1 A : , 3 2 1 A I 11 3 I: Fourth order unit tensor where, exp[(1 Se)n ln a] (5) The yield stress of unsaturated soil under an isotropic stress state is specified as a product of saturated yield stress, psat , and the intensity parameter for the yield stress, , obtained from an effective degree of saturation with the fitting parameters n and a. In the equation below, M represents the critical stress ratio expressed by q/p and D denotes the dilatancy coefficient. MD
When water balance involving precipitation and evaporation is expressed in simulations, changes in suction, dependent upon changes in the water content, can be obtained from soil-water retention characteristic curves. However, analysis of such results using Karube & Kawais constitutive model (used by Fujikawa et al.) can be highly sensitive to the shape of the soil-water retention characteristic curves. This makes calculation difficult over a wide suction region. In this study, the constitutive model proposed by Ohno et al. is used since the shapes of the soil-water retention characteristic curve has no influence on identification of the constitutive model parameters in their model. Ohno et al. showed that yield function of unsaturated soil is given by the effective stress and a parameter contributing to stiffness, and selected effective degree of saturation as a parameter contributing to stiffness. They proposed a yield function of unsaturated soil as an extension of the Cam-clay model. Effective stress is determined by the following equation: net ps1 where, net pa1, ps Ses s pa pw , Se Sr Src 1 Src (1) (2) (3)
1 e0
(6)
where, , , e0 and vp denote the coefficient of compressibility, coefficient of expansion, initial void ratio and plastic volumetric strain, respectively. Figure 2 shows a conceptual diagram of the yield surface for unsaturated soil expressed by Equation (4). 2.2 Infiltration equation and material transport equation Fujikawa et al. derived the following infiltration and material transport equations from the law of conservation of mass regarding the phase of solution within pore-water: n S vSr neS r e r
v 0 c 1 c
(7) (8)
cn S c vSr c neS r e r vc ne Sr J = 0 where
: Effective stress tensor of unsaturated soil net: Net stress tensor 1: Second order unit tensor : Total stress tensor s: Suction ps: Suction stress pa: Pore air pressure pw: Pore water pressure Sr: Degree of saturation Se: Effective degree of saturation Src: Residual degree of saturation
v: Volumetric strain ne: Porosity : Saturation solubility c: Concentration ratio (0 c 1: normalized with concentration of saturated solution being c 1) v k h: Flow velocity (Darcys law) k: Permeability coefficient tensor h: Total head h Dc c: Matter flux (Fickss law) Dc: Dispersion coefficient tensor
1134
mm/month
250.00
Evaporable quantity
200.00
Precipitation
150.00
100.00
50.00
0.00
Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Figure 2. Conceptual diagram of yield surface for unsaturated soil.
Dry season
Wet season
Dry season
Figure 4.
Input climate conditions.
In these two equations, deformation and strain are related by a generalized expression enabling an arbitrary selection of a constitutive model for unsaturated soil. 3 SALT ACCUMULATION SIMULATION
Table 1. Material parameters used in analysis (fine silt). 0.06 0.01 M 1.33 0.30 a 150 n 1.0 k (cm/s) 1.0 104 m 0.3
m: Parameter is Mualems equation (1976).
The elevation of salt from deep ground is simulated with the soil/water coupled F.E. analysis code DACSAR-Uad, as applied to the governing equations introduced in the previous section. Here, we consider the climatic condition (precipitation and evaporation) of the northeastern Thailand, where salt damage is rampant. 3.1 Analysis condition
Figure 3 shows a mesh used for the simulation. A total water head of zero was applied to the lower boundary so that the groundwater level is kept at the bottom of the mesh. Drainage through the right and left boundaries is not allowed. With the upper boundary being a flux boundary, precipitation and evaporation are assigned as
Evaporation Precipitation
0.05m 80 elements)
Concentration=0.03
5.0m (0.2m x 25m elements)
Figure 3.
Analysis mesh and boundary conditions.
downward and upward flows, respectively, based on the climate of Thailands capital city of Bangkok, as shown in Figure 4. Actual evaporation from the ground surface depends on soil moisture. Therefore, in the simulation, the product of the potential evaporation and the effective degree of saturation is applied as actual evaporation. Under the influence of subtropical depression, the atmospheric temperature in Bangkok remains high, between 26 and 30 degrees Celsius, throughout the year, with only minor seasonal changes. Monsoon prevails between May and September, delivering a substantial rainfall and placing the city in semiarid climate conditions with a clear distinction between wet and dry seasons. Groundwater below the lower boundary of the analysis range is assumed to contain salt at a concentration ratio of C 0.03. Here, a material boundary of C 0.03 is provided to represent this scenario. The daily and the seasonal changes of the groundwater level are neglected. The soil material making up the ground is assumed to be fine silt, which has water retentivity similar to the local ground. Table 1 and Figure 5 show material parameters and soil-water retention characteristics, respectively, used in the analysis. The soil-water retention characteristic curve model proposed by Kawai et al. (2007) is used here. In their model, the hysteresis appearing between the drying and wetting processes can be expressed with the logistic curve equation proposed by Sugii et al. (2002). The initial degree of saturation is assigned between the primary drying and wetting curves, assuming that negative water pressure corresponding to the height from the groundwater level occurs in the ground.
4.0m
1135
3.2
Analysis results and discussion
Recrystallizationratio
Figure 6 shows the distribution profile of salt concentration within the ground. Apparently salt moves toward the ground surface. Figure 7 compares the changes in salt concentration as a function of vertical distance from the groundwater level. As one gets closer to the ground surface, the more noticeable are the effects of repeated wet and dry seasons. The salt concentration ratio converges to the initial salt concentration ratio (C 0.03) of deep groundwater for the entire ground area. Subsequently, salt concentration increases first in the nearer to the ground surface. In the third year, the concentration finally exceeds the concentration of saturated solution (C 1.00), resulting in salt
deposits (Figure 8). Figure 9 shows the monthly changes in distribution of the total head. The directions of flow are opposite for wet and dry seasons. Furthermore, the outflow from the ground due to evaporation prevails against the inflow to the ground due to rainfall. This is shown in Figure 10, representing a water balance of ground surface elements. The direction of salt movement depends on the direction of the groundwater flow. This causes salt to accumulate on the ground surface.
0.05 0.04 0.03 0.02 0.01 0
Y=3.95m
500 1000 1500 2000 2500 3000 0
Srf =1.00
W :A ing ett 3. =6, B= 1.9
0 10
Dr yin
0.8
g: A= 16 .3,
0.6
Time(day)
B= 5.9
0.4
Figure 8. Quantity of salt deposits on ground surface.
0.2 Logistic curve Eq.
Src =0.31
Dry season(Jan-Apr,Oct-Dec) Wet season(May-Sep)
Srf - Src Sr = +S 1+exp( A+Blns) rc

20
Height from G.W.L. Y(m)
4.00 3.50 3.00 2.50 2.00 1.50 1.00 0.50

Jan Feb Mar Apr May Jun Jul Aug Sep Oct Nov Dec
Suction (Negative pressure head) (m)
Figure 5.
Soil-water retention characteristics (fine silt).
0 -0.8 -0.7 -0.6 -0.5 -0.4 -0.3 -0.2 -0.1 0.0
0.1 0.2
0.3
Total head(m)
Figure 9. Monthly changes in total head. Figure 6.

0.05
Concentration ratio
Salt concentration distribution.
5.0 4.0 3.0 2.0 1.0 0.0 -1.0 -2.0 -3.0 0 mm/day
0.04 0.03 0.02 0.01 00

500 1000 1500 2000
Y=3.95m Y=2.95m Y=1.95m Y=0.95m Y=0.05m
Outflow Inflow
time (day) T
2500
3000
500
1000 1500 2000 2500 3000 Time(day)
Figure 7.
Changes in salt concentration.
Figure 10. Water balance of ground.
1136
STUDY OF STRATIGRAPHIC ARCHITECTURE RESISTANT TO SALT DAMAGE
Srf=1.00
W
Dr
yi n g
0.8
ett in
:A
As stated in the previous section, when groundwater contains a certain concentration of salt, the salt accumulation occurs due to imbalance between inflow and outflow. This section examines how a combination of the soil materials with different water retentivities can be used to influence water balance. 4.1 Ground improvements for salt damage
g: A= -3
=-1 6. 3 ,
B=7.0 Drying: A=-0.5, B=4.3 Wetting: A=1.6,

0
0.6
.6, B= 1 .9
B= 5.9
0.4
Sr =
Logistic curve Eq. Srf - Src +S 1+exp(A+ Blns) rc Src =0.23
0.2
10
15
Evaporation from ground surface depends on soil moisture. This characteristic is often used in agriculture. The technique is known as mulching, whereby material of a relatively large particle size is overlaid on the ground surface. The material used for mulching prevents the ground from drying because its low water retentivity decreases both soil moisture of the ground surface and the actual evaporation. Here, the effects of mulching are revealed through the simulation shown below. Furthermore, this simulation includes placement of a planting ground over a material for mulching, termed as improved mulching (Figure 11). In these two kinds of simulations, the effect of thickness of the mulching material and the depth of the mulching location were studied. Due to its high water permeability and low water retentivity, silica sand was considered as the mulching material. Figure 12 shows soil-water characteristics for the mulching materials used in the analysis. For comparison, the soil-water characteristic profile of fine silt is also shown (represented by the broken line). Table 2 shows material parameters. 4.2 Analysis results and discussion
Suction (Negative pressure head) (m)
Figure 12. Soil-water retention characteristics (material for mulching). Table 2. Material parameters used in analysis (material for mulching). 0.06 0.01 M 1.33 0.30 a 150 n 1.0 k (cm/s) 8.0 10
4
m 0.3
Figure 14 shows changes in concentration ratio at different heights from groundwater level. The Figure reveals that changes in concentration occur only in the elements immediately above the groundwater level (Y 0.05 m). No changes in concentration were observed in the upper layer. Moreover, changes in concentration in the layer at Y 0.05 m are caused by diffusion, which is not considered to cause serious salt damage since it is not affected by evaporation. 4.2.2 Improved mulching As stated earlier, mulching is promising as a means of deterring salt damage. However, mulching material with poor water retentivity or that which can only hold moisture in small amounts is not suitable for planting, and causes a high environmental load. To solve this problem, improved mulching can be devised and used as a type of stratigraphic architecture that would favor greening effects and be able to deter salt accumulation. The main crop in Thailand is rice, which has root systems that grow as long as 50 cm on the average. Therefore, the effects of different thicknesses in the mulching layer were examined with the upper level set at 50 cm beneath the ground. As in the previous case on mulching, deterring effects of salt damage were examined by reducing the thickness of the layer from 50 cm. Results show that a mulching layer of 20 cm thickness is sufficiently effective in deterring salt damage with no increase in concentration (Figure 15). Figure 16 shows concentration ratio changes with
4.2.1 Mulching A 50 cm mulching thickness was simulated first. It was found that the thickness of 50 cm is sufficiently effective for the inhibition of salt accumulation. In the next step, the thickness of the mulching overlay was gradually reduced. No increase in ground concentration could be observed, even with a 10 cm thick mulching layer. This achieved sufficient salt-damage deterring effects (Figure 13).
Normal

Mulching

Improved mulching
Figure 11. Schematic diagrams of ground improvement.
1137
Figure 13.
Salt concentration distributions in mulched ground.
Figure 14.
Salt concentration distributions in improved mulched ground.
0.05
0.05
Concentration ratio
Y=3.95m Y=2.95m Y=1.95m Y=0.95m Y=0.05m
Concentration ratio
0.04 0.03 0.02 0.01 00

500 1000 1500 2000
0.04 0.03 0.02 0.01 00

500 1000 1500 2000
Y=3.95m Y=2.95m Y=1.95m Y=0.95m Y=0.05m
2500
3000
Time(day)
Figure 15. Changes in salt concentration (mulching).
Time(day)
2500
3000
Figure 16. mulching).
Changes in salt concentration (improved
the mulching layer at a thickness of 20 cm. Just as in the previous mulching scenario, this Figure reveals that concentration changes do not occur in the upper layer, while some changes in concentration, resulting from diffusion, were observed only in the elements immediately above the groundwater level. Although both mulching and improved mulching have salt damage deterring effects, the mechanisms acting therein are different. Figure 17 compares changes in the degree of saturation of ground surface elements between cases with and without mulching. Fine silt, exhibiting an increased salt content, maintains a high degree of saturation, despite
a higher amount of evaporation than precipitation. In contrast, the degree of saturation of the mulching material stays at a substantially low level. This phenomenon arises due to differences in water retentivity of the materials. Under the same suction condition, the mulching material clearly exhibits a low degree of saturation, as shown in Figure 12. Actual evaporation strongly depends on the soil moisture of the ground. Therefore, in the case of mulching material with a low degree of saturation, the actual amount of evaporation decreases substantially. This explains why the water outflow from the ground is less than the water inflow from precipitation (Figure 18) and the fact that salt does not move upward.
1138
1 Degree of saturation 0.8 0.6 0.4 0.2 0 0 500

Fine silt
Mulching material
Permeability coefficient k(cm/sec)
[10-6] 2.0
1.5
1.0
0.5
1000 1500 2000 2500 3000 Time(day)
0.20 0.25 0.30 0.35 0.40 0.45 0.50 0.55 0.60 0.65 0.70
Figure 17. Changes in degree of saturation of the element of the ground surface (mulching).
Figure 20. Unsaturated permeability coefficient of mulching material.
0 -1.0 mm/day -2.0 -3.0 -4.0 -5.0 -6.0 0

Figure 18.
Inflow
decreases substantially in the dry season, as revealed in the Figure. In short, improved mulching contributes to deterred salt damage due to its low unsaturated permeability coefficient, resulting in the blockage of water flow. 5 CONCLUSIONS
500 1000 1500 2000 2500 3000 Time (day)

Water balance in mulching.
1 Degree of saturation 0.8 0.6 0.4 0.2 0 0 Y=3.40m 500 1000 1500 2000 2500 3000 Time(day)
Salt accumulation, due to repeated cycles of dry and wet seasons, was simulated with soil/water coupled F.E. analysis as applied to the constitutive model for unsaturated soil and material transfer equation. Moreover, a combination of different soil materials are proposed from our analysis that can be used in designing stratigraphic architecture that is resistant to salt damage. REFERENCES
Fujikawa, K., Iizuka, A., Kawai, K. & Thirapong, P. 2006. The mathematical modeling of unsaturated soilwater coupled problem considering mass transfer and its FE formulation, Journal of Applied Mechanics, 9: 377384 (in Japanese). Karube, D. & Kawai, K. 2001. The role of pore water in the mechanical behavior of unsaturated soils, Geotechnical and Geological Engineering, 19(3): 211241. Kawai, K., Iizuka, A., Hayakawa, E. & Wang, W. 2007. Non-uniform settlement of compacted earth structures caused by the deformation characteristics of unsaturated soil on wetting, Soils and Foundations, 47(2): 195206. Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media, Water Resources Research, 12(3): 514522. Ohno, S., Kawai, K. & Tachibana, S. 2007. Elasto-plastic constitutive model for unsaturated soil applied effective degree of saturation as a parameter expressing stiffness, Journal of JSCE, 63(4): 11321141 (in Japanese). Sugii, T., Yamada, K. & Kondou, T. 2002. Relationship between soil-water characteristic curve and void ratio, Proc. 3rd Int. Conf. on Unsaturated Soils, 1: 209214.
Figure 19. Change in degree of saturation of the mulching layer (improved mulching).
Meanwhile, Figure 19 shows changes in the degree of saturation in the mulching layer. The degree of saturation remains low, although not as low as when a mulching layer is placed over the ground surface. Finally, Figure 20 shows the relationship between the unsaturated permeability coefficient and the degree of saturation in the mulching material. The permeability coefficient
1139
An environmental chamber for studying the soil-atmosphere interaction

An-Ninh Ta, Anh-Minh Tang, Yu-Jun Cui & Jrmy Thiriat
Ecole des Ponts ParisTech, France
ABSTRACT: A large scale environmental chamber was developed to study the effect of air conditions on evaporation rate. This chamber was instrumented by suction sensors, volumetric water content sensors and temperature sensors at various depths. The soil studied, the Romainville green clay, was put in the chamber by compaction. In the air, temperature and relative humidity and wind speed were monitored. The soil surface temperature was measured by one infrared thermometer. Several humidification-drying cycles were carried out. During the drying phase, a camera fixed above the chamber allowed monitoring the development of the crack network. The evaporation rate was calculated based on the temperature and humidity data at the inlet and the outlet of the chamber. The performance of the chamber to simulate the evaporation phenomenon was then assessed. 1 INTRODUCTION of volumetric water content were buried every 200 mm; ten relative humidity sensors (RS), ten high-capacity tensiometers (TS) and eight psychrometers (PS) for the measurement of suction were installed every 200 mm. overall these sensors were disposed on two opposite sides of the chamber, allowing one measurement, for each type of sensor, every 100 mm along the height of the soil column. The same sensor disposition was adopted for ten PT1000 sensors that measured the soil temperature. Heave at the soil surface was monitored by a network of 15 displacement sensors (LV). Further information about the reinforcement of the chamber, compaction procedure, sensors calibration, specification and installation, sensors positions can be found in Tang et al. (2009). The first wetting phase aimed at saturating the soil column was performed and took 338 days. The recorded data showed a good performance of all sensors. All the recorded data were presented in Tang et al. (2009). Note that at the end of this phase, most parts of the soil column was saturated at a volumetric water content of nearly 50%, the first 50-mm from the surface where a volumetric water content was much higher (77% at 50-mm depth) because of the large swell of the soil in this zone. After the re-saturation phase, the soil column was dried from its surface. For the drying phase, new development was needed and the schema is presented in Figure 1 and a picture is presented in Figure 2. Air flow rate from the compressed air source is controlled by a flowmeter (2% of accuracy over the working range of 500 l/min).
The soil-atmosphere interaction is usually involved in geotechnical and geo-environmental engineering problems, such as structural and pavement damage due to expansive clays, stability of embankments and dams under seasonal climatic changes, stability of constructions in relation to shrink and swell of expansive soils, safety of surface wastes disposal, etc. A good understanding of the mechanisms involved in the soil-atmosphere interaction is essential when analyzing this kind of problems. In the present work, a large scale environmental chamber was developed to study the soil water evaporation in various climatic scenarios. The soil studied was Romainville green clay. This expansive soil, present in some areas of Paris, has been considered as responsible for a large amount of damage of buildings. 2 EXPERIMENTAL SETUP
The experiments were performed in an environmental chamber that consists of an acrylic transparent tank with 800-mm internal width, 1000-mm internal length and 1550-mm height. Romainville clay at a water content of 5.7% was sieved at 2 mm and compacted in the chamber by layers of 50-mm thick. The soil dry density was 1.35 Mg/m3 that was similar to the in-situ dry density (Cui et al. 2006). During the compaction of soil layers, several sensors were installed in the soil. In total, five ThetaProbe sensors (TP) for the measurement
1141
7 4 3 6 Soil surface 5 2 1
Figure 1. Equipment for drying tests (1. compressed air source, 2. flowmeter, 3. heat pipe, 4. measurement point of relative humidity and temperature, 5. air distributor, 6. air collector, 7. infrared thermometer, 8. camera).
holes of the air distributor. The wetted air was then gathered in the air collector where temperature and relative humidity were again measured. Note also that, monitoring the temperature at the soil surface was ensured by a infrared thermometer. The point of measurement is firstly fixed at the center of the surface and then changed to verify the uniformity of temperature at soil surface. The development of cracking at the soil surface during the drying tests was monitored by a high definition camera with specific lens. This camera allows covering the entire soil surface. Note also that the sun light was first prevented, and the soil surface was lighted by 48 LEDs around the transparent cover. The LED lighter has lower heat emission, minimizing the heat perturbation to the soil-atmosphere interface. Furthermore, various resistance temperature detectors and two RS sensors were installed on the internal wall at different heights. These data will complete other data to establish the temperature and humidity profiles. The ambient temperature and humidity were also monitored. 3 EXPERIMENTAL PROCEDURES AND DATA PROCESSING
The drying test took 30 days. One water layer of about 20 mm was put on the soil surface at the beginning of the test. Air was blown at a rate of 100 l/min and the air was heated to reach 150C at the heat pipe level. Temperature and relative humidity measurement at inlet and at outlet were used to calculate the evaporation rate. For this purpose, the saturated vapor pressure was first calculated from the temperature (Brutsaert 1988).
2 13.3185 tR 1.9760 tR Pvs 101.325 exp 3 4 0.6445 tR 0.1299 tR
(1)
where tR 1 (373,15/T) in which T is the temperature in K. Then, the vapor pressure was calculated: Pv h 100 Pvs (2)
Figure 2.
Picture of the experimental setup.
where h is the air relative humidity. Then, the density of water vapor was deduced, as follows.
Thanks to the pipe heat, air was then heated and the maximum temperature was about 250C. The heated air was then measured by one RS sensor giving its real temperature and relative humidity before being blown into the chamber through eight
v
0.622 Pv RTa
(3)
where the value of 0.622 ( 18.016/28.966) is the ratio of the molecular weights between water and dry air, R is gas constant. Finally, the difference of
1142
38
T_air (C)
32
26
At the inlet At the outlet
In the chamber Out of the chamber
(a)
20 55 50 45 40 35 30 25 20 15 10 5 0 0
h (%)
(b) (Same symbols as temperature's)
Figure 3.
Cracking network at the end of the drying test.
10
Elapsed time (day)
20
30
water vapor density of air at inlet v_inlet and at outlet v_outlet was used to calculate the evaporation rate: Q q (v_outlet v_inlet) (4)
Figure 4. Air temperature (a) and air relative humidity (b) variation versus time.
where q is the air flow rate. This method was also used by Aluwihare and Watanabe (2003). In order to monitor the cracking development at the soil surface, one photo every 90 min was taken and one photo every two days was analyzed by image processing. The crack network was analyzed using two parameters: surface crack ratio (the ratio between cracking surface over initial surface) and weighted width. One picture of the soil surface after 30 drying days is presented in Figure 3 showing that the network is mainly made up of wide cracks. Further information about the image processing and cracking parameters identification could be found in (Tang et al. 2008a). 4 EXPERIMENTAL RESULTS
The evolution of the air temperature and the air relative humidity at the inlet, at the outlet, in the chamber, out of the chamber were plotted versus time in Figure 4. It can be noted that the mean temperature at the inlet was about 34C with a small variation of 4C. The ambient temperature varied from 20 to 25C, slightly lower than the temperatures in the chamber and at the outlet during 20 first days. It seems that the ambient temperature conditioned the variation trend of the three others. The values of temperature in the chamber and at the outlet are very close. As far as the relative humidity is concerned, the value at the inlet is tiny and close to 0. The ambient relative humidity presents important variations (2550%) and seems to have no influence on the three others. As the temperature, the values of
relative humidity at the outlet and in the chamber are closes. Indeed, they both present a regular decrease from 53 to 45%. Figure 5 presents the changes in soil volumetric water content, soil suction and soil temperature versus time. For the volumetric water content, it is observed that only the sensor at 50-mm depth presents a significant variation: 65 to 15% of volumetric water content after 5 days of evaporation, others values of volumetric water content at deeper positions remaining unchanged and all close to 50%. Figure 5(b) presents significant variations of suction measured by two PS sensors and one TS sensor during the last 10 days. An important increase of suction at 50-mm depth can be noted. Indeed, the suction exceeded 4000 kPa at the end of the test. At 150-mm depth, the suction measured by the second psychrometer was about 150 kPa. The TS sensor at 200-mm depth started working at t 20 days and it provided one suction value higher than 1000 kPa at the end. At deeper level, the suction was lower than the lower limit of PS sensor (50 kPa) and than the limit of TS sensor (10 kPa) and therefore no result is available. It is important to notice that PS sensors were installed in the soil column while TS sensors were installed on the two opposite sides of the soil column. Furthermore, it seems that at the sides, the soil was dried at a higher rate than the soil at the central part; this is probably due to wide cracks at the sides. Regarding the soil temperature, it is observed that the lowest temperature is at the soil surface. In the soil column, the temperature at 50-mm depth is lower than the temperature at a higher depth. In addition, comparison the soil temperature to air
1143
70 60 50
(a) 50 mm 250 mm 450 mm 650 mm 850 mm (b)
40 30 20 10
10000 1000
100 10 1 23 22 21 20 19 18 17 16 0
TS_200 mm PS_150 mm PS_50 mm 50 mm 0 mm (c)
is influenced by the external temperature) is high, all temperature curves moves to the right showing a temperature increase and vice versa. On the other hand, another temperature gradient is also observed in the zone of 650 to 950-mm depth. This one is smaller than those at the soil surface. Figure 7 presents the evolution of two cracking parameters: cracking surface ratio Rcs (a) and weighted width Ww (b). It can be noted that these two curves are similar. Indeed, they are both clearly divided into two parts: a first initial value (until t 10 days, because of preexisted cracks), and a second increasing phase. The final value of Ww is about 27 mm and that of Rcs is about 35%. The calculated values of actual evaporation rate and accumulated evaporation are presented in Figure 8. The evaporation rate decreases from 1.65 (t 0 day) to 1.25 mm/day (t 30 days). This decrease is progressive without significant changes. A small bounce is however observed at t 20 days
30
T_soil (C)
s (kPa)
(%)
10
Elapsed time (day)
20
40
(a)
Figure 5. Soil volumetric water content (a), soil suction (b) and soil temperature (c) variation versus time.
Air Soil t = 0 day t = 5 days t = 10 days t = 15 days t = 20 days t = 25 days t = 30 days
17 18 19 20 21 22 23 24
Rcs (%)
30 20 10 0 30 25 20 15 10 5 0
50 -50 -150 -250 -350 -450 -550 -650 -750 -850 -950
Level (mm)
Ww (mm)
(b)
10
Elapsed time (day)
15
20
25
30
Figure 7. Cracking characterization: cracking surface ration (a) and weighted width (b).
1.7
Tempe rature (C)
Figure 6.
Temperature profiles at various times.
AE (mm/jour)
1.6 1.5 1.4 1.3 1.2 50
(a)
temperature shows that all temperature values were influenced by ambient temperature. Note also that the surface temperature value presents the highest decrease at the beginning of the test. In order to better understand the evolution of the temperature in the chamber during the drying process, the recorded temperature data is used to draw up temperature profiles for every 5 days (Figure 6). Two temperature gradients coming to the soil surface can be noticed. These gradients increase with the evaporation progress. In addition, it is observed that the external temperature conditions all temperatures in the chamber. Indeed, when the air temperature at 50 mm (which
CAE (mm)
40 30 20 10 0 0 10 20
(b)
Elapsed time (day)
30
Figure 8. Actual evaporation rate AE (a) and cumulated actual evaporation CAE (b) versus time.
1144
which is probably due to the air temperature increase. In total, 40-mm water was evaporated. 5 DISCUSSION
Environmental chamber is usually used when studying phenomena related to atmosphere-soil intercation. Kohsiek (1981) developed the first chamber long time ago and it was then used by several authors (Aluwihare and Watanabe 2003; van de Griend and Owe 1994). This chamber mainly involves air part and the part for soil investigation is poor. Wind-tunnel system is well-equipped but often needs high requirements (Wang 2006). Note also that the wind-tunnel developed by Yamanaka et al. (1997) is similar to the environmental chamber developed in this study, but does not allow the control of inlet temperature. In most cases, temperature is monitored or controlled in terms of ambient temperature (Kondo et al. 1990; Wilson et al. 1994; Wilson et al. 1997). On the other hand, the soils studied in these studies were usually granular soils as sand (Aluwihare and Watanabe 2003; Kondo et al. 1992; Wilson 1990; Wilson et al. 1994; Yamanaka et al. 1997) and loam (Kondo et al. 1992). Wilson et al. (1997) used clays in their studies but the sample was only a thin layer of less than 1 mm. In fact, experimental study on evaporation phenomenon using big clay column has been rarely carried out in laboratory. Note also that the influence of cracks network on evaporation has been rarely studied in laboratory. Some preliminary studies on this question in the field conditions can be found (Ritchie and Adams 1974; Selim and Kirkham 1970). The wind speed in the chamber was about 0.4 m/s. This value is not very high but in the same order of magnitude of the wind speed at some sites in France, 2.0 m/s (Cui et al. 2005) and of those in the literature: 1.0 m/s in Yamanaka et al. (1997) and 4.0 m/s in Kondo et al. (1992). Regarding the temperature, with a temperature of 150C at the heat pipe, a significant temperature at the inlet of the chamber can be generated: 32C. Note that this value has been also usually found in drought periods in France (Cojean et al. 2009; Cui et al. 2005). Continuous monitoring of the soil surface by a high resolution camera made possible the analysis of the cracks network. The image system employed allowed detection of the smallest cracks in the order of 1 mm. The evolutions of two cracking characterization parameters have been found to be similar. This seems logical since the tow parameters are related one to another. The natural light is usually used to take photos (Lakshmikantha 2009; Tang et al. 2008b). In this
study, in order to minimize ambient effect on the chamber, the natural light was replaced by LED light. The effect of the boundary conditions was observed on the recorded data in term of temperature. Indeed, these results show an important influence of the ambient temperature on all temperatures in the chamber: in the air and in the soil. This kind of influence can be observed either in situ (Alvenas and Jansson 1997) or in laboratory (Kondo et al. 1992) where the ambient temperature is not controlled. The recorded data in the air show a good performance of RS sensors. Indeed, this sensor enables relative humidity measurement in a large range: 1% at the inlet to 60% in the chamber over a long time. Note also that, during infiltration phase, RS sensors indicated relative humidity value being close to 100% (Tang et al. 2009). This performance complies with specification given by the manufacturer (Comet 2008). Almost all sensors have worked well and have succeeded in covering their full measurement range. Indeed, for suction measurement, PS sensors can exceed 4000 kPa and TS sensors can reach 1500 kPa. In addition, in terms of volumetric water content, TP sensor operated well in a wide range from 5 to 80%. A longer drying time is need to obtain a full evolution curve of evaporation rate that could be represented by three distinct phases: maximum rate in the first phase followed by a second decreasing evaporation rate phase and a third stabilization phase (Hillel 1980; Hillel 2004; Lal and Shukla 2004). Moreover, a longer time test permits to have more results in terms of suction and water content variations in deeper levels. 6 CONCLUSION
A 30-day drying test was carried out in a large scale environmental chamber showing an important potential of this chamber to study the evaporation phenomenon. The full monitoring of air parameters (temperature, relative humidity) and soil parameters (temperature, relative humidity, suction, water content and soil cracking) enables the assessment of evaporation evolution in relation to the soil cracking effect and the effect of air conditions. More tests with longer time and with various air conditions are ongoing. ACKNOWLEDGMENT This work was carried out within the research project Ala et Risque Scheresse supported by the MAIF Foundation.
1145
REFERENCES
Aluwihare, S. & Watanabe, K. 2003. Measurement of evaporation on bare soil and estimating surface resistance. Journal of Environmental Engineering 129(12): 11571168. Alvenas, G. & Jansson, P.E. 1997. Model for evaporation, moisture and temperature of bare soil: calibration and sensitivity analysis. Agricultural and Forest Meteorology 88(14): 4756. Brutsaert, W. 1988. Evaporation into the atmosphere. Theory, history, and applications D. Reidel Publishing Company. Cojean, R., Audiguier, M., Geremew, Z., Laribi, S., Thnevin, I., Cui, Y.J., Tang, A.M., Ta, A.N., J.-V., H., Taillefer, N., Deffontaines, B., Kaveh, F. & Fruneau, B. (2009). Rapport collectif coordonn par Cojean R., en partenariat entre Armines-Mines ParisTech/Centre de Gosciences, lEcolde des Ponts ParisTech/UR NavierGotechnique, le CSTB et UPE-MLV/ OTIG. Rapport de synthse final du projet Ala et risque scheresseProjet Fondation MAIF. R090130RCOJ, Centre de Gosciences, Mines ParisTech, Fontainebleau, France. Comet. (2008). T3111 Transmitters: Programmable transmitter of temperature, relative humidity and other derived hymidity values with 420 mA outputs. Instruction Manual. 8. Cui, Y.J., Lu, Y.F., Delage, P. & Riffard, M. 2005. Field simulation of in situ water content and temperature changes due to ground-atmospheric interactions. Gotechnique 55(7): 557567. Cui, Y.J., Mantho, A.-T., Cui, K. & Audiguier, M. 2006. Water retention properties and volume change behaviour of natural romainville clay. In G.A. Miller, C.E. Zapata, S.L. Houston & D.G. Fredlund (eds), Fourth International Conference on Unsaturated Soils (UNSAT2006): 873882. American Society of Civil Engineers, Reston, VA 201914400, United States. Hillel, D. 1980. Applications of soil physics. New York: Academic press. Hillel, D. (2004). Evaporation from bare soil and wind erosion. Introduction to environmental soil physics, Elsevier, ed., 337361. Kohsiek, W. 1981. A rapid-circulation evaporation chamber for measuring bulk stomatal-resistance. Journal of Applied Meteorology 20(1): 4252. Kondo, J., Saigusa, N. & Sato, T. 1990. A parameterization of evaporation from bare soil surfaces. Journal of Applied Meteorology 29(5): 385389. Kondo, J., Saigusa, N. & Sato, T. 1992. A model and experimental-study of evaporation from bare-soil surfaces. Journal of Applied Meteorology 31(3): 304312.
Lakshmikantha, M.R. (2009). Experimental and theoretical analysis of cracking in drying soils, Ph.D. thesis, Universitat Politechnica de Catalunya, Barcelona. Lal, R. & Shukla, M.K. 2004. Principles of Soil Physics. NewYork, Basel: Marcel Dekker, Inc. Ritchie, J.T. & Adams, J.E. 1974. Field measurement of evaporation from soil shrinkage cracks. Proceedings of the Soil Science Society of America 38(1): 131134. Selim, H.W. & Kirkham, D. 1970. Soil temperature and water content changes during drying as influenced by cracks: A laboratory experiment. Soil Science Society of America Proceeding 34(4): 565569. Tang, A.M., Ta, A.N., Cui, Y.J. & Thiriat, J. 2009. Development of a large scale infiltration tank for determination of the hydraulic properties of expansive clays. Geotechnical Testing Journal 32(5): 385396. Tang, C., Shi, B., Liu, C., Zhao, L. & Wang, B. 2008a. Influencing factors of geometrical structure of surface shrinkage cracks in clayey soils. Engineering Geology 101(34): 204217. Tang, C.S., Shi, B., Cui, Y.J. & Tang, A.M. 2008b. Effect of temperature on desiccation cracking behaviour of clayey soil. In J.P. Magnan, R. Cojean, Y.J. Cui & P. Mestat (eds), International Symposium Drought and Constructions: 115122. van de Griend, A.A. & Owe, M. 1994. Bare soil surfaceresistance to evaporation by vapor diffusion under semiarid conditions. Water Resources Research 30(2): 181188. Wang, W.C. (2006). Wind Tunnel Experiments on Bare Soil Evaporation, National Central University, Taiwan. Wilson, G.W. (1990). Soil evaporative fluxes for geotechnical engineering problems, Ph.D. thesis, University of Saskatchewan, Saskatchewan. Wilson, G.W., Fredlund, D.G. & Barbour, S.L. 1994. Coupled soil-atmosphere modelling for soil evaporation. Canadian Geotechnical Journal 31(2): 151161. Wilson, G.W., Fredlund, D.G. & Barbour, S.L. 1997. Effect of soil suction on evaporative fluxes from soil surfaces. Canadian Geotechnical Journal 34(1): 145155. Yamanaka, T., Takeda, A. & Sugita, F. 1997. A modified surface resistance approach for representing bare-soil evaporation: Wind tunnel experiments under various atmospheric conditions. Water Resources Research 33(9): 21172128.
1146
Foundations
Settlement prediction of cyclically loaded footings resting on partially saturated sand by means of a kinematic hardening model
A. Becker
Technical University of Kaiserslautern, Division of Soil Mechanics and Foundation Engineering, Kaiserslautern, Germany
ABSTRACT: Partially saturated sand shows at low saturation degrees particular bonding-mechanisms corresponding to a kind of cohesion. To describe the incremental increase of settlement during cyclic loading in this type of soil a typical elasto-plastic material model has been expanded. The model includes an isotropic-kinematic hardening rule for deviatoric loading conditions. The volumetric components are described by a suitable dilatation function. The constitutive model belongs to the general class of elasto-plastic multi-surface models. In addition to the yield surface another bounding surface is used to distinguish between first loading as well as re- and unloading conditions. The degree of saturation is not explicitly used but expressed in terms of an apparent cohesion as determined from suitable cyclic triaxial-tests. In numerical sensitivity analyses using the FEM the behaviour of vertically loaded strip footings is investigated. Based on dimensional analysis empirical correlations are determined, which allow prediction of accumulated settlements in dependency of the number of cycles, the cyclic load amplitude, the static pre-loading of the footing, the initial void ratio, the degree of saturation, and the system geometry. Results are presented in terms of polynomial expressions for direct use in geotechnical design. 1 INTRODUCTION
p p0
Load
In the case of partially saturation surface tension of pore fluid leads to compressive stresses acting on the grain structure of non-cohesive soils. This effect of increase in strength, known as capillary cohesion, is described in numerous publications for static loading. In the field of cyclic loading tests on partially saturated sands the experiments of Wu (1983) and Qian (1990) are known. Both authors presented the influence of water content w on the dynamic shear modulus at small strain amplitudes. At low degrees of saturation values of shear modulus rise up over about 60% compared to values for dry sand. Holzlhner (1978) investigated permanent settlements of cyclically loaded foundations in model tests on partially saturated sand. The experimental results show that due to the capillary cohesion the permanent subsidence due to cyclic loading were smaller using moist sand with w 4% compared to dry conditions. Besides static loading, foundations may often be loaded cyclically. In addition to settlements due to static loads permanent settlements with increasing number of cycles can occur and lead in this way to problem with serviceability of the building. The development of cyclically induced settlements with cycles number can be described by a logarithmic relation
p p0
e0, w, b
Time
Figure 1. Strip foundation under cyclic loading.
depending on the number of cycles N, e.g. Hettler (1981), Holzlhner (1978), Wichtmann (2005). Figure 1 shows a strip foundation of width b and embedment t. Soil is defined by unit weight [kN/m3], initial void ratio e0 [] and water content w [%]. Static loading is given by p0 [kPa] and the amplitude of cyclic load by p [kPa]. In this paper the development of settlement depending on N and the above mentioned parameters is determined on the basis of FEM calculations. The results are approximated by analytical relationships. 2 CONSTITUTIVE MODEL FOR THE SOIL
The model briefly presented is based on an elasto-plastic formulation, whose validity has been demonstrated in the literature for static loading
1149
conditions including capillary effects (Meiner & Borm, 1988). Strains which occur during unloading and reloading of sandy samples as well as near the isotropic state of stress during the initial stage of loading are defined as quasi-elastic. Thus, the parameters K and G of the linear elastic material law may be determined. It was found, that the bulk modulus K as well as the shear modulus G depend on both, the stress condition and the void ratio e0. For the moduli are obtained: K dI 645 E pa (1 I a )1 3 dI 1 2 e0 0.1
dsi 2 dei
e
-1
Memory surface f(, eP, e0)
-s1
ij
IIs.c1/2 IIs
1/2
-s 2
-s 3
(1)
- 3
2
- 2
G
200 e0
2
pa (1 0.01 I a 1.5 10 I a )
5
(2)
Figure 2.
Two surface model in principle stress space.
where: I first invariant of the strain tensor of the quasi-elastic conditions I first invariant of the stress tensor si deviatoric stress components eie quasi-elastic deviatoric strain components pa 100 kPa; atmospheric pressure Ia I/pa. In the case of unloading and reloading the bulk modulus is given by the relationship 3333 pa K ur (1 0.04 I a ) (e0 0.1) (3)
tensor ij directed to the cones central axis. According to this, the stress tensor has to be divided into a volumetric term (kk kk) ij as well as a deviatoric part sij . The flow rule is formulated as f ( , e p , e0 ) II s
12
II s,c
12
s A I 1 B III 3 2 II s
m
(4) with the stress invariants, II s sij . s ji , III s sil slj s ji , and sij ij ij I /3 The term containing cohesion ccycl is given by B 1 6 ccycl 1 B III s 3/ 2 II s
m
where a I Iu, and Iu the first stress invariant as obtained before unloading. Besides a non-associated flow rule, the model contains a yield criterion depending on plastic strains P. Modification and expansion of the existing isotropic hardening hypothesis in form of a kinematic-isotropic part was done to describe changes of the yield surface during cyclic loading. Determinant parameter depicting the increasing change in shape of the yield surface is the second invariant of plastic strain tensor IIep1/2. The description of the elasto-plastic material response is based on a multi-surface model, e.g. Chen (1990). In addition to the yield surface, which limits the elastic region, there is a surrounding isomorphic surface to distinguish between first loading, unloading as well as reloading. During first loading this so-called memory surface expands isotropically. Stress states reaching the yield surface lead to plastic deformations, and a change of yield surface in size as well as with respect to position within the memory surface. Figure 2 shows the yield surface in principal stress system. The yield function is described by stress invariants that are not related to the vector of space diagonal ij but to a triagonal
II s,c
1/ 2
(5)
The parameters A, B, m control the shape of the yield surface both under hardening as well as softening conditions. A detailed presentation of the formulation is given by Becker (2002). The parameter ccycl describes the cohesion as a function of void ratio e0 and water content w with the following relationship: ccycl wexp(w) (6)
The parameters , , are determined from cyclic triaxial tests on partially saturated sand: 0.228e06.582, 3, 0.7. Maximum cohesion value has been achieved in these experiments at a water content of about 4.5%. Table 1 shows some properties of the sand used in the laboratory tests (Becker, 2002).
1150
Table 1. Material properties. Maximum void ratio e0,max [] Minimum void ratio e0,min [] Effective size d10 [mm] Uniformity coefficient CU [] Table 2. Range of parameters in numerical study. 0.65 0.80 0 15.5 15 90 100 400 14 02 0.92 0.59 0.18 1.47
is approximated by analytical expressions as a function of cycles N by following well-known logarithmic law: N 1 uz ( N ) U1 ln 1 b U2 (7)
Void ratio e0 [] Capillary cohesion ccycl [kPa] Amplitude p [kPa] Static load p0 [kPa] Foundation width b [m] Foundation embedment t [m]
Here, U1 and U2 are dimensionless functions, depending on the parameters listed above: t p p ccykl ,e0 ) , U1 1( , 0 , b b b b t p p ccykl U2 2 ( , 0 , , ,e0 ) b b b b
(8)
NUMERICAL PARAMETER STUDY

1 0,000
NumberZyklenzahl of cycles N [-]

100 10000 1000000
For a rigid foundation founded on partially saturated soil, as shown in Figure 1, parametric studies with the finite element method have been performed (Becker, 2002). Within this frame, the influence of different parameters on the development of settlement uz(N) in dependency on the number of cycles N has been investigated. The range of parameters used in the study is specified in Table 2. The unit weight of the soil is assumed to be constant for all calculations with 16.5 kN/m3. The foundation is described by means of elastic material parameters Ec 30.000 MPa, 0.20 and the unit weight c 25 kN/m3. As a result of this study Figure 3 shows the normalized settlement uz(N)/b versus the number of cycles N for selected examples. The following effects can be noticed: i. With increasing initial void ratio e0 cyclic settlement uz(N)/b increases. ii. Even a small water content of about 2.5% leads to a significant reduction of settlements due to the influence of the capillary cohesion. iii. With increasing load amplitude p settlements increase approximately linear with p. iv. At the same number of cycles N settlements decreases with increasing static load p0. v. Increase of foundation width b leads to an increase in settlement at the same number of cycles N. vi. The increase of embedment t results in a significant reduction of the settlements. 4 APPROACH FOR PREDICTION CYCLIC SETTLEMENT
e0=0,65
uz(N)/b [-]
0,005
15kPa 100kPa
e0=0,70
0,010
w=2%, e0 1m
1m
e0=0,80 e0=0,75
0,015
1 0,000
100
10000
1000000 w=4,5% w=3,5%
uz(N)/b [-]
0,005
15kPa 100kPa
w=2,5%
0,010
w, e0=0.65 1m
1m
w=0,0%
0,015
1 0,000 100 10000 1000000 p=15kPa
uz(N)/b [-]
0,020
p 100kPa
p=30kPa p=60kPa
0,040
w=2% e0=0.65 1m
1m
p=90kPa
0,060
1 0,000 p0=400kPa 100 10000 1000000
uz(N)/b [-]
15kPa
p0=300kPa
0,005
w=2% e0=0.65 1m
p0 1m
p0=100kPa p0=200kPa
0,010
Based on the results from the FEM calculations, the evolution of cyclic induced settlement uz(N)/b
Figure 3. Influence of some parameters on the development of cyclic settlement.
1151
For some parameter combinations the graphs for U1 and U2 versus one parameter are depictet exemplarily in Figure 4. These curves show that an approximation of the variables U1 and U2 by power functions is sufficiently accurate. Introducing the dimensionless variables (b 1):
t t b , p0 p0 b ,p p b , ccykl ccykl b ,b b b
and, Bn 2n 61 2(n 1) 60 b n 1 ... 32. The constants 63 126 are given by Becker (2002). Accordingly U2 is approximated by:
U 2 A1 A2 c A3 t A4 c t A5 p0 A6 c p0 A7 t p0 A8 c t p0 e0 (A9 A10 c A11 t A12 c t A13 p0 A14 c p0 A15 t p0 A16 c t p0 ) p (A17 A18 c A19 t A20 c t A21 p0 A22 c p0 A23 t p0 A24 c t p0 e0 (A25 A26 c A27 t A28 c t A29 p0 A30 c p0 A31 t p0 A32 c t p0 ))
(9)
leads to
U1 p (B1 B2 c B3 t B4 c t B5 p0 B6 c p0 B6 c p0 B7 t p0 B8 c t p0 e0 (B9 B10 c B12 c t B13 p0 B14 c p0 B15 t p0 B16 c t p0 ))
(12)
(10)
with the exponent :

B17 B18 c B19 t B20 c t B21 p0 B22 c p0 B23 t p0 B24 c t p0 e0 (B25 B26 c B27 t B28 c t B29 p0 B30 c p0 B31 t p0 B32 c t p0 )
where, An 2n 61 2(n 1) 60 b , n 1 ... 32 and 63 to 126 are constants (Becker, 2002). 5 EXAMPLE
(11)
100
b/t = 1/1 m, p0 = 100 kPa, p = 15 kPa w = 4.5%, = 16.5 kN/m3

U2
The application of the relations given above is shown in the following example, Figure 5. According to Equation 2 the capillary cohesion can be assessed to be ccycl 11.2 kPa. The dimensionless parameters, Equation 9, are given as: p t 1.30 250 1.30 p0 0 13.16 b 1.0 b 1.0 19 p ( )20 b 3.0 p 1.05 b 3 b 1.0 19 b 1.0 ccykl 11.2 ccykl 0.59 b 1.0 19 t Following Equations 10 and 11 we obtain U1 with the constants i as: B1 63 64 b 13.546 7.012 3.0 34.583 B2 65 66 b 13.954 7.109 3.0 35.282
U1 10 5 , U 2 10 -1
75
50
U1
25
0 0,60
0,65
0,70
0,75
0,80
0,85
e 0 [-]
450
b/t = 1/1 m, p0 = 100 kPa, e0 = 0.65 w = 4.5%, = 16.5 kN/m3

U1
U1 10 5 , U 2 10 -1
300
150
U2
0 0,00
25,00
50,00
75,00
100,00
, ..., B32 0.004 U1 9.047104.
p [kPa]
75
b/t = 1/1 m, p = 15 kPa, e0 = 0.65 w = 4.5%, = 16.5 kN/m3

U2
20kPa 250kPa
U1 10 5 , U 2 10 -1
50
25
U1
0 0 100 200 300 400 500
e0=0.68 w=4%
1,3m 3m
p 0 [-]
Figure 4. U1 and U2 versus some parameters.
Figure 5.
Illustrative example.
1152
Number of cycles N [-]

1 0,000 100 10000 1000000
uz(N)/b [-]
=1m b=1m 0,005 =2m b=2m b=3m =3m
0,010
Figure 6. Settlement uz(N)/b vs. N for different values of b.
The validation of the model is based on cyclic triaxial tests. In numerical studies the development of settlements of a strip foundation induced by a cyclic vertical load is investigated. The results of this FEM calculations yield to a equation that allow prediction of settlements. Besides constant values given in tables, the geometry in form of embedment and foundation width is needed as input. Loading is described by a static as well as a cyclic component. Behaviour of soil is covered by the initial void ratio, water content and density. The solution is straightforward, and it can be implemented manually or in any spreadsheet program. REFERENCES
Similarly, we find U2 using Equation 12: A1 63 64 b 400.404 1.193 3.0 403.983 A2 65 66 b 373.39 90 5.004 3.0 388.401 , ..., A32 0.891 and thus U2 367.1. The normalized settlement uz(N)/b is given by Equation 7:
N 1 uz ( N ) N 1 U1 ln 1 9.047 10 4 ln 1 367.1 b U 2
This curve is shown in Figure 6. After N 106 cycles a settlement uz(N) for the foundation can be estimated to: 106 1 uz (106 ) 3.0 9.047 10 4 ln 1 0.022 m 367.1 A variation of the foundation width b for this example leads to: b 1 m: b 2 m: U1 4.483104, U1 6.765104, U2 287.5 U2 327.3
Becker, A. 2002. Stoffmodell und numerisches Modell fr zyklisch beanspruchte, teilgesttigte Sande. Verffentlichung des Fachgebietes Bodenmechanik und Grundbau der Technischen Universitt Kaiserslautern. Chen, W.F. 1990. Nonlinear Analysis in Soil Mechanics. Amsterdam: Elsevier. Hettler, A. 1981. Verschiebungen starrer und elastischer Grndungskrper in Sand bei monotoner und zyklischer Belastung. Verffentlichung des Instituts fr Bodenmechanik und Felsmechanik der Universitt Karlsruhe. Holzlhner, U. 1978. Bleibende Setzung von Fundamenten infolge dynamischer Last. Bautechnik 55: 150154. Meiner, H. & Borm, G. 1988. Construction of a double tunnel with ground windows. Proc. 2nd Int. Conf. on Case Histories in Geotechnical Engineering, St. Louis. Qian, X. 1990. Dynamic behaviour of unsaturated cohesionless soil. Ph. D. Dissertation. Univ. of Michigan, Ann Arbor. Michigan. Wichtmann, T. 2005. Explizites Akkumulationsmodell fr nichtbindige Bden unter zyklischer Belastung, Schriftenreihe des Institutes fr Grundbau und Bodenmechanik der Ruhr-Universitt Bochum. Wu, S. 1983. Capillary effects on dynamic modulus of fine-grained cohesionless soils. Ph. D. Dissertation. Univ. of Michigan. Ann Arbor. Michigan.
Results for these two foundation widths are shown in Figure 6. 6 CONCLUDING REMARKS
An appropriate constitutive model is presented, which describes the behaviour of partially saturated sand under the influence of cyclic loading.
1153
Performance and design of foundations on unsaturated expansive soil

J.T. Bryant & M.A. Haque
Bryant Consultants Inc., Carrollton, Texas, USA
ABSTRACT: The design of foundation systems on expansive soils requires knowledge of unsaturated soil mechanics. Expansive clay minerals are electrochemically active particles, especially in the presence of water. This affinity for water leads to soil mass volume changes with changes in soil moisture content. When a cohesive soil mass hydrates, the materials experience a volume change (termed swelling or heaving), the magnitude of which depends on the clay chemistry and the initial degree of saturation. The results from the swell test method, which tends to provide a conservatively higher estimate of movement due to the lateral confinement of the soil sample in a steel ring, is used as a basis to design most of the foundation systems. However, this test is not representative of conditions in nature, where the vertical swell strain is only a portion of the total volume change. This test method can also produce conservative estimates of movement due to complete inundation of the sample with water throughout the entire active zone depth, which is not generally typical in-situ. In order to avoid overly conservative design and to provide better performing foundation systems, in this study a heuristic approach is revealed for producing guidelines to design efficient foundation systems at a reasonable cost. The heuristic approach will also be shown to comport with principles of unsaturated soil mechanics. 1 1.1 INTRODUCTION Foundation design parameters 2 POTENTIAL HEAVE
Expansive soils are widely distributed throughout the different countries in the world. The materials that experience a volume change as the moisture profiles vary are termed as expansive soils. These soils typically exhibit shrinking (volume reduction) due to drying and heave (volume increase) during wetting. This differential movement influenced by seasonal variations can cause severe distress to the foundation and super structure built upon them (Nelson & Miller, 1992). The factors that affect the expansive nature of these soils can be referenced using site-specific environmental conditions and soil characteristics such as the initial degree of saturation, Plasticity Index (PI), type of mineral present, and soil suction. The intrinsic properties of soil mass are clay mineralogy, particle size and Plasticity Index (PI) and the external interdependent properties are permeability, pore water pressure and matric suction. The intrinsic properties of soil mass do not change over time, whereas the interdependent properties do change over time. These changes are what results in volume change. In general, the foundation systems on expansive soils are designed based on the bearing capacity of the soil and the amount of heave expected from the subgrade soils at that site.
Typically, the increase in soil volume caused by seasonal variations is referred to as heave. Heaving of expansive soils occurs when the soil encounters an increase in its moisture profile. The potential heave of expansive soils can be predicted using several methods such as the swell test methods, PVR method, Suction based methods etc. There are different types of swell test methods such as free swell, constant volume method, and overburden stress swell test method (Al-Mhaidib, 1999). However, the one-dimensional overburden swell test method is the most commonly used method due to its ease of operation and general availability of the equipment in most of the soil labs (Kassiff and Ben-Shalom, 1971). 2.1 One-dimensional swell test
The one-dimensional swell test or the oedometer swell test is performed according to the procedure specified by the ASTM D4546 standard. The prepared specimen is placed in the ring with a saturated porous stone at the top and bottom. Then the sample is completely immersed in water and the expansion of the specimen is measured until there is no significant change. Finally, the water content from the top, bottom, and middle of the specimen is measured.
1155
2.2
Volumetric vs Vertical strain
This swell test method tends to provide a conservatively higher estimate of movement due to the lateral confinement of the soil sample in a steel ring, which confines all of the movement to the vertical direction. However, this condition is not representative of conditions in nature, where the vertical swell strain is only a portion of the total volume change. Other conservative assumption inherent in the swell test method includes the assumption of uniform wetting and vertical boundary conditions with no exponential suction decay with depth. This test method can also produce conservative estimates of movement due to complete inundation of the sample with water throughout the entire active zone depth, which is not generally typical in-situ. In order to determine the influence of laterally restraining the sample in the oedometer swell tests, (Al-Shamrani & Al-Mhaidib, 1998) compared the swell percentages from the one-dimensional oedometer swell tests and three-dimensional swell tests carried out using Bishop and Wesleys tri-axial stress path cell. The three-dimensional swell tests measured both the volumetric strain and the vertical strain of the soil sample. The oedometer samples were 70 mm in diameter and 19 mm thick and the triaxial swell test samples were 38 mm in diameter and 76 mm in length. Both the samples were compacted at the same initial moisture content and dry density values and the experiments were performed using various confining pressures. Based on the swell strains recorded it was apparent that the vertical strains encountered in the oedometer swell tests were relatively higher than the vertical strains recorded by the triaxial stress path equipment. The oedometer vertical strains values were almost 1.6 to 2 times higher than the vertical
strains recorded by the tri-axial stress path cell. Therefore, foundation designs based on predicted heave by oedometer swell test method would be overly conservative designs. 2.3 Active zone depth In addition, another conservative assumption inherent in the oedometer swell test method includes the assumption of uniform wetting and vertical boundary conditions with no exponential suction decay with depth. The depth in a soil profile up to which periodic changes of moisture occur is defined as the active zone. As long as a particular area experiences normal extremes of climatic conditions the depth of this zone will not vary. Soils below the active zone do not experience drastic changes in moisture content and thus do not contribute to soil expansion. The definition of an active zone must be broken down into two parts: 1) the moisture active zone and the 2) movement active zone. Here, we are of the opinion based upon theory founded upon Aubeny and Long (2007) and empirical evidence by McOmber and Thompson (2000) that the moisture active zone would be limited from surface infiltration due to gradient and permeability considerations. However, based upon considerations of real estimates of final stress states of the soil including the residual negative pore water pressure that has been measured in the Denver area, it is likely that the moisture active zone will be limited and that the movement active zone will be even more limited due to the residual equilibrium pore water pressure (final suction state). In order to more fully understand the complex soil-water interactions of clay-bearing soils and the resulting soil mass volume and pressure changes associated with these interactions, one needs to examine the phenomena of capillarity and soil suction in unsaturated cohesive soils. The unsaturated zone is the portion of soil mass between the ground surface and the water table. Water in this zone has a pressure head less than the atmospheric pressure, and is retained by a combination of adhesion and capillary action. Soil volume change occurs when the suction changes with time. In other words, volume change occurs only in the unsaturated zone or zone of seasonal moisture change. Figures 2 and 3 shows the typical suction profile and soil suction data of over 26,000 suction tests measured by BCI over the past three years in DFW area, respectively. This figure shows a decay in soil suction profile with depth, which implies that the wetting is not uniform with depth and hence the swelling pressure cannot be uniform over the entire depth. Various case studies have also confirmed suction decay with depth.
Figure 1. Comparison of swell test results from the oedometer and tri-axial swell tests. (Al-Shamrani & Al-Mhaidib, 1998).
1156
Figure 2.
Typical in-situ soil suction profile. Figure 4. content. Suction cohesion with varying moisture
soils vary based on the moisture content. The bearing capacity of unsaturated soils is made up of two parts, which are the bearing capacity of the saturated soil portion and the bearing capacity caused by soil suction (Yongfu, 2003). qu sccNc sq qNq s 2 BN
Figure 3. Seasonal Influence on measured suction values.
1
(1)
The depth of this active zone will vary depending on the location of the soil profile. It does not change at any one single location and on a particular soil profile. However, weathering and subsequent lowering in water table would change the water table but not the heave process. Various researchers have agreed that the active zone depth can easily be determined by plotting the liquidity index against the depth of the soil profile. 3 3.1 BEARING CAPACITY Saturated soil
The bearing capacity of a soil is defined as the maximum load that the soil can support before undergoing shear failure. The bearing capacity of the soil is based on the strength characteristics such as cohesion and friction angle of the soil, depth of embedment, overburden stress, etc. The bearing capacity of saturated soils is a constant values based on the above mentioned parameters. 3.2 Unsaturated soil
where c total cohesion which includes the suction cohesion (cs) and undrained cohesion of the saturated soil, Nc, Nq, & N are the bearing capacity factors, q surcharge, B foundation dimension, and Sc, Sq, & S are the shape factors. Yongfu, 2003 performed several unconsolidated undrained triaxial tests on unsaturated soil samples at different moisture contents and on similar saturated soils. The difference between the unsaturated shear strength and the saturated shear strength was calculated as the suction cohesion. Figure 4 presents the suction cohesion values with varying moisture contents. It can be observed that as the water content increases the suction decreases whereby decreasing the bearing capacity provided by the matric suction. In the same study, it was also found that increasing swelling pressure also decreased the bearing capacity values. The experimental values were also compared with plate load tests, which resulted in good correlations between the measured and predicted values. 4 REALISTIC APPROACH
Although the bearing capacity of the saturated soils is a constant the bearing capacity of unsaturated
As mentioned earlier since the oedometer method provided relatively higher swell strains that what would typically be representative of in-situ conditions, we would have to utilize methods that
1157
incorporate unsaturated soil mechanics principles such as simulating suction decay profiles with depth to predict realistic heave values. Such values would then prevent over conservative design of foundation systems on unsaturated expansive soils. Several researches have reported swell prediction models based on empirical correlations to predict the vertical swelling movements in unsaturated expansive soils. Some of these correlations are as follows (Snethen, 1979) C H ( A Bwo ) log( mf f ) H 1 eo (Hamberg, 1985) H N Hi Ch log h i H i 1 1 eo i (3) (2)
Some of the basic assumption with these correlations (Hamberg, 1985) is that the volumetric strain in expansive soils is mainly controlled by changes in the soil suction applicable in the sitespecific active zones. In addition, the potential heave and respective changes in the water content is directly related to the suction stress changes. In equation (2) the term Ch represents the suction index with respect to the void ratio and h represents the soil suction value. An additional method that incorporates suction change to predict the potential heave of soils is provided by the 3rd edition of the Post-Tensioning Institute manual (PTI Method). The Post Tensioning Institute in 1976, commenced research efforts at the Texas A & M University to develop guidelines for the design of slab-on-grade foundation systems on expansive soils. This included analyses of existing design procedures and computer modeling of post-tensioned foundations in expansive soils, which resulted in publication of the Design and Construction of Post-Tensioned Slabs-onGround in 1980. This design method is commonly known as the PTI method. This was added to the Uniform Building Code Standards and Standard Building Code for the design of ribbed foundation in expansive soils in 1988. The third edition Design of Post-Tensioned Slabs-on-Ground was developed in association with the Texas Department of Transportation and published in 2004. The PTI method for the prediction of potential differential movement is based on several design parameters such as PI, percent Fine Clay, Activity Ratio, Suction Compression Index, Slope of Soil Water Characteristic Curve (SWCC), Unsaturated Diffusion Coefficient, Thornthwaite Moisture Index, Maximum Unrestrained Swell, and Edge Moisture Variation Distance. Most of the
parameters are determined from empirical equations provided in the PTI design publications. The suction compression index for 100% fine clay content is initially calculated from the PI, Liquid Limit (LL), percent fine clay, and mineral classification. The suction compression index can be defined as the change in volume with respect to change in suction of a particular soil. This value is then corrected for the actual percent clay content. The slope of the estimated SWCC curve and unsaturated diffusion coefficient of the soil is calculated based on empirical equations. Finally, the estimated edge variation distance is determined based on the Thornthwaite moisture index value and the weighted modified unsaturated diffusion coefficient. The Thornthwaite moisture index is obtained from agricultural soil sciences based on the average rainfall and evapotranspiration rates for 20 to 30 years. A Thornthwaite index value of zero indicates that the average rainfall equals the average evapotranspiration rate over an extended period. A positive Thornthwaite index value indicates higher rainfall rates when compared to the evapotranspiration rate averaged over a long period. Then the maximum unrestrained swell value (Ym) is calculated based on the constant suction value, maximum suction change, stress change factors, and the suction compression index. A computer program termed VOLFLO 1.5 also calculates the potential differential soil movement applying the unsaturated soil mechanics principal as in the PTI method. In fact, this differential movement is used in the design of post-tensioned slab-on-grade foundation systems by the PTI method. 5 CASE STUDY: POTENTIAL HEAVE PREDICTION
The case study deals with upward differential movement that affected a slab-on-grade foundation system for a residential building located in Frisco, TX. The building was constructed approximately in 2002. Figure 5 provides the floor plan of the residential building. The foundation system supporting the superstructure consists of a ground supported, foundation floor slab with exterior grade beams. The foundation slab was designed based on the soil properties encountered in the preliminary geotechnical investigations. Differential foundation movement and slight distress was reported by the homeowners soon after construction. Therefore, in order to determine the magnitude of movement a geotechnical exploration was conducted at the residential structure. Three geotechnical borings were performed at the approximate locations as shown in Figure 5. Undisturbed specimens were obtained with
1158
Figure 6. building.
Relative elevation survey of the residential
Figure 5. Boring location and floor plan of the residential building.
thin walled Shelby tube samplers and they were extruded in the field before being logged, sealed and packaged by field staff to help protect them from disturbance and to maintain their insitu moisture content. In addition to the geotechnical borings, a recently performed relative elevation survey was available. The relative elevation survey readings plotted in a plan view format is presented in Figure 6. It can be observed form this figure that higher elevation readings were generally recorded along the east and southeast corners of the structure and lower elevation readings were recorded at the west perimeter of the residence. In order to determine the differential movement in the foundation slab, the soil samples from the geotechnical borings were subjected to several laboratory tests such as oedometer swell tests, dry unit weights, moisture contents, Atterberg limits, Unconfined Compressive Strength (UCS) tests, total soil suction, and hydrometer analysis. The moisture content liquidity index values were higher near the surface and decreased with depth, indicative of a net downward movement of moisture. Correspondingly, the total
soil suction values were also lower near the surface, which increased with depth. Therefore, based on the geotechnical data, it most likely appeared that the differential foundation movement had occurred due to moisture-induced heave caused by surficial moisture influences. Since both the oedometer swell test values and the soil suction values were available, the potential heave at this site was calculated based on both the values. The computer program VOLFLO 1.5 was used in order to calculate the potential heave of the soil since construction by incorporation soil suction and unsaturated soil mechanics theory. Some of the input parameters required for the VOLFLO program included the dry densities, initial and final suction, constant suction value and its respective depth % passing # 200 sieve, % passing 2 micron, Atterberg limits of the different soil layers encountered, and Thornthwaite moisture index value. The Thornthwaite index value is based on the average rainfall and average evapotranspiration rates, PostTensioning Institute (2004). The initial soil suction values were determined based on the site conditions extant over the residential lot prior to construction. The initial site conditions were determined from aerial photographs of the site prior to construction. Based on these photographs, significant vegetation was identified over the entire foundation footprint prior to construction. It has been reported that the soil suction values near the surface adjacent to trees are approximately 4.5 pF, Post-Tensioning Institute (2004). Therefore, the initial soil suction was taken
1159
Figure 7.
Initial and final suction profiles.
as 4.5 pF near the surface and the final suction value near the surface was taken as 3.1 pF based on the current total soil suction values encountered in the geotechnical borings. The initial and final soil suction profiles incorporating soil suction decay with depth was based on the unsaturated moisture flow through the soil and the weighted average modified unsaturated diffusion coefficient by the computer program utilizing several empirical formulas. A screen shot of the initial and final moisture suction values simulated by VOLFLO 1.5 is presented in Figure 7. The predicted heave based on the VOLFLO program ranged between 2.4 to 2.5 inches. However, the potential heave at the site predicted by using the oedometer swell test results resulted in relatively higher values ranging between less than 0.5 inches to 4.0 inches. Based on Figure 6 the maximum differential movement identified at the residential building was approximately 2.5 inches. Comparing the predicted heaves by both the suction based method and the oedometer sell test results, it is apparent that the observed differential movement comports well with the heave values by the suction based method. 6 CONCLUSION
higher than the heave values encountered on site in real world conditions. This is mainly due to the complete inundation of the soil sample resulting in complete saturation of the soil sample, which is not typically in-situ at certain depths. Further, due to the lateral restraint on the samples, an appropriate reduction factor must be applied to the predicted heave values in order to incorporate the volumetric change of the soil. Although there are several suction based models and methods to predict moisture induced heave values, the VOLFLO 1.5 program was used based on the prior experience and continued use of the program by the authors. Based on the above case study, the heave values predicted by the suction based method was in comport with the observed differential movement at the residential building. In addition, the swell test method predicted a relatively higher heave value. Therefore, it is recommended that suction based methods incorporating the principles of unsaturated soil mechanics be used to predict potential heave of soils, which would ultimately result in economical foundation designs. REFERENCES
Al-Shamrani, M.A. and Al-Mhaidib, A.I. (1998). Prediction Of Potential Vertical Swell of Expansive Soils Using a Triaxial Stress Path Cell, The Quarterly Journal of Engineering Geology, Vol. 18. Al-Mhaidb, A.I. (1999) Swelling Behavior of Expansive Shales from the Middle Region of Saudi Arabia, Geotechnical and Geological Engineering, Vol. 16. Aubeny, C. and Long, X. (2007). Moisture Diffusion in Shallow Clay Masses, J. of Geotechnical and Geoenvironmental Engineering, Vol. 133. Hamberg, D.J. (1985). A Simplified Method for Predicting Heave in Expansive Soils, MS thesis, Colorado State University, Fort Collins, CO, USA. Kassiff, G. & Ben-Shalom, A. (1971). Apparatus for Measuring Swell Potential under Controlled Moisture Intake, Journal of Materials, Vol. 6, 1, 315. McOmber, R.M. and Thompson, R.W. (2000). Verification of Depth of Wetting for Potential Heave Calculations, Advances in Unsaturated Geotechnics, Geotechnical Special Publication 99, Geo Denver 2000. Nelson, J.D., Miller, D.J. (1992). Expansive Soils: Problem and Practice in Foundation and Pavement Engineering.Wiley, New York. PTI (2004). Design of Post-Tensioned Slabs-onGround, Post-Tensioning Institute, Phoenix, AZ. Snethen, D.R. (1979). An Evaluation of Methodology For Predicting And Minimization of Detrimental Volume Change of Expansive Soils in Highway Subgrades. Rep. No. FHWA-RD-79-49, Final Rep., Washington, D.C. Yongfu (2004). Bearing Capacity of Unsaturated Expansive Soils, J. of Geotechnical and Geological Engineering, Vol. 22.
One of the main factors that influence the design and type of foundation systems is the potential heave value. The potential heave at a particular site can be predicted by several methods including the oedometer swell test method and suction based method. However, it is important to realize the prediction of realistic heave values that would be encountered at the site, which would in turn result in economical design of foundation systems. As mentioned earlier several researchers have proved that the oedometer swell test method results in predicted heave values that are relatively
1160
Identification, characterization and analysis of deep foundation in collapsible soilRegional Prison of Eunpolis, Bahia/Brazil
R.Q. Coutinho, B.P.F. Castro & K.C.A. Dourado
Department of Civil Engineering, Federal University of Pernambuco, Recife, Brazil
ABSTRACT: Foundations implanted in collapsible soil can suffer one additional settlement of considerable magnitude due to the flooding process of the soil. This paper presents results of geotechnical investigation in a Regional Prison construction that presented damages during the execution of the deep foundation. The presence of collapsible soil was considered to be one of the causes of the problem. The objective of this work was to identify, characterize the soil and investigate the possible of influence on the pile foundation. The soil profile basically is sandy clay presenting NSPT values lower than 10 until the depth of 7 m and after that increasing with depth. The soil behavior is presented as moderate to high compressibility and the collapsible soil is classified as being problematic. Two load pile tests were performed with the soil in natural condition and after a flooding process. The curves showed a similar behavior until the work load presenting some difference in the lateral and total failure load. 1 INTRODUCTION is a Regional Prison construction that presented damages (cracks) and broken of concrete piles during the execution of the deep foundation of the buildings. The presence of collapsible soil was considered to be one of the possible causes of the problem. The geology of the Eunaplis city is inserted in a Geological Formation named Barreiras Formation. The Barreiras Formation is constituted of coastal continental sediments of Tertiary age that they form extensive plateaux, frequently cut for falesias next to shoreline. The main lithology constituent in the area are matures sandstone that show varied colors, generally with plan-parallel stratification and crossed stratification. The clay and silt fraction is represented by shales and siltstone intercalated to sandstones in layers with thicknesses varying of some centimeters to sets of ten meters. In Eunaplis city the dry season is from May to September, with August and September the driest months. The annual average rainfall in the last thirteen years observed was 1,283.8 mm. In these thirteen years, the maximum accumulated monthly rainfall observed (405.8 mm) was in November of 1998, being above average historical of this month (186.9 mm). It is observed that the month (February/2008) where the first field activities were carried out is in the rainy period of the region with an average rainfall of 88.7 mm.
Collapsible soils are found in various regions of the world (especially in tropical climates) and encompass a great variety of geological materials. They are characterized by an open unstable structure and water contents less than necessary for saturation. Particles are maintained in their positions in the soil structure by means of bonds capable of providing temporary additional strength. This bond may be created by suction and/or such cementing substances as iron oxides and carbonates. As these forces tend to decrease or even vanish after wetting, shear failure and volume decrease may occur (Coutinho et al. 2004). The foundations implanted in collapsible soil can behave satisfactorily for some time. However they can suffer one additional settlement of considerable magnitude, due to the accidental appearance of a water source that starts to flood through the soil. This paper presents results of geotechnical investigation (laboratory and field tests) in one site, with the objective to identify, characterize the collapsible soil behavior and investigate the possible of influence on the construction foundation (pile foundation) with the wetting of the soil/deposit. 2 CHARACTERISTICS OF THE AREA
The studied area is located in Eunaplis City, State of Bahia, Northeastern of Brazil. The site studied
1161
GEOTECHNICAL CHARACTERIZATION
0 1 2 3 4 5 6 7 8 9 10 11 12 13 14 15 16 17
Classification of the material

0 Silty clayed soft. Sand with silty brown.
SPT N 10 20 30 40 50
Water Content (%) 0 10 20 30 40
The geotechnical investigation campaign of field and laboratory tests was performed in two steps. The first geotechnical campaign consisted basically of borehole SPT, sampling, ExpansCollapsometer (ECT) Tests, characterization tests (grain size distribution and Atterbergs Limits), Proctor normal compaction test and oedometer tests to investigate the collapsible behavior. Figure 1 shows the location of the activities performed in the studied area. The objective of the first campaign was to identify and characterize the collapsible soil and their possible influence in the superficial and pile foundation behavior. 3.1 Soil profile and characterization test
Silty clayed medium to hard (red).
Depth (m)
Silty clayed medium to hard (red)
Clay without silty hard (gray)

NATURAL
water level 14.77m. Limit of the sounding 16.37 m
FLOODED
Figure 2.
Geotechnical profile and water.
Figure 2 presents the geotechnical profile of the area. The results from SPT showed values of N 10, for depths until 6.0 m. The water table level was near from 15.0 m of depth. This figure also shows the water content profiles obtained before and after the conclusion of the ECT tests. The sampling was carried out in February/2008. Five blocks and disturbed samples for accomplishment of laboratory tests were collected in the depth
of 2.53.0 m. The water content obtained during the extraction of the undisturbed samples was in the range of 18%. The soil is composed of sandy clay with silt with liquid limit (wL) in the range of 45.3062.60%. The values of plastic limit (wP) are in the range of 28.0535.81%. 3.2 Compressibility and collapsibility To investigate the compressibility and collapsibility behavior double oedometer tests were performed in in two groups of samples: a. natural water content (EDN tests) and the correspondent soaking tests (EDI tests); b. natural water content in the air dried conditions (EDNSa) and the correspondent soaking tests (EDISa). Figure 3 presents typical curves of the variation of volumetric deformation versus vertical stress with the correspondents parameters of compressibility. It can be observed that the volumetric deformation can be sufficiently significant in function of the vertical stress applied to the soil. The volumetric deformations are approximately of 4% (natural) and 9% (soaking) for stress of 100 kPa; and 7.5% (natural) and 14% (soaking) for 200 kPa. The compressibility characteristics (Cc and Cs) in the natural or flooded condition were obtained and are presented in Figure 3. The studied material is presented as sandy clay with silt. According to the classification of Kulhawy & Mayne (1990) the behavior of the soil from Eunaplis is presented as moderate to high compressibility. It is important to register that the values of the observed compressibility are function of the initial condition of the samples (natural water content, natural water content in the air dried conditions and the correspondent soaked tests).
Figure 1. area.
Location of the field activities in the studied
1162
v (%)
Sandy Clay with Silt

25
Collapse deformation c (%)
5 10 15 20 25 30 1 10 100 1000 10000
Volumetric deformation
20 15 10 5 0 -5
EDNSa EDN EDISa

100000
EDI-EDN EDI-EDNSa EDS 10 kPa 50 kPa 100 kPa 200 kPa 10 kPa dry 200 kPa dry
Vertical Stress (kPa)
Test EDI EDN

Figure 3.
vm (kPa) 46 110
CC 0,383 0,415
Cs 0,031 0,026
10
100
1000
10000
Vertical Stress ( kPa)
Curves v versus v log (3,0 m of depth).
Figure 5. (a) Variation of the collapse deformation c (%) with the vertical stressResults of the double and single.
Volumetric deformation v (%)
10
15
10 kPa 50 kPa
The c values increased with the stress level and depend on the initial condition (water content) of the sample. Figure 5 shows the collapse deformation c obtained from double oedometer tests and points obtained in the single oedometer tests. The results from both tests (double and single) are similar showing increase in c with stress and in the double tests presenting a maximum value, particularly for the natural water content sample. The natural air dried sample presented higher c values as expected. 4
10 100 1000
20
100 kPa 200 kPa
25
IDENTIFICATION AND CLASSIFICATION OF COLLAPSIBLE SOIL
Vertical load (kPa)
vi (kPa) 10 50 100 200 10 200

Figure 4. tests.
Type of test EDS EDS EDS EDS EDSSa EDSSa
c (%) 1.27 1.87 2.55 4.85 1.94 16.78
Curves v versus v log from EDS and EDSSa
The results of a simple oedometer test will indicate whether a soil is collapsible and at the same time give a direct measure of collapse strain potential (c) (Houston, 1996). Figure 4 presents results from single oedometer test in samples of 3.0 m depth.
Many methods for identifying collapsible soils are only qualitative in nature, providing no information on the magnitude of the collapse strain potential. These methods also called indirect methods are typically based on information of dry density, moisture content, void ratio, specific gravity, and Atterbergs limits. In terms of quantitative methods (direct methods) of identification the most common test consists of taking the best quality undisturbed sample possible and subjecting it to a one-dimensional response to wetting test in the laboratory under anticipated field stress conditions. Some variations on the onedimensional test have been utilized, including the double oedometer test proposed by Jennings and Knight (1957) and the single specimen collapse test as described by Houston et al. (1988). 4.1 Indirect methods
Based in previous experiences in other sites (see Dourado et al. 2007) the authors consider the
1163
methods of Basma and Tuncer (1992) and Futai (2000) more indicated for the soil studied. In this way, according to empirical equations (Equations 2 and 3, respectively) the samples were classified according to criterion of Jennings and Knight (1975), which classifies the soil according to gravity of the damages waited in building construction. The soil is classified as Problematic (P) if 5 PC% 10 and Moderate Problematic (PM) if 1 PC% 5. In the cases of the indirect methods of Basma and Tuncer (1992) (Equation 1), the deposit is classified as being serious. According to the proposal of Futai (2000) (Equation 2) this soil is classified as with moderate problem. PC% 48.506 0.072.(S-C) 0.439.wo 3.123.s 2.85.lnvi e Sr c max 4, 2 A(1 IP) PC e 100 1 e0
0 ,6
Proctor test. The soil is classified as potentially collapsible if the value of GC is smaller than 80%. In general the results of GC obtained for the soil from Eunpolis were smaller than 80%, characterizing the soil as potentially collapsible. 5 FIELD TESTS
(1) (2)
(3)
where: e variation of the void index due to flooding under constant tension; and e0 initial void index. 4.2 Direct methods based on oedometer test The direct methods are based on the measure of the potential of collapse of the soil, through the accomplishment of simple and double oedometer tests. From the potential of collapse (deformation of collapse) and of the maximum deformation of collapse the soil is classified in accordance to gravity of the damages waited in a building construction. The criterion of Reginatto and Ferrero (1973) evaluates, from double oedometer tests, the collapse susceptibility of the soil for one determined vertical stress, being as reference the overburden stress and the preconsolidation or yield stress for two conditions limits (natural moisture content and soaked specimen). Considering the value of the coefficient of Collapsibility (C) obtained, the Eunaplis soil is classified as Truly Collapsible (TC), where C 0, i.e., the soils collapses without any external loading. According to the Criterion of Jennings and Knight (1975) this soil is classified as being Problematic (P). According the proposal of Lutenegger and Saber (1988) the soil studied is classified as serious (G). 4.3 Empirical method De Mello (1973) proposed one empirical method based in the degree of compaction (GC) by the
The second campaign of investigation consisted of a field program with complementary borehole SPT tests on natural water content and flooded condition, ExpansoCollapsometer Tests (ECT) and static pile load tests on natural content and flooded condition. Figure 6 presents a comparation between the boreholes profiles from the SPT tests. In this figure also are presented the water content values obtained from the boreholes samples. The values of NSPT were greater for the natural condition until the depth of 9 m. The biggest differences occurred in the depth of 3 m (in the range of 2 to 3 times bigger) and in the depths of 78 m. In the flooded condition the water content values were higher in special until the depth of 11 m. The values of NSPT were greater for the natural condition until the depth of 9 m. The biggest differences occurred in the depth of 3 m (in the range of 2 to 3 times bigger) and in the depths of 78 m. In the flooded condition the water content values were higher in special until the depth of 11 m. Due to some distance between the two SPT performed (natural-SPT09 and flooded-SPT10see Figure 1) it was observed similar soil profiles but with two important differences: a. In the SPT09 was observed the existence of one layer of clayey silt, with sand and gravel, iron concretion (78 m) indicating a laterite soil;
Figure 6. SPT and water content profileSPT09 (natural) and SPT10 (flooded).
1164
b. In the SPT10 it was observed in the depth of 11 meters, a very resisting layer of a soil with NSPT in the order of 50 blows. 5.1 Static pile load tests
The driven precast concrete pile has a square section of 200 mm, projected with characteristic uniaxial strength (fck) of 35 MPa and length of 10.30 m. Two static compression load pile tests were performed in the slow maintained tests condition in accordance to the Brazilian NBR 12131 (1996) standard. The natural load tests were performed applying 10 successive steps loading of 80 kN. The unloading was performed in 4 successive steps, with each smaller than 25% of the total load reached in the loading. The flooded static compression load test was carried in a pile with final length of 10.00 m and the same characteristics of the natural load test. A system of soaking for the flooded load test was used. For soaking the soil a pit was built with drains. This pit has as objective to keep constant a water head of 0.05 m, through of a water reservoir connected to a car. The total time of soaking was 72 hours. During this period was consumed a volume in order of 12,000 liters of water. The flooded load test was started after the ending of the soaking process. Twelve steps of loading were necessary to reach the maximum deformation. For the unloading it was performed a 4 steps, with each load smaller than 25% of the total load. Comparing the profiles of natural and flooded water content (Figure 6) it is observed that flooding revealed efficient until the 10 m, certifying that all the length of the pile was flooded. However due the difference observed between the NSPT in layer of the tip of the piles, the soaking did not reduce the resistance of tip of the flooded pile. Figure 7 presents a comparison between the results of the two load tests. Some points of the curve were used for help in the interpretation and comments: 1. For the vertical work of load the pile design (400 kN) the two curves practically do not present significant differences; 2. Differences in the behavior in the two curves are observed from the applied load of 510 kN; 3. In relation to the settlements, for example, for the load of 640 kN, a settlement of 10 mm is observed in the natural curve and of 15 mm for the flooded curve; 4. The load of 720 kN corresponds to a settlements in the natural curve of 15 mm and in the flooded curve of 20 mm. In this case that
Figure 7. Curves load settlements.
it would be waits bigger difference. However in this point the pile starts to work by the tip and in this layer of tip presents NSPT highly resistant in the order of 50 blows, for the case of the flooded area. The extrapolated curve flooded in the blue color represents the situation where the layer of observed more resistant tip in the flooded borehole had not occurred. Considering in the natural curve, the rupture load of 800 kN, this value would correspond to a settlement of 27.10 mm. For this value of settlement to occur, analyzing the flooded curve corrected (blue hatched line), would be necessary a load of 720 kN. This would represent a reduction in the rupture load of 80 kN in the order of 10%. It is important to register that a situation of intense flooding would be possible with the occurrence of significant internal tubing emptying in the prison. In precipitation terms a great intensity of rain and in a bigger time would be necessary. 6 ANALYSES OF THE LOAD TESTS RESULTS
6.1 Estimate of the failure load In this work, it was used the method of van Der Veen (1953), Dcourt & Quaresma (1978) and Dcourt (1996) for the estimate of the failure load. a) Natural load test (PCNatural) For the estimation of the failure load of the natural load test was considered the borehole SP09 as reference.
1165
According Dcourt & Quaresma (1978) and Dcourt (1996), the profile was separated in layers. The unity lateral resistance (rl) was calculated in accordance with the average value of the NSPT for each layer. For the tip resistance, it was adopted a soil characteristic factor of C 250 kN/m2. In lateritic soils the lateral resistance of the piles can be of 2 to 3 times the resistance calculated by the conventional prediction methods (Dcourt 2002). The RL in the stretch with iron concretion (between 7.0 and 8.0 m) was calculated separately using a factor of increase of 2.5. It was obtained a total lateral resistance (RL) of 475.4 kN and a tip resistance (RP) of 227.0 kN, totalizing a failure load (Qf) of 702.4 kN (Table 1). b) Flooded load test (PCFlooded) For the estimation of the failure load of the flooded load test was considered the borehole SP10 as reference and the same methodology adopted in the natural load for to estimate of the failure load. It was obtained a total lateral resistance (RL) of 291.5 kN and a tip resistance (RP) of 291.5 kN, totalizing a failure load (Qf) of 631.5 kN. The results of the load tests in both conditions are presented in the Table 1. By Dcourt & Quaresma (1978) method, it can be observed in the predict lateral resistance a significant difference: reduction of 475.4 to 291.5 kN corresponding to 38% due the flooding. The value of the flooded tip resistance (340.0 kN) was bigger that in the natural condition due to the tip resistance in consequence of the differences between the soil profiles as described in the item 5 and Figure 6. However it was observed a reduction in the flooded failure of 10% (702.4 to 631.5 kN). The van Der Veen (1953) method was used to estimated the failure load. It was possible a good interpretation in the both curves presenting values of Qf 809 kN (natural) and Qf 836 kN (flooded). van Der Veen method was also applied for the hypothetical case of the corrected flooded curve (Figure 7). The results showed a Qf in the order of 700 kN, emphasizing the conclusion above (reduction of 809 for 700 13.5%). Table 2 presents the results of the failure loads obtained through the three methods utilized in this
Table 1. Summary of the results. Dcourt & Quaresma (1978)results (kN) Load test Natural Flooded Lateral resistance (RL) 475.4 291.5 Tip resistance (RP) 227.0 340.0 Failure load (Qf) 702.4 631.5
Table 2. Estimates of Qf obtained and measured. Rupture load (Qr) (kN) Load test Natural Dcourt (1996) van Der Dcourt & Veen Quaresma Dynamic (1953) (1978) load test 702.4 631.5 770
809 812 RL 504 kN Flooded 836
work (van Der Veen, 1953; Dcourt & Quaresma, 1978 and Dcourt, 1996). The result obtained by a dynamic load test in the same pile statically tested on natural condition is also presented. It can be observed that the values obtained for the failure load were in the same order around of 800 kN. The exception was the prediction by Dcourt & Quaresma (1978) method for the flooded pile. 6.2 Analysis of results of the total failure load For the total failure load the difference is relatively small (10%) but it occurred due to the tip resistance in consequence of the differences between the soil profiles, as described in the item 5 and Figure 6. The values obtained in the static load tests (Table 2) were in the same order of magnitude. However, if did not exist the presence of the very resistant layer below of the tip of the pile flooding, an influence in the order of 10%, after flooding was waited. Similar result is observed in the prediction by Dcourt & Quaresma (1978) method. Some cases reported in Brazilian literature shows a reduction in the results of the total failure load due to the flooding process in the order of 20%. 7 CONCLUSIONS
This paper presents results of geotechnical investigation in a construction site that presented damages during the execution the foundation. The soil profile basically is sandy clay with silt presenting NSPT values lower than 10 until the depth of 78 m and after that increasing with depth. The behavior of the soil is presented as moderate to high compressibility and in terms of collapsibility the soil is classified as being problematic. The flooding process revealed efficient. It is observed with the flooding a significant difference in the predict RL (reduction of 38%). The value of the flooded tip resistance (340.0 kN) was bigger that in the natural condition due to the tip resistance
1166
in consequence of the differences between the soil profiles. Therefore it was observed a reduction in the flooded failure of 10% (702.4 to 631.5 kN). REFERENCES
Basma, A.A. & Tuncer, E.R. 1992. Evaluation and Control of Collapsible Soils. Journal of Geotechnical Engineering Division, ASCE, Vol. 118, n 10, October, p. 14911504. Coutinho, R.Q., Souza Neto, J.B & Dourado, K.C.A. 2004. General Report: Characterization of Nontextbook Geomaterials, Proc. ISC2, Millpress, Vol. 2, p. 12331266. Dcourt, L. 1996. Foundations: Theory and Practice: Chapter 8. PINI Editors, So Paulo, p. 265301. (In Portuguese). Dcourt, L. 2002. Capacidade de Carga de Estacas Executadas no Campo Experimental de Engenharia Geotcnica da U. E. de Londrina. Algumas Ponderaes. XII COBRAMSEG, Vol. III, pp. 15451555, So Paulo. (In Portuguese). Dcourt, L. & Quaresma, A.R. 1978. Capacidade de Carga de Estacas a partir de Valores de SPT. Anais, 6 COBRAMSEF, RJ, Vol.1, p. 4553. (In Portuguese). De Mello, V.F.B. 1973. Soil Mechanics apontments. Escola Politcnica da Universidade de So Paulo (In Portuguese). Dourado, K.C.A., Souza Neto, J.B. & Coutinho, R.Q. 2007. Avaliao dos Mtodos de Identificao e Classificao da Colapsibilidade do Solo. In: VI Simpsio Brasileiro de Solos No Saturados. V. 1. p. 123132 (In Portuguese).
Futai, M.M. 2000. Comportamento Colapsvel de Solos Tropicais Brasileiros. VII Congresso Nacional de Geotecnia, Porto, Vol. 1, pp. 193204. (In Portuguese). Houston, S.L. 1996. Foundations and Pavements on Unsaturated SoilsPart One: Collapsible Soils. Unsaturated Soils, Alonso & Delage eds., pp. 14211439. Houston, S.L., Houston, W.N. & Spadola, D.J. 1988. Prediction of Field Collapse of Soils Due to Wetting. ASCE, Vol. 114, n 1, January, p. 4058. Jennings, J.E. & Knight, K. 1957. The Additional Settlement of Foundations due to a Collapse of Structure of Sand Subsoils on Wetting. Proc. of the 4th ISSMFE, vol. 1, pp. 316319. Jennings, J.E. & Knight, K. 1975. A Guide to Construction on or with Materials Exhibiting Additional Settlement Due to a Collapse of Grain Structure. 4th Regional Conf. for Africa on Soil Mechanics and Foundation Engineering, p. 99105. Kulhawy, F.H. & Mayne, P.W. 1990. Manual on Estimating Soil Properties for Foundation Design, Report No. EPRI EL-6800, Cornell University. Lutenegger, A.J. & Saber, R.T. 1988. Determination of Collapse Potential of Soils. Geotechnical Testing Journal, ASTM, Vol. 11, n 3, September, p. 173178. NBR 12131. 1996. Pilesstatic loading test. Brazilian Association of Technical Norms. (In Portuguese). Reginatto, A.R. & Ferrero 1973. Collapse Potential of Soils and Water Chemistry. VIII ICSMFE, Vol. 2, p. 177183. van der Veen 1953. The Bearing Capacity of a Pile. III ICSMFE, Zurich, Vol. 2, p. 8490.
1167
Pile horizontal bearing capacity in tropical soil under distinct moisture contents
Maycon Andr de Almeida Miriam Gonalves Miguel
University of Campinas, Brazil State University of Londrina, Londrina, Brazil
Sidnei Helder Cardoso Teixeira
Federal University of Parana, Curitiba, Brazil
ABSTRACT: This research presents a study about the behavior of Strauss pile, driven in a tropical soil, when submitted to static horizontal load and different soil moisture contents. The horizontal load tests were performed at the Geotechnical Engineering Experimental Field (GEEF) in the State University of Londrina campus, located in the north of the State of ParanaBrazil. The GEEF geological and geotechnical profile can be defined as a basaltic silty-clay residual soil and it is typical of a large area in the Parana State. The superficial layer consists of a lateritic and collapsible soilapproximately 12 meters thick. Above this layer, there is a saprolitic rigid silty clay soil layer. The water level is approximately 21 meters deep. Several 12 m-long Strauss piles were driven at GEEF soil. These piles were totally embedded in the lateritic and collapsible layer of the soil. Horizontal load tests were performed on two piles at the same time and for three soil conditions: a) in natural content moisture (not wetted); b) pre-wetted for 36 and 12 hours and c) controlled 1-hour wetting, maintaining a constant horizontal load, applied before the wetting, with magnitude equal to the admissible load. The load was applied in a fast manner, so as not to allow horizontal displacements stabilization. Soil matric suction and moisture content were determined during all load tests. Matric suction data was obtained by conventional 1 m-length tensiometers, adequately calibrated and installed in the soil near the piles. Water content data was determined in laboratory, using disturbed soil samples collected 1 m deep into the soil, using a manual auger. 1 INTRODUCTION Tropical soils have properties and mechanical behavior quite distinct from other typical soils. In addition to that, in most cases, it is found in an unsaturated condition, such as lateritic soils (Gidigasu 1976, Nogami and Villibor 1995). An important feature of lateritic soils is their high stiffness. Soils from different regions, but with the same level of SPT resistance may differ in behavior. Predominantly clay soils may show behaviors that are characteristic of sandy soils. As a result, many methods for estimating the foundations bearing capacity underestimate the soil strength, making it necessary to carry out load tests in order to determine the actual behavior of piles installed in such soils. Collapsible soils, on the other hand, have a porous and unstable structure due to the connections between particles by bonds attributed to matric suction and cementing substances, giving to the soil mass an additional resistance in non-saturated conditions. However, when these soils are moist, the bonds between particles are destroyed, causing a sudden
In civil constructions, foundations are frequently submitted to horizontal forces. This occurs, for example, in foundations of bridges, transmission towers, off-shore structures, retaining walls, etc. Horizontal forces can occur due to the action of wind, sea waves, lateral earth pressure, or due to more than one simultaneous effect, as in bridge columns that suffer the action of vehicles braking, wind and water flow. In Brazil, mainly in south and southeast regions, Strauss piles are widely used as structural elements of building foundations and retaining walls. However, these piles are designed based on empirical models that do not take into account the strength and compressibility of soil. This is fundamentally important in areas where lateritic and collapsible soils are found, as in the city of LondrinaBrazil, leading to pile size overestimation and elevating the final work cost as well as an unnecessary use of natural resources.
1169
loss of resistance and, therefore, causing large relative motion between particles (Jennings & Knight 1957, Dudley 1970, Barden et al. 1973, Alonso et al. 1990, Houston 1996, Houston et al. 2001). The behavior of piles subjected to horizontal loading is strongly influenced by the mechanical properties of superficial layers of soil. In the case of piles embedded in collapsible soil, the variation of matric suction in the field, which is more expressive in the more superficial soil layers, significantly affects the geotechnical behavior of the piles. It is, therefore, essential to evaluate the variations of bearing capacity and settlements as function of variation of moisture content for these layers. With that in mind, six load tests were carried out on 12 m long Strauss piles, with 32 cm in diameter, subjected to horizontal loading on Geotechnical Engineering Experimental Field Saburo Morimoto (GEEF), located at the State University of Londrina, in the southern region of Brazil. The geotechnical profile of GEEF is typical of the northern region of the state of Paran, Brazil, constituted by porous lateritic and collapsible clay, about 12 m deep. The horizontal load tests were performed with soil with field conditions, i.e., unsaturated (tests S2, S8 and S9), with pre-wetting for 12 hours (test S10), with pre-wetting for 36 hours (test S2), and with controlled soil wetting around the piles, while they were submitted to allowed load (S6 and S7 tests). The tests were carried out to determine the influence of matric suction in the behavior of piles, i.e., to verify the effect of collapsibility in piles subjected to horizontal forces. The measurement of soil matric suction was made by means of conventional tensiometers, installed around the piles, during the entire horizontal load test. 2 GEOTECHNICAL CHARACTERISTICS OF SOIL IN THE LONDRINA REGION
12 meters thick and is widely used to support small and medium structural elements of foundation. This explains the fact that this layer is well characterized and studied in the region. The second layer, located between depths of 12 to 16 m, is also composed of red silty clay, with firm to hard consistency.
3 3.1
MATERIALS AND METHODS Pile execution
For this study, six 12 m long piles were executed in GEEF, built entirely in the lateritic soil layer. Three of these piles were reinforced with three 8 mm steel bars (S2, S6 and S7) and three with seven 8 mm steel bars (S8, S9 and S10). The steel bars were installed along the entire pile length, with 4.2 mm stirrups, every 30 cm (Figure 1). All piles were 1.15 m long and 32 mm in diameter Dywidag steel bar (ST85/105) centered on the cross section. The piles execution process followed a regional procedure, which consists on using the casing in only 3 meters below soil surface. This is because the local soil presents cohesion and water level is below the pile base, which prevents the shaft from locking during its execution. The regional executive process was used in order to simulate the real situation that occurs in piles subjected to horizontal forces in the Londrina area. Due to the necessity of reinforcing piles along their length, plastic concrete with compression stress value in concrete equal to 20 MPa was used after placing of the reinforcement steel bars 3.2 Horizontal load tests
The used load application system, proposed by Miguel (1996), is composed of a load cell with nominal capacity of 200 kN, a deformation indicator, a manual hydraulic jack with a maximum
The city of Londrina is located in the north of Paran State, southern Brazil, where the bedrock is more prevalently represented by basalt. The subsoil of this region is well represented by the profile of the Experimental Geotechnical Engineering of State University of Londrina, where this research was performed. Based on results of SPT and in laboratory geotechnical characterization tests, Miguel et al. (2002) found that this profile consists on a residual soil of basalt, with four distinct layers, being the first two on unsaturated condition. The first layer (most superficial) consists of silty clay and presents lateritic origin and behavior, with high permeability and resistance due to the presence of iron oxides cementing agents. This layer is
Figure 1.
Location of tested piles.
1170
Figure 2.
Horizontal load application scheme.
capacity of 500 kN and four mechanical extensometers with 0.01 mm sensitivity: one 50 mm and three 30 mm extensometers. Figure 2 shows a scheme of the adopted horizontal load test, using a pair of piles. A steel bar crossed the openings in the load cell and hydraulic jack. The ends of the steel bar were properly connected to the piles shafts, allowing free horizontal pile displacement. This bar was always used in order to maintain the stability of the load application set. The horizontal displacements of each pile were measured using two mechanical extensometers installed in reference beams, fixed in magnetic bases, as shown in the scheme in Figure 2. Fast horizontal loading tests were carried out in accordance with NBR 12131/1991, using a loading time increment of 5 min. The displacement readings on mechanical extensometers for each loading stage were done at time zero and at 5 minutes. The unloading was done in two 10 minute stages. Horizontal load tests regarding field soil moisture condition (unsaturated) were carried out on pairs of piles S8 S2 and S8 S9. The influence of soil collapse in horizontal loading capacity was analyzed through a re-test performed on piles S8 and S2, but with soil pre-wetted around S2 for 36 hours, and by a load test performed between piles S8 and S10, with pre-wetting just around pile S10 for 12 hours. Unlike the other piles, the adopted methodology for testing piles S6 and S7 considered the application of 1 kN load increments in the set until reaching the allowable horizontal load, which was determined by prior load test on another structurally identical pile, with soil in field moisture condition, maintaining this load for an hour up to reach displacement stabilization. After that, by means of two independent hoses, the soil around piles was humidified, with a flow rate equal to 0.060 liters per second during one hour, performing horizontal displacements readings each five minutes. Soil samples were collected from the ground about 1.0 meter below the surface in all load tests, allowing
the soil moisture to be verified during testing. In the load test with controlled soil wetting (S6 and S7) two samples were collected, one before wetting and another after the completion of load test in order to verify changes in soil water moisture. 4 TEST RESULTS AND ANALYSIS
With the data obtained from the tests, load versus displacement curves for each horizontal tested Strauss pile were prepared. They are shown in Figures 3 and 4, and grouped according to their stiffness (3 and 7 longitudinal bars, respectively). Abbreviations used in the graphic are: NW for piles subjected to load tests without soil wetting, W12H and W36H for pile tests with pre-wetting of 12 or 36 hours, respectively, and CW for piles tested with controlled wetting, to verify the influence of suction on the results of horizontal loading tests. Table 1 shows the results of load tests, soil moisture and matric suction readings. On that table, Pu is the horizontal loading capacity, obtained by van der Veen (1953) method and Pc is the collapse load, estimated based on evolution of load versus displacement curves from the horizontal load tests carried out under soil pre-wetting. Analyzing the curves in Figure 3 and the values in Table 1, one can notice a reduction of 25% in the bearing capacity of pile S2 (20 kN to 15 kN) when re-tested with the pre-wetted soil for 36 hours. The average soil water content, obtained from samples collected at one meter depth near pile S2, was 32% in the first test and 57% in the second test. The reading of matric suction obtained in tensiometers installed near pile S2, one meter deep, presented in the first test a constant value of 17 kPa, while in re-test with pre-wetted soil around the pile, the observed suction was 1 kPa (already considering the initial tensiometer zero reading). In the test performed with piles S6 and S7, an average water content equal to 35% was obtained
1171
Figure 3. Curves of load versus horizontal displacement for piles S2, S6 e S7 (lower stiffness).
Figure 4. Curves of load versus horizontal displacement for piles S8, S9 e S10 (higher stiffness).
for the samples collected near pile S6, in the nonwetted (not saturated) condition and, after wetting controlled procedure that content was 55%. Next to pile S7, the average initial water content obtained in the laboratory was equal to 37%, while after soil wetting around the pile, it was 56%. The allowable load considered in the test was 10 kN (using factor of safety equal to 2 and a maximum load capacity of 20 kN, obtained in a load test performed on S2). Load increments of 1 kN were applied in every 5 minutes up to reaching 10 kN. After displacement stabilization (2 mm), the soil wetting was conducted around the piles and tensiometers for one hour. However, during the entire test time, no additional displacement was observed in the horizontal displacements, even though a sharp decline in the matric suction value was observed, as indicated by the tensiometers. After that, new load increments were applied until reaching the maximum capacity of the horizontal piles (20 kN), with the soil already wet, in order to enable the development of load versus horizontal displacement curves.
The matric suction of the soil around piles S6 and S7 in the beginning of the test was approximately 15 kPa. After soil controlled wetting for one hour, it was recorded a matric suction value of 2 kPa. It is likely, therefore, that the 1 hour wetting time is not enough for the soil to present any reaction. The readings obtained for suction piles S6/S7 and S2 are almost equal. However, after a 36 hour pre-wetting period, the collapse was observed at 15 kN with pile S2. Another hypothesis is that the allowable load or horizontal displacements may have not been the critical factors to promote pile collapse. The average horizontal bearing capacity for piles S8 and S9 (Figure 4 and Table 1) was approximately 30 kN, with values of matric suction near pile S8 equal to 21 kPa, throughout the entire test time. While near pile S9 these readings ranged from 22 kPa at the beginning of the test to 25 kPa at the end. The average water content of soil near the two piles was equal to 37%. The load test involving pile S10 was carried out along with pile S8, which, in this case, was used only as reaction. The readings of matric suction of soil under pile S10 (10 kPa), subject to pre-wetting for 12 hours, were not as low as those found in the test performed for pile S2 (1 kPa), when re-tested under pre-wetting for 36 hours. This may possibly have occurred due to the 24 hour period between the two soil wetting procedures. The average soil water content, obtained by soil sampling at one meter below ground surface near pile S10 showed values of around 41%. The collapse load of pile S10 was equal to 27 kN. Based on results obtained in the test with pile S10, it is possible to verify that the decrease in suction due to pre-wetting for 12 hours was not enough to significantly reduce the value of the maximum load in the load test, as it was about 27 kN, only 10% below the average horizontal bearing capacity found for the piles with same stiffness (S8 and S9). In contrast, the S2 pile presented a reduction of about 25%, under a suction of 1 kPa. Table 1 summarizes the results collected during the horizontal load tests carried out in GEEF. The table shows the test results for the soil around the piles in unsaturated, pre-wetted and controlled wetted for 1 hour conditions. It can be verified that for the less stiff piles (S6 and S7) a reduction in matric suction of 95% caused a reduction of approximately 25% on the horizontal bearing capacity of piles, which under field conditions of soil moisture (unsaturated) were about 20 kN and, after the previous wetting, it decreased to 15 kN. However, piles S6 and S7, subjected to loading under controlled soil wetting for 1 hour did not have their load capacity decreased during that time; however, suction dropped about 85% in tensiometers close to them.
1172
Table 1. Results obtained in load tests and reductions referring to load and matric suction regarding load tests perforned with soil under unsaturated and pre-wetted conditons. Soil moisture during horizontal load test Unsaturated and pre-wetted 36 hrs Controlled wetting Controlled wetting Unsaturated Unsaturated Pre-wetted 12 hrs Unsaturated state Pu (kN) w (%) Suction 20 20 20 32 28 32.0 35.5 37.0 37.3 37.1 17 19 13 21 25 Pre-wetted Pc (kN) w (%) Suction 15 20 20 27 57.0 55.0 56.0 41.0 1 3 2 10 Reductions Pu 25% 0 0 10% Suction 95% 84% 85% 60%
Pile S2 S6 S7 S8 S9 S10
CONCLUSIONS
The less stiff piles had an average horizontal bearing capacity of approximately 20 kN in the field soil moisture, suffering small horizontal displacements (maximum 1.0 mm) in loads corresponding to 50% of maximum horizontal loading capacity. For loads higher than 10 kN, these displacements have increased to a maximum of 30 mm. The collapse load for this same pile configuration, when subjected to the soil pre-wetting for 36 hours, was 15 kN. Tensiometer readings taken during the load tests carried out in soil under field moisture and pre-wetted for 36 hours, showed, for the less stiff piles, a reduction in matric suction of 95% while it was verified a reduction of approximately 25% in the horizontal bearing capacity of piles. Piles S6 and S7 were horizontally loaded under filed moisture up to reach the allowable load (10 kN) and the horizontal displacements stabilization under applied load. After stabilization, the allowable load remained applied and the soil was wetted under constant flow rate for 1 hour, in order to verify whether the reduction of soil matric suction corresponded to the destabilization of horizontal displacements. After that time, stages of loading were applied to achieve maximum horizontal displacements of the extensometers, under constant soil moisture. A reduction of 85% in the value of matric suction was observed, but there were no displacement increase during the expected time (1 hour), but only after this time, when further loading stages were applied. Among the possible causes for the low influence of the soil moisture condition over piles under allowable load, two hypotheses may be considered: a) the possibility that wetting was not performed with pile under critical horizontal load or critical horizontal displacement, below which the collapse does not occur, or b) the soil needs more time to respond to a drop in soil matric suction, in order to start the collapse process and to find a balance under the new state of tension.
Under a 12 hour pre-wetting procedure, pile S10, presented a collapse load of 27 kN, i.e., a load capacity reduction of approximately 10%. This small reduction can be explained by the fact that more stiff piles may require larger portions of soil due to pile bending stiffness. REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soil. Geothecnique, 40:(3), 405430. Barden, L.; MCgrown, A. & Collins, K. 1973. The collapse mechanism in partly saturated soil. Engineering Geology, 7, 4960. Committee on Tropical Soils of ISSMFE. 1985. Peculiarities of Geotechnical Behavior of Tropical Lateritic and Saprolitic Soils Progress. Edile: So Paulo/SP. Dudley, J.H. (1970) Review of Collapsing Soils. Journal of the Soil Mechanics and Foundation Division. ASCE, 96: (SM3), 925947. Gidigasu, M.D. 1976. Laterite Soil Engineering: Pedogenesis and Engineering Principles. Developments in Geotechnical Engineering, 9. Elsevier Scientific Publishing Company. Houston, S.L., Houston, W.N., Zapata, C.E. & Lawrence, C. 2001. Geotechnical engineering practice for collapsible soils. Geotechnical Geological Engineering, 19, 355355. Houston, S.W. 1996. Foundations and pavements on unsaturated soilspart one: collapsible soils. Proc., 1st International Conference on Unsaturated Soils, Paris, 3, 14211439. Jennings, J.E. & Knight, K. 1957. The additional settlement of foundations due to a collapse of structure of sandy sub soils on wetting. Proc. IV International Conference on Soil Mechanics and Foundation Engineering. 1, London, 316319. Nogami, J.S. & Villibor, D.F. Pavimentos de baixo custo com solos laterticos. So Paulo, Villibor, 1995. (In Portuguese). van Der Veen, C. 1953. The bearing capacity of a pile. Proc. of the Third International Conference on Soil Mechanics and Foundation Engineering. Zurich, 2, 8490.
1173
The effect of ripping clay soil on swell behaviour

Stephen Fityus, Olivier Buzzi, Michael Holt & Theodore Gunther
ABSTRACT: One approach to mitigate the effects of expansive soil foundation movements on light structures, is to rip the soil foundation prior to the construction of a slab-on-ground, supported on deep piles. This study evaluates the potential for ripped soil to swell or collapse beneath a foundation slab, cast on a ripped soil foundation. A series of tests are described in which a series of different expansive clay soils are prepared to different clod sizes and different initial water contents, prior to being placed in containers of various scales, surcharged with a nominal pressure of 9 kPa and wetted. The results show that the soil response may include swelling, collapse or both, depending upon the soil type, the initial water content, the ripped soil texture and the method of wetting used. The practical limitation of this type of experimental research are discussed in detail. 1 INTRODUCTION allows the soil heave to be accommodated within its own bulk volume. This paper presents an experimental study undertaken to assess the effectiveness of such an approach. 2 BACKGROUND
The swell realized by an expansive clay soil, in response to wetting, depends on many factors. These include the expansiveness of the soil, the initial water content, the amount of wetting that occurs and the stress/strain boundary conditions (nature and extent of confinement). Strategies to control or limit swell usually involve trying to control these factors, and commonly, control of water content change is the aim. Control of swell strain through control of stressstrain boundary conditions is a much less widely considered option. A situation where it arises is that of pile-supported raft slab foundations. The idea behind this approach is that the effects of an expansive foundation can be avoided, if a stiffened raft is suspended on a grid of piles, installed beyond the active depth. Such an arrangement reliably overcomes problems due to soil shrinkage, since the slab remains supported by the piles as the soil pulls away. However, wetting of the soil beneath the raft produces heave that will either lift the slab off the piles, or else cause structural failure of the piles, slab or both. One solution to this problem is to temporarily support the slab on void formers as it is cast, so that future swelling of the soil can fill the voids without lifting the slab. This is achieved by making the void formers from something that collapses after construction, such as cardboard. An alternative approach (Heltech, 2009) is to rip (scarify) the soil with a ripping tyne to achieve a structure of loosely packed clods, interspersed with large macrovoids. The aim of the approach is to either achieve a soil structure that collapses in the long term, or at least, achieve a structure that
As far as the authors are aware, there are no existing studies that have considered the swelling potential of ripped clay soils on a field scale, although a number of studies have looked at the effect that reduced bulk density has on swell strain and/ or swelling pressure. (Popescu, 1986; Uppal and Pallit, 1969; Braudeau and Mohtar, 2006). These, however, have been limited to small scale samples where the soil has been variously recompacted after first being broken down to crumbs of millimeter dimensions. While they have generally established that the potential for swelling is reduced in soils with higher void ratios, and to some extent they have quantified this, their results cannot be automatically extrapolated to field scale soils which are ripped by conventional earthmoving equipment. Assessment of the swell potential of a ripped soil at field scale is not a trivial exercise. As the ripped clods of soil can be as big as tens of centimeters, a large sample is needed to ensure that any outcome is representative of the larger soil mass. In addition to this, the existence of the soil as a matrix of irregular lumps, in point contact with other lumps, leads to a variety of possible soil behaviours in the event that the soil becomes wet. An initially dry soil will take up water readily, and in doing so, swell. A relatively dry soil is also relatively stiff, and so, it can exert a significant force against its confinement,
1175
soil heaves
clod deforms without causing heave
water content
Figure 1. Schematic representation of the balance between swelling pressure and soil softening in a clay clod as it wets.
and potentially, cause vertical and horizontal movements. As the soil becomes wetter still, it continues to increase in volume, but its strength decreases (ie, it softens: Vanapalli, 2009). At some point, the yield strength of the soil clod will be exceeded, and as it swells further, it will be unable to lift the overlying soil, but instead, it will deform plastically and remold to close the interstitial voids. This idea is illustrated schematically in Figure 1. As it is drawn, Figure 1 suggests that clay soil clods should yield relatively quickly and that their potential to produce heave is likely to be small. However, the relative balance between lifting and yielding is poorly understood, and difficult to quantify, being affected by the initial water content, the rate and manner of wetting, the size of the clods and the relationship of the clod to its neighbours (ie, its situation of confinement). 3 APPROACH
Considerable thought and several attempts were given to design an experimental arrangement through which the swell potential of ripped soils could be assessed in a reasonable time (days to weeks) on a large scale (0.51 m sample size). In the context of foundation soils beneath a slab, it is known that such soils generally acquire an equilibrium moisture condition that is relatively wet, when compared to the practical extremes of air dry and field saturation. It would be expected that initially dry ripped foundation soil would become progressively wetter over a 35 year period (Fityus et al. 2004), and potentially, heave as a consequence. In this case, the clods would maintain relatively uniform water contents, as they become wetter, and as a consequence they should have the greater opportunity to produce heave. In seeking to simulate this experimentally, however, a more rapid wetting time was required. If soil clods are wet rapidly, through exposure to free water, then it may occur that the exterior of the clod becomes very wet while the middle remains
dry. A situation can be envisaged where different parts of the same clod may be swelling and lifting, swelling and yielding and not swelling, simultaneously, as indicated by Figure 2. It follows then, that the greatest potential to swell will occur in a relatively dry soil that is wet gradually, so that the rate of moisture diffusion within the clod is greater than the rate at which moisture is delivered to the soil. The first attempt to simulate this behaviour involved blowing humidified air (relative humidity 96% at 20C) through a bulk sample of ripped clay. This was continued for 2 weeks, but in that time, little wetting of the soil was achieved, mostly because the amount of water delivered by this air was relatively small, and because the soil could only extract enough airborne moisture to lower the humidity to 89%. Although this approach could potentially achieve the uniform wetting observed beneath foundation slabs, the approach was abandoned because it would take months or years to achieve a result. The second approach trialed involved injecting steam into the box of ripped soil. Whilst steam seemed to be able to deliver the rapid water content increases desired, wet (cooler) steam had to be used for safety reasons, and so the associated localized free water meant that the wetting was far from uniform. Also, issues were identified with the effect of steam reducing the swelling potential of the clay (Couture, 1985). The third approach trialed, and the one which ultimately yielded meaningful results, involved water delivery to the soil through a type of drip irrigation system. The idea for this came from observations that water could be efficiently drawn up a considerable distance into a fragmented soil through simple capillary suction. So, to achieve an appropriate experimental design, a series of smaller scale scoping experiments on smaller pots of fragmented clay, was firstly undertaken to determine the efficient distance that water could be drawn into a clay soil, and the rate at which it should be delivered.
clod yield stress
swell pressure
wetting and unyielded dry
rapid wetting
wetting and yielded
Figure 2. Schematic representation of a clod with a steep water content gradient leading to locally varying behaviours (refer also to Figure 1).
1176
EXPERIMENTAL ARRANGEMENT
displacement transducers cap metal plate pot stainless mesh
9 kPa
9 kPa
Two clay soils were selected for the experimental work. These were a residual clay from Maryland (Fityus and Smith, 2004) and a black soil from Redbank Plains, Queensland, Australia. The basic soil properties are presented in Table 1. Quantities of each of these were excavated by ripping with a backhoe with a single tyne, and then returned to the laboratory where they were spread on the laboratory floor and allowed to air dry. These soils are typical Australian expansive soils causing problems for residential foundations. 4.1 Small scale pot tests In order to study the capillary conduction behaviour of fragmented soil, as series of swell tests were undertaken using samples of air-dried Maryland clay, with a 1020 mm fragment size. Water was introduced to the base of the samples through a bed of sand that was kept saturated up to its interface with the clay. A surcharge of 9 kPa was added to the soil, to represent a typical foundation pressure beneath a light residential structure. A total of 8 tests were performed, with the tests terminated at different times between 0.25 and 12 days, so that the extent of capillary rise could be evaluated. The experimental arrangement is shown in Figure 3. At the termination of each test, the samples were extracted from the pots in 10 mm layers, and the water content of each layer was determined. The results are shown in Figure 4. It can be seen from Figure 4 that water was readily and rapidly drawn up through the fragmented soil to a height of 50 mm in around 2 days, and to 6070 mm in less than 4 days. Hence, if water is introduced into a soil layer at a particular level, it could be expected to soak up and down and effectively wet an interval of around 120150 mm. The small pot data was further analysed to determine the rate of water uptake, by integrating the water in the soil layers above the base of the sample. Figure 5 shows the volume of water absorbed by the soil in the pot as a function of time. From this data it was determined that water could be delivered initially at a rate of 0.1 ml/min/kg, for a period of around 16 hours, but then, the uptake
ripped soil
water reservior sand
Figure 3. Schematic representation of the small pot tests.
100 90 80 70 60 50 40 30 20 10 0 0 0.25 days 4.79 days 10 20 0.67 days 6.67 days 30 40 water content (%) 1.20 days 11.67 days 2.20 days initial 50 60 2.67 days
Figure 4. Water uptake profiles in the small pot tests.
height above base
300
Volume of Water (mL)
250 200 150 100 50 0 0 1 0.1 ml/min/kg 0.004 ml/min/
5 6 7 8 Number of Days
10
11
12
13
Figure 5. Water uptake rate in the small pot tests.
Table 1. Properties of Maryland and Redbank Plains black soil used in this research. Liquid limit (%) Maryland Redbank Plains 69.8 73.2 Plastic limit (%) 24.1 27.6 Plasticity index (%) 45.7 45.6 Linear shrinkage (%) 14 20 Shrink-swell index (undisturbed) % strain/pF (Fityus et al. 2005) 5.2 7.1 Shrink-swell index (remolded) % strain/pF (Fityus et al. 2005) 5.5 5.3 Air-dry water content % 5.2 15.2
1177
rate dropped to around 0.04 ml/min/kg. This is the necessary information required to design the water delivery system for the large test. 4.2 Large scale box tests
In order to assess the effect of ripping on swell behaviour on a large scale, a reinforced plywood box, 1 m 1 m was fabricated in the Civil Engineering laboratory of the University of Newcastle. The box consisted of 4 rigidly-connected vertical walls and a floating base. The base was supported on blocks that prevented relative downward movement, while allowing upward movement to occur, if required. This aspect of the design was to allow the base to be jacked upward into the box, to simulate the effect of deeper swelling of the intact clay beneath the ripped zone. This aspect will not be considered further here. The testing arrangement is shown in Figure 6 and Figure 7. The sample was placed loosely in the box, starting with a 60 mm layer of ripped soil,
then a sheet of cloth fabric with overlaid 3 mm nylon delivery tubes this was followed by another 120150 mm of ripped soil, then another layer of cloth, etc. A flannelette cloth was chosen to distribute the water laterally, based on trials of several different cloth materials. Water dripped onto a horizontal layer of this cloth was found to spread more than 200 mm from the dripping point, and so, to produce a wetted circle more than 400 mm in diameter. Water was delivered to each layer of cloth through nine small diameter nylon tubes, driven by a peristaltic pump, and spaced out across the cloth so that the 400 mm wetted circles they produced would overlap, to make the entire cloth wet. In each test, three irrigation layers (comprising a cloth sheet and nine drip points) were embedded in around 480 mm of clay soil. Particular care was taken to level the surface of the upper layer, so that a concrete slab, imposing a load of 9 kPa could be placed evenly on the soil. Again, 9 kPa was chosen to be typical of the loads transmitted to the soil through the foundation slabs of lightly
NE Slab
displacement transducers (see inset) slab hanger
NW SP W slab SW SP centre SP
NE SP
SE SP
SE Slab
concrete slab (9kPa)
settlement plate conductive cloth nylon tubes peristaltic pump
silicone tubes
peristaltic pump
s u p p o r t
detached base hydraulic jacks
s u p p o r t
peristaltic pump
reinforced plywood box (1m x 1m)
water reservior
Figure 6.
Schematic representation of the large scale box tests.
1178
centre SP
3 Displacement (mm: up +ve)
2
NW SP
0 0 -1 50 100
NE SP SW SP
150 W slab
200
NE slab SE slab
-2 Time (hours)
Figure 7.
The large scale box tests in operation.
Figure 8.
Results from the box test on Maryland clay.
loaded residential structures. To simulate the situation of a slab sitting on embedded piles, the slab was suspended from hangers that prevented it falling below its initial position, even if the soil, below collapsed. The slab was, however, free to move upward if the soil heaved, and three displacement transducers were mounted to record and such vertical movements. So that any collapse behaviour of the soil below could be monitored, 5 settlement plates (indicated by SP in Figure 6) were positioned through the slab to reflect any under-slab soil movements. These were monitored by displacement transducers mounted on the slab itself, so that only relative soil movements were recorded. 5 RESULTS
3
SW SP
2 Displacement (mm: +ve up)

SE slab
1
NE slab W slab NW SP NE SP
0 0 -1 50 100
150
200
-2
-3 Time (hours)
Tests on both of the soils were completed successfully. In both cases, water was delivered at the faster rate of 0.1 ml/min/kg for more than 12 hours, with more than 80 litres delivered in this time before any free water was observed to seep from the box. At this time, the drip rate was reduced and water continued to be taken up at this slower rate. The effect of ripping on swelling can be assessed from the results for Maryland clay in Figure 8 and for Redbank Plains clay in Figure 9. For Maryland clay, slab settlements of 0.6 to 1.0 mm were measured, representing some small relaxation of the slab hangers as the soil below softened and deformed. The settlement plates either remained stationary relative to the slab, or else they rose up by between 1 and 4 mm, as soil beneath them heaved. These behaviours are consistent with observations that the surface of the clay was not perfectly level at the base of the slab, and that some of the plates were not hard against the base of the slab when the wetting commenced. In summary, ripped Maryland
Figure 9. clay.
Results from the box test on Redbank Plains
clay was not able to heave sufficiently to cause lifting of the 9 kPa slab. For Redbank Plains clay, slab heaving of 0.3 to 1.4 mm was measured, confirming that, despite being heavily voided, the Redbank Plains clay was able to lift the 9 kPa load. The settlement plates displayed a variety of responses, with the complicated response of the NorthWest Settlement Plate (NW SP) firstly being heaved by swelling clay, then falling due to locallised collapse before being gradually lifted again. This behaviour suggests that complex interactions, including rearrangements within the matrix of clods, occur throughout the body of ripped soil as it swells. Final water content profiles from both tests are shown in Figure 10, indicating that relatively consistent distributions were obtained. These results are discussed further in the next section.
1179
Maryland Clay Final Moisture Content Irrigation Layer Initial Water Content 0 Depth from top of sample (mm) 0 50 100 150 200 250 300 350 400 450 500 Moisture Content (%) 20 40
0 Depth from top of sample (mm) 50 100 150 200 250 300 350 400 450 500
Redbank Clay Final Moisture Content Irrigation Layer Initial Water Content Moisture Content (%) 0 20 40
Figure 10.
Water uptake in the box tests.
DISCUSSION AND CONCLUSIONS
The swelling potential of a ripped clay soil is not readily assessed, by experimental means or from a theoretical basis. Whether or not a clod of ripped soil swells against an applied load to produce heave, or it remolds as it swells, depends on many factors, most of which cannot be easily quantified experimentally or theoretically. These include the size, shape and diffusivity of the clods, the stressstrain boundary conditions, the soil strength/water content relationship and the way in which water interacts with the soil. To evaluate the heave potential of a ripped soil experimentally, it is necessary to achieve relatively rapid, uniform wetting of a soil sample that is sufficiently large to reasonably represent the greater soil mass. Achieving this experimentally is not a trivial exercise. Wetting using humid air proved relatively ineffective, as humid air does not carry the load of water vapour required to wet the soil rapidly. Further, despite the relative mobility of the vapour phase in a ripped soil, it appeared that air flow did not permeate all areas of the sample equally. Wetting of the soil using steam proved to be more effective, with steam able to introduce water to the soil at a rapid rate, but it was again observed that the wetting of the soil mass was unexpectedly non-uniform. Wetting through conduction/diffusion proved to be the most effective wetting method of those considered. Despite the relatively small degree of connectivity between the soil lumps in the ripped soil, conduction over distances of 5070 mm proved to be relatively rapid and predictable. Provided that simultaneous wetting was undertaken at several levels in the sample, the results suggest that relatively uniform water content distributions are obtained.
Some consideration should be given to the presence of the cloth fabric layers used to spread the water across the sample, as their tensile strength could potentially affect the mechanical response of the sample. However, given the one dimensional swelling characteristics of the experimental arrangement used, it is considered that the presence of the cloth layers would not have any significant influence on the heaving behaviour of the sample. The results of this study suggest that ripping can significantly reduce the heave potential of an expansive clay soil, and in some cases, it could even eliminate it. From the results of the shrink swell tests, it is estimated that the one dimensional heave strain of undisturbed Maryland clay from an airdry condition is of the order of 9.5%. The test results indicate that this can be eliminated by ripping the soil. Similarly, the one dimensional heave strain of undisturbed Redbank Plains clay from an air-dry condition it is estimated to be of the order of 12.8%. The test results indicate that this could not be entirely eliminated by ripping the soil, but that the heave could be reduced to less than 1.5 mm in 450 mm, or 0.3%. Strains of this magnitude, even developed in a deep soil layer, are readily tolerated by typical light with structures. Hence, the results of the work indicate that there is good potential for soil ripping to be used as an effective strategy to reduce the uplift pressure on foundation slabs. The results are, however, insufficient to conclude that ripping will always be an effective strategy, in every imaginable case. While they indicate that the strategy is likely to control heave in most cases, the slow, uniform wetting required to ensure that the greatest heave potential was realized, could not be achieved perfectly in this research. It should also be considered, in evaluating these results, that ripping may have consequences that could offset some of the benefits reported here. Principally, ripping may have the undesirable side effect of making the soil more permeable, with the consequence that poor drainage could lead to deeper, unripped levels in the soil (ie, below the active depth) being exposed to considerably increased water availability. Hence, broader consideration should be given to the ripping strategy when applied to situations other than those with shallow rock. ACKNOWLEDGEMENTS The authors would like to thank the Heltech foundation company for its financial support of this research and Mr Lachlan Bates and Ms Paola Dutto for their assistance in undertaking the fieldwork.
1180
REFERENCES
Braudeau, E. & Mohtar, R. 2006. Modelling the swelling curve for packed soil aggregates using the pedostructure concept. Soil Science Society of America Journal, 70: 494502. Couture R.A. 1985. Steam rapidly reduces the swelling capacity of bentonite. Nature, 318 no 6041, 5052. Fityus, S., Cameron, D. & Walsh, P. 2005. The shrink swell test. The ASTM Geotechnical Testing Journal, Volume 28. 92101. Fityus, S.G. & Smith, D.W. 2004. The development of a residual soil profile from a mudstone in a temperate climate, Engineering Geology, Vol. 74, No. 1, 3956. Fityus, S.G., Smith, D.W. & Allman, M.A. 2004. An expansive soil test site near Newcastle., ASCE Journal of Geotechnical and Geoenvironmental Engineering, Vol. 130, No. 7, 686695.
Heltech, 2009. http://www.heltech.com.au/media/Stable Slab_SystemOverview.pdf Popescu, M.E. 1986. A Comparison between the behaviour of swelling and of collapsing soils. Engineering Geology, 23: 145163 Uppal, H.L. & Palit, P.M. 1969. Measurement of swelling pressure of expansive soils. In: Proceedings of the Second International Research and Engineering Conference on Expansive Clay Soils, pp. 250255. Vanapalli, S. 2009. Shear strength of unsaturated soils and its applications in geotechnical engineering practice. Unsaturated Soils: theoretical and numerical advances in unsaturated soil mechanics, 579598.
1181
Comparison of bearing capacity of unsaturated sand using Cone Penetration Tests (CPT) and Plate Load Tests (PLT)
F.M.O. Mohamed, S.K. Vanapalli & M. Saatcioglu
Department of Civil Engineering, University of Ottawa, Ottawa, Ontario, Canada
ABSTRACT: The bearing capacity of saturated soils is usually evaluated from Cone Penetration Tests (CPTs) or Plate Load Tests (PLTs). CPTs are commonly used in conventional geotechnical engineering practice due to their reliability, simplicity and associated low costs. CPTs are also performed in unsaturated soils; however, the influence of matric suction towards the contribution of the bearing capacity is not evaluated or taken into account. An experimental program was undertaken in a specially designed equipment to determine the bearing capacity of sand under both saturated and unsaturated conditions using model PLTs and CPTs. These studies demonstrate that the bearing capacity of sands is significantly influenced by matric suction. Based on the studies presented in this paper, simple relationships are proposed between CPTs and the ultimate bearing capacity of sands for both saturated and unsaturated conditions. 1 INTRODUCTION CPTs under both saturated and unsaturated conditions are presented and analyzed. These results are compared with the model footing tests (i.e. model PLTs) conducted on the same sand from an earlier study in the OUBCE (Mohamed & Vanapalli 2006). Simple relationships are proposed for estimating the bearing capacity of sands in both saturated and unsaturated conditions deriving correlations between the PLTs and CPTs using results from the present study and other studies published in the literature. The proposed approach is simple and consistent with the conventional techniques followed in engineering practice towards the estimation of the bearing capacity of sandy soils using CPTs. 2 BACKGROUND
The field or in situ bearing capacity of soils can be determined from the Plate Load Tests (PLTs) or the Cone Penetration Tests (CPTs). The PLTs are reliable but are cumbersome and expensive and hence are not commonly used in geotechnical engineering practice. The CPTs are conventionally used in practice because of their reliability, simplicity and associated low costs. A number of studies are reported in the literature for evaluating the bearing capacity of saturated soils based on the CPTs results using empirical equations (for example, Meyerhof 1956, Eslaamizaad & Robertson 1996, Lee & Salgado 2005, Eslami & Gholami 2006, CFEM 2006). These methods however are not suitable for arid and semi-arid regions where the natural ground water table is deep and soils are typically in a state of unsaturated condition. The bearing capacity of unsaturated soils is significantly influenced by matric suction. Experimental studies show that even low matric suction values of 2 to 6 kPa can increase the bearing capacity of sands by 4 to 6 times in comparison to saturated conditions (Mohamed & Vanapalli 2006). Steensen-Bach et al. 1987 commented that ignoring the influence of capillary stresses in the bearing capacity of unsaturated soils would be equivalent to disregarding the influence of reinforcement in the design of reinforced concrete. In the present study, results of an experimental program conducted on a sandy soil in a specially designed tank referred to as University of Ottawa Bearing Capacity Equipment (OUBCE) using the
There are several correlations proposed by a number of investigators to relate the bearing capacity of saturated sands to CPTs results (Meyerhof 1956, Campanella et al. 1983, Eslaamizaad & Robertson 1996, Lee & Salgado 2005). The above procedures are conservative when they are extended for unsaturated sands. Recently, Russell & Khalili (2006) have conducted CPTs in unsaturated soils and concluded that the matric suction approximately doubles the cone penetration resistance in sandy soils. The cone tip resistance and sleeve friction of CPTs results in both saturated and unsaturated sands were investigated by Mohamed & Vanapalli (2009). The results presented in that study demonstrate negligible sleeve friction contribution to the
1183
load carrying capacity; however, dramatic increase of the cone resistance was observed when CPTs were performed in sand under unsaturated conditions. The significant increase in the cone resistance was attributed to the contribution of matric suction. Relationships consistent to the procedures followed in practice for saturated soils towards estimating the bearing capacity of unsaturated soils from the CPTs would be valuable. An attempt is made in this paper to evaluate the bearing capacity of sandy soils both in saturated and unsaturated conditions from CPTs. Such techniques would encourage extending the mechanics of unsaturated soils in geotechnical engineering practice, particularly for the design of shallow foundations in sandy soils where the ground water table is at a greater depth from the natural ground surface. 3 EQUIPMENT AND METHODOLOGY
LVDT
LC
Sand
430mm 20mm
430mm 430mm 430mm 900mm
PLT
CPT
100mm
40mm
CP
60 (a)
(b)
900mm
Figure 2. Schematic of top view of the UOBCE along with; (a) section of the model plate load; (b) model cone penetrometer.
Table 1. Properties of the tested soil. Property Coefficient of uniformity, Cu Coefficient of curvature, Cc Average dry unit weight, d, kN/m Average relative density, % Void ratio, e (after compaction) Effective cohesion, c, kPa Effective internal friction angle, , Value 1.83 1.23 16 65 0.620.64 0.6 35
A schematic of test setup designed to determine the variation of bearing capacity of sands with respect to matric suction using the model PLTs and the CPTs in the UOBCE is shown in Figure 1. The setup consists of a rigid-steel tank of 900 mm-length 900 mm-width 750 mm-depth. The test tank can hold up to 1000 kg of soil and the capacity of the loading machine is 15 kN. The schematic of the top view of test tank along with sections of the model CPT and the model PLT is shown in Figure 2. A strain rate of 1.2 mm/min was used for loading the cone in the present study. An equivalent strain rate was also used for conducting the model
1500mm
PLTs in the UOBCE for determining the bearing capacity of the same sand from an earlier study by Mohamed & Vanapalli (2006). More details about the equipment design, setup and testing procedure are available in Mohamed & Vanapalli (2006), Mohamed & Vanapalli (2009). 4 SOIL PROPERTIES
Reaction Loads
Reaction Loads
120mm
Loading LVDT LC
2465mm
Equipment
The soil used in the present study can be classified as poorly graded sand, SP according to the USCS. The soil has approximately 5% of silt. More details of the properties of the soil used in the present research program are summarized in Table 1. 5 EXPERIMENTAL RESULTS
50mm
Tensiometer DAQS
750mm
40mm
900mm Containers
5.1 Cone Penetration Tests (CPTs)

PC
Sand
Valve
Base
2240mm
Figure 1. Schematic of the University of Ottawa Bearing Capacity Equipment (UOBCE).
Mohamed & Vanapalli (2009) conducted a number of CPTs on compacted sand in the UOBCE under saturated and unsaturated conditions. The first series of tests were carried out under saturated conditions (i.e. matric suction value equal to zero) and the second series of tests were conducted under unsaturated conditions (i.e. matric suction values of 1 kPa, 2 kPa and 6 kPa). Figure 3 shows a crosssection of the test tank along with the placement of Tensiometers and other accessories including
460mm
1184
CPT LVDT LC Tensiometer
Piezometer (clear pipe)
10 7kPa 150 4kPa 300 2kPa 500 600 1kPa
Soil surface
5.8kPa 4.4kPa 2.8kPa
IZ ~ 1.5B
WTL
WTL for Test 4
1kPa
900mm
Measured matric suction using Tensiometers Hydrostatic variation of matric suction
Figure 3. Schematic illustrates the procedure used for estimating the average matric suction within the influence zone, IZ.
0.0
Cone resistance, qc (MPa)

1.0 2.0
0 50 100 Penetration, d (mm) 150 200 250 300 350 400
3.0
4.0
6 kPa
2 kPa
0 kPa (saturated) 1 kPa
Figure 4. Variation of the cone resistance from CPTs with penetration depth under saturated and unsaturated sand conditions.
which stresses are predominant for shallow square footings (Poulos & Davis 1974). The average matric suction value is considered to be the matric suction value at the centroid of the suction distribution diagram from 0 to 1.5B depth region in the analysis of the results (see Figure 5). This technique has been extended successfully for interpreting the bearing capacity of unsaturated soils for both coarse- and fine-grained soils (Mohamed & Vanapalli 2006, Vanapalli & Mohamed 2007 and Vanapalli et al. 2007). The variation of the measured matric suction with respect to depth is typically nonlinear for unsaturated soils (see left hand side of Figure 5). The variation of matric suction with depth for the tested sandy soil however may be approximated to be linear in nature (i.e. variation was similar to hydrostatic equilibrium condition with respect to natural ground water table; see right hand side of Figure 5). Such a conservative assumption can be extended to other sands when matric suction values are not determined. The bearing capacity values of the compacted sand using the model PLTs at different average matric suction values are plotted in Figure 6. The measured bearing capacity values of the compacted unsaturated sand was 4 to 6 times higher than the bearing capacity values of the same sand under saturated condition. Vanapalli & Mohamed (2007) proposed a semiempirical model for both interpreting and predicting the variation of the bearing capacity of a sandy soil with respect to matric suction under drained loading conditions using the PLT results. This model uses the saturated shear strength parameters (c and ) and the Soil-Water Characteristic Curve (SWCC).
PLT LVDT
(ua- u )
w 1
750mm
Piezometer (clear pipe)
the LVDT and a load cell. Figure 4 shows the experimental results of the variation of cone resistance, qc with respect to the penetration depth for the tested sand under both saturated and unsaturated conditions. 5.2 Model Plate Load Tests (PLTs)
LC Tensiometer
10 150 7kPa 300 4kPa 2kPa 500 600 1kPa 5.8kPa Stress 4.4kPa
Bulb Zone
2.8kPa 1kPa
w 2
Mohamed & Vanapalli (2006) conducted a number of PLTs using model footings on the same sand under saturated and unsaturated conditions. This paper also details experimental procedures for conducting these tests in the OUBCE along with other testing details. The model PLTs results were analyzed considering the influence of average matric suction value in the vicinity of the stress bulb zone which is equal to depth 1.5B (i.e. 1.5 times the width of the plate used for testing). This is also the depth zone in
WTL
WTL for Test 4
900mm
Measured matric suction using Tensiometers Hydrostatic variation of matric suction
Figure 5. Schematic illustrates the procedure used for estimating the average matric suction below the PLTs (Mohamed & Vanapalli 2006).
1185
750mm
(ua- u )
1200
Bearing capacity, qult (kPa)
1000 800 600 400 200 0
Measured BC using 100 mm 100 mm PLT
Best fit
More recently, Lee & Salgado (2005) have proposed another relationship to estimate the ultimate bearing capacity from CPTs data as shown below. qbL (qcAVR) (3) where qbL limit unit bearing capacity, qc AVR average of qc values from the footing base to a depth from footing base equal to footing diameter B; and constant function of lateral earth pressure ratio, K and relative density, Dr.
10
Saturated bearing capacity 0 2 4 6 8 Matric suction, (ua - uw) (kPa)
Figure 6. Measured bearing capacity versus matric suction relationship for 100 mm 100 mm model PLTs.
6.2 Relationships for estimating the bearing capacity of sands under both saturated and unsaturated conditions from CPTs Empirical methods to correlate CPTs and the ultimate bearing capacity of shallow foundations valid for both saturated and unsaturated soils are valuable for practicing engineers. As a first step in this direction, Mohamed & Vanapalli (2009) carried out CPTs in a laboratory environment to propose such correlations providing comparison with the model PLTs results. Three different influence zone depths were considered in the analysis (i.e. referred to as Case (i), Case (ii) and Case (iii)) to provide comparisons between the cone penetration resistance, qc and the measured bearing capacity from the model PLTs (see Figure 7). A difference of 20% to 50% between the CPTs results and the PLTs results was observed when the sand was under unsaturated condition for the influence zone depths, IZ presented in Case (i) and Case (ii) respectively. However, better comparisons were possible between the cone resistance, qc and the bearing capacity values from the model PLTs using the influence zone depth, IZ is 1.5B or slightly greater (see Case (iii)). In the present study, a penetration depth (i.e. influence zone, IZ) for averaging qc values is chosen to be equal to 1.5B after performing several trial studies on the sandy soil in saturated and unsaturated conditions. This depth corresponds to 150 mm in the present study (i.e. 1.5 B 150 mm;
D B Model Footing (PLT) Soil Surface Stress Bulb Case (iii) 1.5B 1.5B
The measured bearing capacity values from the model PLTs are compared with CPT results in later sections to show the influence of matric suction (Figure 6). More details of the model PLTs results and discussions are available in Mohamed & Vanapalli (2006). The results of these studies were used to propose simple equations to estimate the bearing capacity of both saturated and unsaturated sands from conventional CPT results. 6 CORRELATIONS BETWEEN THE CPT AND THE BEARING CAPACITY OF SANDS General
6.1
Meyerhof (1956) proposed a correlation between the cone penetration resistance, qc and bearing capacity of shallow foundations, qult for saturated sands as follows: q D qult c ( B ) 1 B C (1)
1.5D D D
Case (i)
Case (ii)
B qult 0.1849(qc ) D
0.5093
B 1 F D
(2)
D = Cone diameter (40 mm) B = Width of model footing (100 mm)
where qult ultimate bearing capacity at relative displacement of 10%; qc average cone resistance value over a depth of B; B and D width and depth of footing respectively; F shape factor and N /Nq.
Figure 7. Schematic illustrates three different cases (i.e. Case (i), Case (ii) and Case (iii)) of influence zones (modified after Mohamed & Vanapalli 2009).
1186
1.5D D
where qult ultimate bearing capacity; qc arithmetic average of cone resistance values in the influence zone equal to 1.5B; B footing width; C constant equal to 12.2 and D footing embedded depth. Eslaamizaad & Robertson (1996) have proposed a relationship to estimate the ultimate bearing capacity of shallow foundations in sands from the CPTs data. The correlation between qult and qc was expressed as:
Cone Penetrometer (CPT) D D
Bearing capacity,qult or cone resistance, qc (kPa)
note B 100 mm width, see Figure 5). The average CPTs values over a depth IZ 150 mm were used to develop correlations with the measured bearing capacity values from model PLTs including some field data from the literature. The depth of penetration of 1.5B is consistent with the depth zone in which the stresses are predominant (Poulos & Davis 1974, Chen 1999). The influence zone (i.e. a depth of 1.5B from the footing base) provides good correlations between average qc and ultimate bearing capacity of sand in both saturated and unsaturated conditions. This depth also provides requisite confinement to the shaft to which cone is connected. Similar influence zone depth of 1.5B was used by Meyerhof (1956) and Schmertmann (1978) to relate the ultimate bearing capacity of shallow footings to an average of cone penetration resistance value. Two different equations are proposed based on the CPTs results to estimate the bearing capacity of shallow foundations of surface footings on saturated and unsaturated homogenous sand respectively. Equation (4) is suggested to estimate the ultimate bearing capacity for saturated sands based on the studies undertaken through the present research program (i.e. (ua uw) 0 kPa) from CPT results: qult (sat) (qc sat) (4) where qult (sat) ultimate bearing capacity for saturated homogenous sand, 0.15/B0.63 (i.e. correlation factor), qc sat average cone resistance under saturated sand condition (e.g. within influence zone, IZ equal to 1.5B from the footing base level) and B footing width. Similar forms of expressions as Eq. 4 were suggested in the literature by various investigators. For example, Meyerhof (1956) and Eslaamizaad & Robertson (1996) suggested correlation factors range from 0.17 to 0.25 for estimating the bearing capacity of saturated sands using CPTs. A correlation factor value of 0.1 was recommended by the Canadian Foundation Engineering Manual (CFEM) for the evaluation of allowable bearing capacity of a soil using CPT results. Equation (5) is suggested to estimate the ultimate bearing capacity for unsaturated sands (i.e. (ua uw) 0 kPa): qult (unsat) (qc unsat) (5) where qult (unsat) ultimate bearing capacity for unsaturated homogenous sand, 0.19/B0.68 (i.e. correlation factor), qc unsat average cone resistance under unsaturated sand conditions (e.g. within influence zone, IZ equal to 1.5B from the footing base level), B footing width. The two correlation factors and presented in Eq. 4 and Eq. 5 respectively are derived from regression analysis of the experimental results obtained from model PLTs and CPTs for the same sand under saturated and unsaturated sand conditions.
TEST RESULTS AND DISCUSSIONS
Studies undertaken through this research program show that a correlation factor, value of 0.1 to 0.64 is required using Eq. 4 for estimating the bearing capacity of sands that are in a state of saturated condition. The factor, for reliable estimation of the bearing capacity of unsaturated sands using Eq. 5 lies between 0.1 and 0.9. The need for using a larger range of values in comparison to demonstrates the contribution of matric suction towards the bearing capacity of unsaturated sands. The factors, and for saturated and unsaturated sands respectively are dependent on the footing width B. In both the cases, the correlation factor increases as the footing width, B decreases. Figure 8 shows the measured results from CPTs for both saturated and unsaturated conditions over
1000 800 600 400
Measured BC from PLTs
Average qc over IZ = 1.5B
200 0
Estimated Sat. BC using Eq. 4 Estimated Unsat. BC using Eq. 5
2 3 4 5 Matric suction, (ua - uw) (kPa)
Figure 8. Typical results of the estimated BC using the proposed equations and the measured BC from 100 mm 100 mm PLTs with respect to matric suction. Table 2. Measured BC from PLTs or footing and estimated BC from average CPTs. Soil type SP SP SP SP SP SP S# S# S# Matric suction (kPa) 0 2 6 0 2 6 33 25 20 qc AVR (kPa) 118 565 805 270 900 1235 2800 3530 10000 Measured BC (kPa) at S 25 (mm) 121* 570* 840* 125** 630** 745** 500 525 950 Estimated BC (kPa) 81 (Eq. 4) 504 (Eq. 5) 769 (Eq. 5) 133 (Eq. 4) 621 (Eq. 5) 852 (Eq. 5) 532 (Eq. 5) 419 (Eq. 5) 900 (Eq. 5)
* 100 mm 100 mm; ** 150 mm 150 mm; 1000 mm 1000 mm; 2000 mm 2000 mm; 3000 mm 3000 mm; Estimated average suction (along influence zone depth of 1.5B) values assuming hydrostatic profile. This is based on extending the assumption of suction profile results from the present study which is close to hydrostatic in nature; # Sand from Lee & Salgado (2005).
1187
1250
Estimated bearing capacity, BC (kPa)
Estimated BC = Measured BC 1000 750 500 250 R = 0.94 0 0 250 500 750 1000 1250 Measured bearing capacity, BC (kPa)
2
AKNOWLEDGMENTS The first author thanks the General Secretariat of Higher Education in Libya for the financial support. REFERENCES
Campanella, R.G., Robertson, P.K. & Gillespie, D. 1983. Cone penetration testing in deltaic soils. Canadian Geotechnical Journal 20(1): 2335. Canadian Foundation Engineering Manual, CFEM (4th ed.). 2006. Canadian Geotechnical Society. Vancouver, Bitech Publisher, P 512. Eslaamizaad, S. & Robertson, P.K. 1996. Cone penetration test to evaluate bearing capacity of foundations in sands. Proceedings of 49th Canadian Geotechnical Conference, St, Johns, NF: 429438. Eslami, A. & Gholami, M. 2006. Bearing capacity analysis of shallow foundations from CPT data. Scientia Iranica 13(3): 223233. Lee, J. & Salgado, R. 2005. Estimation of bearing capacity of circular footings on sands based on cone penetration test. Journal of Geotechnical and Geoenvironmental Engineering 131(4): 442452. Meyerhof, G.G. 1956. Penetration tests and bearing capacity of cohesionless soils. Journal of the Soil Mechanics and Foundations Division, ASCE 82 (SM1): 119. Mohamed, F.M.O. & Vanapalli, S.K. 2006. Laboratory investigations for the measurement of the bearing capacity of an unsaturated coarse-grained soil. Proceedings of 59th Canadian Geotechnical Conference, Vancouver, BC: 219227. Mohamed, F.M.O. & Vanapalli, S.K. 2009. An experimental investigation of the bearing capacity of an unsaturated sand using cone penetration tests. Proceedings of 62nd Canadian Geotechnical Conference, Halifax, NS: 205213. Poulos, H.D. & Davis, E.H. 1974. Elastic Solutions for Soil and Rock Mechanics. New York: John Wiley and Sons. Russell, A.R. & Khalili, N. 2006. Cavity expansion theory and the cone penetration test in unsaturated sand. Geotechnical Special Publication, ASCE 147 (2): 24562556. Schmertmann, J.H. 1978. Guidelines for cone penetration test, performance and design, Report No. FHWA-TS78-209, US Department of Transportation, Washington, D.C., P. 145. Steensen-Bach, J.O., Foged, N. & Steenfelt, J.S. 1987. Capillary induced stresses-fact or fiction? 9th ECSMFE, Groundwater Effects in Geotechnical Engineering, Dublin: 8389. Terzaghi, K. 1943. Theoretical Soil Mechanics. New York, USA: John Wiley and Sons. Vanapalli, S.K. & Mohamed, F.M.O. 2007. Bearing capacity of model footings in unsaturated soils. Proceedings of the 2nd International Conference on Unsaturated Soils, 79 March, Weimar, Germany: 483493. Vanapalli, S.K., Oh, W.T. & Puppala, A.J. 2007. Determination of bearing capacity of unsaturated soils under undrained conditions. Proceedings of the 60th Geotechnical Conference, Ottawa, ON: 10021009.
Figure 9. Comparison between estimated bearing capacity using the proposed equations and measured bearing capacity obtained from experimental and field data.
an influence zone (IZ) of 1.5B depth. In addition, the bearing capacity values of sands calculated using Eq. 4 (for sands in saturated condition) and Eq. 5 (for sands in unsaturated conditions) using CPTs data are also plotted along with the model PLT results. There is good comparison between the estimated bearing capacity values (from CPTs) and the measured bearing capacity values (from PLTs) (see Figure 8). Table 2 provides comparison between the estimated bearing capacity values (e.g. using Eq. 4 for saturated sands and Eq. 5 for unsaturated sands) and the measured bearing capacity values from six model PLTs (from this study) and three footing test results reported in literature by Lee & Salgado (2005). The results (i.e. plotted in Figure 9) show reasonably a good comparison between the estimated and measured bearing capacity values. 8 SUMMARY AND CONCLUSION
The CPTs are simple and inexpensive to conduct in the field and hence are conventionally used in engineering practice to determine the bearing capacity of soils. These results are also reliable in the estimation of the bearing capacity compared to Terzaghi (1943) equation as uncertainty associated with the measurement of the internal friction angle, and estimating the bearing capacity factors (i.e. Nc, Nq and N) can be alleviated. In this paper, simple relationships are proposed between the average cone resistance, qc and the ultimate bearing capacity, qult of sandy soil in saturated and unsaturated conditions. The proposed relationships provide reasonable comparisons between the estimated and measured bearing capacity values for sands in both saturated and unsaturated conditions. Further research work is however necessary to be conducted to check the validity of the proposed relationships using different types of sands both in the laboratory and in field. Several studies in this direction are in progress at the University of Ottawa, Canada.
1188
Water content profiles for design of foundations on expansive soils

D.D. Overton, K.C. Chao & J.D. Nelson
ABSTRACT: Construction of light structures in arid climates generally results in an increase in water content of the subsoil profile due primarily to irrigation and a reduction of evapo-transpiration. Water introduced by irrigation at the surface migrates downward and increases the water content at depth as it advances. Although the subsurface is wetted significantly, the soil above the wetting front may not become saturated. If the wetting front encounters an aquatard, a perched water condition may develop and the soils above the aquatardmay become saturated. Determination of a final water content profile for design of foundations on expansive soils is an important design parameter. In the absence of other data, it is conservative to assume that the entire profile is saturated. Alternately, modeling of water migration in the vadose zone can be performed to determine water content profiles for various site conditions. The results of this rigorous analysis can be used to determine a final water content profile for use in optimizing the design of foundations. However, this rigorous analysis is not always economical for small projects. Thus, a simplified procedure is presented that can be used to hand calculate final unsaturated water content profiles for design of foundations on expansive soils. The results of the water migration modeling and hand calculation methods are presented and the applicability of these methods is discussed. The resulting foundation design based on the final water content profiles determined from these two methods, as well as the assumption of full saturation, is presented and discussed. 1 1.1 INTRODUCTION Definition of expansive soils it is appropriate to base the foundation design on the amount of wetting and resulting heave that will occur at the end of the design life. A design life of 100 years for foundations was considered in this paper (Schmalz & Stiemer, 1995). Design of foundations for expansive soils must consider the timewise migration of the water in the vadose zone, and the extent to which subsoil wetting will occur during the design life of the structure. These factors will govern the amount of heave that will occur over the design life. 1.2 Zones of subsoil wetting
The design of foundations for sites where expansive soils exist is one of the most important aspects of geotechnical engineering in arid areas. If the heave of a foundation on expansive soils reaches an intolerable amount the structural integrity of the foundation and the structure that is supported by the foundation may be compromised. Residential development in arid climates often results in an increase in water content of the soil profile due primarily to irrigation and a reduction of evapo-transpiration. Water introduced by irrigation at the surface migrates downward and increases the water content at depth as it advances. If the wetting front encounters an aquatard, a perched water condition may develop and the soils may become saturated. For these conditions, it is common practice to consider that the soils will be saturated for the entire depth of potential heave for design of foundations. For cases where heave can extend to large depths and no aquatard exists, the subsurface can also be wetted significantly. However, the soil above the wetting front may not become saturated. Additionally, the time required for the wetting front to migrate through the entire depth of potential heave may exceed the design life of the structure. Thus,
The Active Zone is a term that has been used differently in various applications, and confusion frequently arises as to its intended meaning. There is a need for a clear definitive designation of the active zone and other associated zones. The following four definitions have been put forth (Nelson et al. 2001): Active Zone, za, is that zone of soil that is contributing to heave due to soil expansion at a particular point in time. The depth of the active zone will vary as heave progresses, and therefore, it varies with time. Zone of Seasonal Moisture Fluctuation, zs, is that zone of soil in which water contents change seasonally due to climate changes.
1189
Zone of Wetting, zw, is that zone in which water contents have increased over the pre-construction equilibrium conditions. Factors contributing to this could include capillary rise after the elimination of evapo-transpiration from the surface, infiltration due to irrigation or precipitation, or introduction of water from off-site. Underground sources may include broken water lines, development of perched water tables, or flow through permeable strata that are recharged at distant locations. Depth of Potential Heave, zp, is the depth to which the overburden vertical stress equals or exceeds the swelling pressure of the soil. This represents the maximum depth of the Active Zone that could occur.
2 2.1
CONCEPTUAL UNSATURATED FLOW MODELS Saturated final water content profile
of saturation of the soils. However, as shown in previous studies (Chao et al. 2006; Overton et al. 2006), the degree of wetting typically will be less than fully saturated, and may not extend to the depth of potential heave. Design of foundations for these conditions must consider the design life of the structure, the depth of wetting that can occur during the design life, and the degree of saturation, and thus the portion of potential heave that can develop during the design life. The same conditions also exist for collapsible soils as noted by Houston et al. (2001): In contrast, full collapse potential would not be expected in zones that are wetted by downward infiltration only and which are not submerged by groundwater The regions that experience complete submergence would experience full collapse potential, but the zones of wetting by downward infiltration alone would exhibit collapse consistent with the partial collapse relationship for that particular soil. 3 FINAL WATER CONTENT PROFILES
Development in arid regions typically results in construction of slabs-on-grade which reduce evaporation, and irrigation which exceeds the evapotranspiration of the yard. These factors as well as others result in the development of a wetting front. Below the wetting front the water content is the same as that which existed prior to introduction of the water source. However, above the wetting front the water contents are higher, and the soil typically will be unsaturated. The difference in soil suction between the wetter and drier zones will result in downward flow of water, and the wetting front will continue to move downward until an impermeable boundary or a water table is reached (McWhorter & Nelson, 1979). Once an impermeable boundary is reached, the soils will be saturated and the water table will propagate back upward to the surface. Full wetting of the soil profile would be expected only when rising groundwater or perched water conditions develop. Where a rising groundwater table is anticipated, the full wetting conditions should be used to make calculations (Houston et al. 2001). Thus, if a water table or an impermeable boundary is within the depth of potential heave and the groundwater is anticipated to rise to the elevation of the surface wetting profile, then it would be reasonable to assume the entire depth of potential heave could be fully wetted during the design life of the structure. 2.2 Unsaturated final water content profile
The depth of wetting and corresponding degree of saturation can be calculated using readily available computer codes, such as Vadose/W, SVFlux, and Hydrus 2-D. Using the results of these analyses, the amount of heave that occurs in the partially wetted zone can be calculated. This amount of heave will be less than that calculated assuming wetting and saturation for the entire depth of potential heave. Alternatively, simple hand calculations can be performed to determine the final water content of the soils. However, the calculations do not readily led themselves to determine the depth of wetting with time. Thus, these calculations will provide a final constant water content profile, but that profile must be used for the entire depth of potential heave. The use of the constant water content profile estimated from the hand calculations will result in an overly conservative foundation design if the actual depth of wetting does not migrate through the depth of potential heave during the design life of the structure. 3.1 Saturated final water content
For sites which are subject to irrigation and other surface and near surface water sources, and which do not have an aquatard at depth, the wetting front will move downward increasing the degree
The entire depth of potential heave should be assumed to be saturated during the design life of the structure if a water table or an impermeable boundary is within the depth of potential heave. Additionally, if calculations are not performed to determine a depth of wetting or degree of saturation, then it should be assumed that the entire depth of potential heave would be saturated. A saturated volumetric water content equal to the porosity can be used for design in these conditions.
1190
3.2
Vadose zone modeling

0
10
15
20
25
30
35
40
Analyses of the migration of the wetting front were conducted using the computer program Vadose/W Version 7.15 (GEO-SLOPE, 2007). Vadose/W is a finite element program that can be used to model movement and distribution of pore water within porous materials such as soil and bedrock. Vadose/W can model both saturated and unsaturated flow in response to climatic and environmental conditions, making it possible to analyze water migration as a function of time. The program considers the effect of precipitation and infiltration, surface seepage, runoff and ponding, plant transpiration, potential and actual evaporation, snow accumulation and melt, ground freezing and thawing, and groundwater recharge. A one-dimensional profile was used for analyses that consisted of homogeneous claystone. Typical profiles in the Front Range of Colorado would consist of soil or weathered claystone over claystone. However, the soil or weathered claystone typically do not extend to great depths below the basement excavation. Thus, the results of the homogeneous claystone layer presented herein are represent of conditions at a basement elevation. Initial water contents prior to development, post-development landscaping and grading, post-development irrigation values, and climate data applied were determined for an actual subdivision and were applied to the model. Climate data were generated using the ClimGen program (University of Washington, 2002) after training the program with approximately 20 years of historical data from a nearby NOAA certified weather station. The soils properties used in the analyses are presented in Table 1. The Soil Water Characteristic Curve (SWCC) for the claystone used in the analysis was obtained using the filter paper test. The unsaturated hydraulic conductivity function of the claystone was estimated based on the SWCC of the claystone using the Fredlund et al. (1994) equation. The Vadose/W model was run for 100 years. The results of the analyses with regards to the changes of the water content versus time are shown on Figure 1. Figure 1 indicates that the volumetric water content of the claystone increases up to about 32 percent near the ground surface and to decrease to the original water contents at a depth of approximately 11 meters in 100 years.
Table 1. Soil properties for claystone. Saturated hydraulic conductivity, Kv (cm/sec) Porosity, n Displacement pressure, hd (m) Pore size distribution index, 2.0 108 0.37 6 0.218
Depth Below Ground Surface (m)
10
1/2000 8/2010
15
8/2020 8/2040
20
8/2060 8/2080
25
8/2099
30
Figure 1. Water content profiles for claystone calculated by Vadose/W.
3.3
Simplified hand calculations
McWhorter & Nelson (1979) developed a method of analysis to model the movement of the wetting front. Below the wetting front the pressure head in the soil is dictated by the soil suction, which in turn, is a function of the water content. The analysis is complicated somewhat by the fact that for a small zone across the wetting front the water content, and hence, the suction is varying. McWhorter & Nelson (1979) derived the following generalized relationship between the hydraulic conductivity and volumetric water content. McWhorter & Nelson (1979) indicated that Equation (1) is quite insensitive to where is increasing or decreasing. f r ke ks n
r ( 2 3 )
(1)
where, ke effective hydraulic conductivity; ks saturated hydraulic conductivity; f final volumetric water content; r residual volumetric water constant; n porosity; and pore size distribution index. According to Darcys law, for the case where water is continuously ponded to a small depth at the ground surface, the infiltration at the surface, qi, in a homogeneous and isotropic medium is represented by Equation (2): h qi ke 1 z (2)
in which h pressure head and z elevation head, and other symbols are as previously defined.
1191
McWhorter & Nelson (1979) showed that a good approximation for the pressure head, h, at the depth of potential heave, zp, is to assume that the suction is equal to the displacement pressure, hd, of the soil. Assuming the pressure head due to the ponding at the surface is negligible, Equation (2) can be expressed as follows: h qi ke 1 d zp (3)
CALCULATIONS OF HEAVE
The infiltration, qi, does not completely contribute to downward migration of the wetting front. Some of it is retained in the pore spaces of the soil and increases the water content. This increase in water content is, of course, the factor that causes heave. Combining Equations (1) and (3) yields,
q z p 2 3 f r n r i ks z p hd
Heave prediction methodologies were first developed in the early 1960s (e.g. van der Merwe, 1964). More rigorous methods were subsequently developed based on theoretical considerations and results of one-dimensional oedometer tests (Fredlund, 1983; US Army Corps of Engineers, 1983; Nelson & Miller, 1992; Fredlund & Rahardjo, 1993). The methodology presented herein has refined those methods and establishes the definition of the Heave Index, CH. It utilizes the net mechanical stress ( ua) as the controlling stress state variable, and will be referred to as the oedometer method. The heave prediction methodologies all use the same basic equation for calculation of heave. The equation for heave of a soil layer of thickness, zi, as presented in Nelson et al. (2006) is,
(4)
CH zi log
i1

cv vo i
(7)
Equation (4) can be simplified for unsaturated clay soils as follows: k ( f ) z p 2 3 f n ks z p hd

(5)
where, free-field heave; CH heave index; zi thickness of the layer; cv constant volume swelling pressure; and vo overburden plus applied stress at the midpoint of the layer. 4.1 Heave based on saturated final water content profiles
If k(f) ks, Equation (5) can be further simplified as Equation (6). z p 2 3 f n z p hd

(6)
Calculations were performed for the homogeneous claystone profile that was discussed in Section 3.2 of this paper. Using the soil properties presented in Table 1 and a value of zp of 19.1 meters presented in Table 2, the results of these analyses indicated a final constant volumetric water content of 33% can be used for design.
Table 2. Claystone prediction. Total unit weight Mg/m3 1.86 Percent swell, S% % 4.2 parameters Swelling pressure, cs kPa 608 used Heave index, CH 0.046 for heave
The heave of the soil with the saturated final water content, as discussed in Section 3.1 of this paper, was calculated using this methodology and assuming that the entire profile was saturated. The calculations result in 384 mm of predicted free-field heave. The soils parameters used for the analyses are presented in Table 2. 4.2 Heave based on water content profiles from vadose zone modeling
Depth of potential heave, zp m 19.1
The final water contents for various times during the 100 year design as calculated by the Vadose/W program are presented in Figure 1. These final water content profiles can be used for design purposes to estimate the potential heave. The percentage of saturation from the original water content profile to the selected design water content profile must be determined. A summary of the degree of saturation for these profiles normalized to the saturated volumetric water content of 37% is presented in Table 3.
1192
Table 3. Normalized final degree of saturation from Vadose/w model. Depth (meters) 02 24 46 68 810 1012
1.0
Normalized final degree of saturation 0.94 0.94 0.92 0.90 0.85 0.24
this paper. Using the final design volumetric water content of 33 percent, and the methodology presented in Section 4.2 of this paper for calculation of heave for partially saturated soils, the calculated free-field heave was 362 mm. 5 DISCUSSION OF RESULTS
Initial Vol. W. C. = 18%

0.9 0.8
Initial Vol. W.C. = 23% Initial Vol. W.C. = 30% Initial Vol. W. C. = 38% y = -3.2994x 2 + 6.2475x - 1.98 R = 0.9928 y = -5.6649x2 + 12.402x - 5.737 R = 0.9992 y = -4.5032x2 + 14.551x - 9.0581 R = 0.9986
Normalized Percent Swell
0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.3
y = -4.9077x2 + 9.3679x - 3.495 R = 0.9945
0.4
0.5
0.6
0.7
0.8
0.9
1.0
Figure 2. Normalized percent swell vs. degree of saturation (modified from Chao, 2007).
The water content for a saturated claystone used as an example in this paper is 37 percent, while the water content from the hand calculations is 33 percent. Both of these water contents are assumed to be constant through the entire depth of potential heave. The water content in 100 years calculated using the Vadose/w model was determined to be 32 percent near the ground surface and to decrease to the original water contents at a depth of approximately 11 meters. A comparison of these three final water contents that could be used for design is shown on Figure 3. Using the final water contents presented in Figure 3, the free-field heave was calculated. The heave is summarized in Table 4 and indicates that the calculated heave for a saturated condition is reduced by 25 percent by using a partially saturated final water content profile.
10 0 15
Volumetric Water Content (%) 20 25 30
Percent swell and swell pressures can be determined for partially wetted soils using the calculated normalized percentage saturation. Data from Chaos dissertation (Chao, 2007) has been modified to present normalized degree of saturation versus normalized percent swell for initial volumetric water contents of 18, 23, 30, and 38 percent, as shown in Figure 2. These curves can be used to calculate the percent swell for a partially saturated soil using the results of a fully saturated consolidation-swell test. Additionally, the swell pressures can be determined for partially wetted soils. Reichler (1997) indicates that an e-log p curve from a partially saturated consolidation-swell test has the same shape as that from a fully wetted consolidation-swell test. Thus, using the revised percent swell as calculated in the above discussion, the consolidation-swell line can be shifted downward to the new percent swell and a new swell pressure for the partially wetted soil can be calculated. The calculated free-field heave using this methodology was 287 mm for the water content profile in 100 years obtained from the Vadose/W program. 4.3 Heave based on water content profiles from hand calculations
35
40
A
5
Depth Below Ground Surface (m)
10
15
20
25
30
A: Initial Water Content Prof ile B: Saturated Water Content Prof ile C: Water Content Prof ile f rom VADOSE/W Model D: Water Content Prof ile f rom Hand Calculation
Figure 3. Comparison of design water content profiles. Table 4. Summary of heave calculations. Scenario Saturation for full depth of potential heave Water content profile from Vadose/w Water content profile from hand calculation Calculated free-field heave (mm) 384 287 362
The heave can be calculated using the constant final water content as discussed in Section 3.3 of
1193
CONCLUSIONS
The final water content for design can be calculated assuming that the entire depth of potential heave is saturated or by estimating a design water content profile using commercially available computer codes. If project budgets do not allow for extensive modeling, simplified hand calculations can be used to determine a constant design water content. The analytical solutions will provide a residual water content profile in the upper lithographic region with decreasing water contents at depth as shown on Figure 1. The final water contents may be less than saturation, thus giving reduced values of heave versus the heave calculated using a saturated water content profile. Additionally, the water content profile decreases back to the water content that existed prior to development at depths below the limits of the downward water migration. Using this water content profile will result in the calculation of substantially less heave than that calculated assuming saturated conditions. The final water content can also be determined using hand calculations. The results of these calculations will result in a water content profile that is lower than a fully saturated profile, and thus less heave will be calculated versus that calculated using a saturated design water content profile. However, the hand calculations will not be able to predict the reduction from the final water content to the original water content at depth. Thus, heave calculations using the final water content profile may result in larger calculated heave than using the results of an analytical solution if the depth of wetting does not migrate through the depth of potential heave during the design life of the structure. For foundation design the entire depth of potential heave should be assumed to be wetted during the life of the structure, unless site specific analyses have been performed to accurately determine the rate of migration of the wetting front and the resulting time rate of heave. For sites where a perched zone will not develop, and it is not economical to analyze the migration of a wetting front, hand calculations can be performed to determine a final design water content. Heave calculations performed using a final design water content will result in significantly less heave, and result in a more economical foundation, as compared to calculations performed assuming saturation of the entire depth of potential heave. REFERENCES
Chao, K.C. 2007. Design Principles for Foundations on Expansive Soils. Dissertation submitted in partial requirement for the Ph.D. Degree, Colorado State University, Fort Collins, Colorado.
Chao, K.C., Overton, D.D. & Nelson, J.D. 2006. The Effects of Site Conditions on the Predicted Time Rate of Heave, American Society of Civil Engineers, Unsaturated Soils Conf. 2006. Special Publication 147. Fredlund, D.G. 1983. Prediction of Ground Movements in Swelling Clays. Proceedings for the 31st Annual Soil Mechanics and Foundations Engineering Conference, University of Minnesota, Minneapolis, MN. Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soil. John Wiley & Son, Inc., New York, NY. Fredlund, D.G., Xing, A. & Huang, S.Y. 1994. Predicting the Permeability Function for Unsaturated Soils Using the Soil-Water Characteristic Curve. Canadian Geotechnical Journal, 31:533546. GEO-SLOPE International, Ltd. 2007. GEO-STUDIO VADOSE/W Software Package for Seepage Analysis, Version 7.15. Calgary, Alberta, Canada. Houston, S.L., Houston, W., Zapata, C. & Lawrence, C. 2001. Geotechnical Engineering Preactice for Collapsible Soils. Geotechnical and Geological Engineering, Kluwer Academic Publications, 19:333355. McWhorter, D.B. & Nelson, J.D. 1979. Unsaturated Flow Beneath Tailings Impoundments. J. GeotechEngDiv, ASCE, November, 105(GT11):13171334. Nelson, J.D. & Miller, D.J. 1992. Expansive Soils: Problems and Practice in Foundation and Pavement Engineering. John Wiley & Sons, Inc., New York, NY. Nelson, J.D., Overton, D.D. & Durkee, D.B. 2001. Depth of Wetting and the Active Zone. Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE, Houston, Texas, 95109. Nelson, J.D., Reichler, D.K. & Cumbers, J.A. 2006. Parameters for Heave Prediction by Oedometer Tests. Proceedings of the 4th Intl. Conf. on Unsaturated Soils. Arizona. April 26. Overton, D.D., Chao, K.C. & J.D. Nelson. 2006. Time Rate of Heave Prediction for Expansive Soils. American Society of Civil Engineers, Proceedings of GeoCongress 2006 Conference. Atlanta. February. Reichler, D.K. 1997. Investigation of Variation in Swelling Pressure Values for an Expansive Soil. Thesis submitted in partial requirement for the Master Degree, Colorado State University, Fort Collins, Colorado. Schmalz, T.C. & Stiemer, S.F. 1995. Consideration of Design Life of Structures. Journal of Performance of Constructed Facilities, ASCE, 9(3):206219. US Army Corps of Engineers, 1983. Technical Manual TM 58187, Foundations in Expansive Soils. van der Merwe, D.H. 1964. The prediction of heave from the plasticity index and the percentage clay fraction of soils. Civil Eng. South Africa 6:103107. Washington State University, 2002. Biological Systems Engineering Department, ClimgenClimate Data Generator Users Manual, Pullman, Washington.
1194
Bearing capacity of a strip footing on an unsaturated sand

T. Schanz & Y. Lins
Ruhr-Universitt Bochum, Faculty of Civil and Environmental Engineering, Chair for Foundation Engineering, Soil and Rock Mechanics, Bochum, Germany
S.K. Vanapalli
Department of Civil Engineering, University of Ottawa, Ottawa, ON, Canada
ABSTRACT: The bearing capacity of a surface strip footing on Hostun sand was determined under saturated, dry and unsaturated conditions in specially designed equipment. The variation of bearing capacity with respect to matric suction was also predicted by modifying Vanapalli & Mohamed (2007) equation taking account for plane strain conditions and using the Kinematic Element Method (KEM) proposed by Gusmann (1986). There was a reasonably good comparison between the measured experimental results of bearing capacity measurements and predicted values using both the methods. 1 INTRODUCTION located above the groundwater table and the stresses due to loading of these foundations are distributed within the unsaturated soil zone. Such a scenario is more commonly encountered in semi-arid and arid regions of the world. Reliable estimation of the bearing capacity therefore will be dependent on the understanding of the mechanical behavior of unsaturated soils. In other words, the influence of soil suction should be taken into account in the assessment of the bearing capacity. However, in conventional engineering practice, foundations are designed assuming saturated conditions for the soil such that the design approach is conservative. In some scenarios the influence of groundwater table variation is taken into account by using different values of density of the soil in the estimation of bearing capacity (Terzaghi 1943). Simplistic and conservative approaches are conventionally used in engineering practice because a valid framework for interpreting the bearing capacity of unsaturated soils is not available. There is growing evidence to show that the bearing capacity of shallow foundations is significantly influenced by matric or capillary suction (Steensen-Bach et al. 1987, Oloo et al. 1997, Costa et al. 2003, Rojas et al. 2007, Vanapalli & Mohamed 2007). Mohamed & Vanapalli (2006) studies have shown that the bearing capacity of a square model footing on a coarse-grained soil under unsaturated condition is approximately 5 to 7 times higher than the bearing capacity under saturated conditions. Vanapalli & Mohamed (2007) recently provided a framework for both interpreting and predicting the bearing capacity of unsaturated soils as the
The bearing capacity of continuous or strip footings (i.e. shallow foundation) placed on a homogeneous soil is conventionally estimated using Terzaghi (1943) equation based on saturated shear strength parameters (i.e. c and ): qu c Nc qNq 0.5BN (1)
where: qu is the ultimate bearing capacity, Nc, Nq and N are the bearing capacity factors related to cohesion, c of the soil, the surcharge q, which is the overburden pressure (i.e. D) and unit weight, of the soil, respectively. The footing width is B and D is the depth. The bearing capacity factors Nc, Nq and N are dependent on the angle of internal friction, value. The angle of internal friction value used in the analysis should account for plane strain conditions for estimating the bearing capacity of continuous footings; typically, this value is taken to be equal to 1.1 times the value measured from triaxial tests (DS 415 1984). Several investigators proposed bearing capacity factors Nc, Nq and N based on different assumptions (Terzaghi 1943, Meyerhof 1951, Gussmann 1986, Kumbhojkar 1993). The estimated bearing capacity values, particularly for coarse-grained soils with angle of internal friction, values greater than 30 are sensitive as bearing capacity factors can have a significant influence. The bearing capacity and settlement behavior are two key properties required in the design of shallow foundations. Shallow foundations are typically
1195
conventional approach for design of foundations is conservative. The focus of Vanapalli & Mohamed (2007) and other studies in the literature (SteensenBach et al. 1987, Costa et al. 2003, Rojas et al. 2007) were directed towards understanding the bearing capacity of square footings. In this paper, the bearing capacity of continuous model footing was determined on Hostun sand under saturated, dry and unsaturated conditions. A box was specially designed to load model continuous (i.e. strip) footings under plane strain conditions to conduct these tests. The shear strength parameters required for analyzing the data were determined from modified biaxial tests on Hostun sand under saturated and unsaturated conditions (Schanz & Alabdullah 2007 and Alabdullah & Schanz 2009). The experimental results were analyzed using the semi-empirical equation proposed by Vanapalli & Mohamed (2007) and the Kinematic Element Method (KEM) presented by Gussmann (1986). The results of the research presented in this paper suggest that there is a reasonably good comparison between the predicted and measured bearing capacity values for the tested Hostun Sand using both these approaches. 2 MATERIAL USED
Percent passing by weight (%)
100 80 60 40 20
Clay
Silt
fine medium coarse
Sand
Gravel
0 0.001
0.01
0.1
10
100
Grain size (mm)

Figure 1. Grain size analysis of Hostun sand.
Liquid phase
Gas phase Transition zone
60 50 40 30 20 10 0
Drainage process
s
aev
Saturated zone Residual zone
r, r
0.1
10
100
The experimental program for this study was conducted on Hostun Sand, which is quartz sand. The material is poorly-graded medium sand according to the Unified Soil Classification System. The grainsize distribution curve of this soil is given in Fig. 1. The Soil-Water Characteristic Curve (SWCC) is an important tool in understanding the engineering behavior of unsaturated soils. This curve relates the suction in the soil to the gravimetric water content, volumetric water content or degree saturation and is often used along with the saturated soil parameters in the prediction of the engineering properties of unsaturated soils (e.g. hydraulic conductivity, shear strength and also bearing capacity). The SWCC of Hostun sand is plotted as the variation of volumetric water content versus suction (Fig. 2). The SWCC shows that the material desaturates in a narrow range of suction values. The various features from the measured SWCC using the approach suggested in Vanapalli et al. (1999) are summarized below: Saturated volumetric water content: s 40% Air-entry value: aev 1.9 kPa Residual volumetric water content: r 5% Residual suction: r 3.3 kPa Saturated zone: From 0 to 1.9 kPa Transition zone: 1.9 to 3.3 kPa Residual zone: 3.3 kPa
Suction (kPa)
Figure 2.
SWCC of Hostun sand.
BEARING CAPACITY TESTS
3.1 Equipment Bearing capacity of a surface continuous model footing was determined in a specially designed box (Fig. 3). The dimensions of the box are 1000 mm in length, 500 mm in height and 500 mm in width. The box consists of an outer frame made of wood and an inner tank made of plexiglass. The outer frame serves as a stiffener to prevent lateral bending of deformations that may occur while loading the model footing in the tank. The tank has 4 openings (2 on each side), that are used for allowing water below the strip footing. A nonlinear suction profile develops above the ground water level (i.e. below the footing) due to capillary stresses in the bearing capacity box. This can be achieved by hanging water column technique. The box is also equipped with tensiometer sensors at a depth of 100 mm from the soil surface in the tank to measure the matric suction values (i.e. capillary suction) below the footing. The dimensions of the model strip footing used in the study are 477 mm in length, 79 mm in width and 45 mm in height.
1196
25 20
Load (kN)
=2 kPa =3 kPa =4 kPa

dry
S=1.0
Fmax=19.1 kN Fmax=14.7 kN Fmax=14.1 kN Fmax=8.1 kN Fmax=4.9 kN
15 10 5 0 0 25 20
Settlements (mm)
10
15
20
Load (kN)
Figure 3.
Bearing capacity equipment.
15 10 5 0 0
Reihe7 S=1.0 Reihe1 =2 kPa Reihe2 =3 kPa Reihe3 =4 kPa Reihe4 dry
3.2
Experimental program and testing procedure
Bearing capacity tests include the loading of a strip footing model that is placed on the surface of the sand specimen (i.e. either water or air saturated) as well as the unsaturated sand specimen ( 2; 3 and 4 kPa). The footing was loaded at a constant rate of 0.002 mm/s. The dry specimen with a height of about 360 mm was prepared by uniformly pluviating oven dry sand with a funnel (500 ml capacity) into the bearing capacity box in several layers. After laying each layer, the sand was compacted using a 2 kg hand compactor. The water saturated and unsaturated specimens with a height of about 360 mm were prepared by uniformly pluviating oven dry sand using a funnel (500 ml capacity) into the bearing capacity box filled with deaired water. The deaired water was allowed to gradually percolate into the sand box from the bottom. The water level was always kept slightly above the sand specimen in the bearing capacity box to avoid collection of any occluded air bubbles during the specimen preparation. The height of water in both burettes (one on the right hand and one on the left hand side of the bearing capacity) were then set to 200, 300 and 400 mm below the surface of the sand specimen using the hanging water column technique. The matric suction values below the footing up to a depth of 1.5 width of the footing were measured using tensiometers. The bearing capacity of the strip footing was determined by loading it after attaining equilibrium conditions with respect to capillary stresses (i.e. matric suction) values. More details of the testing techniques are discussed in Lins et al. (2009). 3.3 Experimental results and interpretation
Suction (kPa)
Figure 4. Experimental results of bearing capacity tests (a) load versus settlement (b) load versus suction.
Experimental results from the bearing capacity tests on water and air saturated soil as well as unsaturated soil specimens are presented in Fig. 4.
The load versus settlement and the ultimate bearing capacity is plotted versus suction in Fig. 4(a). The results suggest show a linear increase in load with settlement initially (i.e. elastic range). There is a non-linear increase in the load as the settlement further increases and reaches a maximum value (i.e. ultimate bearing capacity). After failure, the settlements are further increasing without any increase in the load carrying capacity of the footing. The strip footing was loaded up to 20 mm, which is 25% of the width of the strip footing. There is reasonably well defined failure load for all the tests suggesting that the mode of failure is close to general shear failure conditions (see Lins et al. 2009 for more details). These results are consistent with other experimental results in literature on coarse-grained soils, for example, Steensen-Bach et al. (1987) and Mohamed & Vanapalli (2006). The maximum load carrying capacity of the strip footing under unsaturated conditions occurs at a suction value of 3 kPa, which is Fmax 19.1 kN (see Fig. 4(b)). This is equivalent to a stress value of qu 517 kN/m . The maximum load for the air saturated (S 0) specimen is Fmax 4.9 kN, which corresponds to a stress of qu 132 kN/m . It is of interest to note that the bearing capacity of the sand at saturated conditions is close to the bearing capacity at dry condition. The results show that the bearing capacity of strip footing of the tested Hostun sand is significantly influenced due to matric suction. The bearing
1197
capacity under unsaturated condition was found to be approximately 2.5 to 4 times higher than for the saturated condition. The average value of matric suction used in the calculations of the bearing capacity is assumed to be at the centroid of the matric suction distribution diagram from 0 to 1.5B. This procedure used is consistent with the earlier research studies by Vanapalli & Mohamed (2007) and Oh & Vanapalli (2008). 3.4 Prediction of variation of bearing capacity with respect to suction using the semiempirical approach proposed by Vanapalli & Mohamed (2007)
Vanapalli & Mohamed (2007) proposed a semiempirical procedure for predicting the bearing capacity of unsaturated soils using the saturated shear strength parameters (i.e. c and ) and the SWCC. The proposed equation is given below: qu [c (ua uw)b (1 S) tan (ua uw)AVR S tan ] Ncc 0.5BN (2)
where: (ua uw)b is equal to the air-entry value, (ua uw)AVR is the average suction value below the footing and is the fitting parameter equal to unity to coarse-grained soils such as sands with no plasticity index. This approach is similar to predicting the shear strength of unsaturated soils using a fitting parameter, value equal to 1 (Vanapalli & Fredlund, 2000). The parameters c and are the shape factors. The shape factor value of 0.5 was used for predicting the bearing capacity of strip footing. This equation takes the same form as Terzaghis equation shown in Eq. (1) without the surcharge contribution if the matric suction, (ua uw) value is set to zero. The contribution of surcharge is not necessary in the present study as the model footing was loaded placing them directly on the sand surface. Predicted bearing capacity and experimental bearing capacity results are shown in Fig. 5. The
1000 Prediction (Vanapalli & Mohamed 2007) Experimental Results
saturated shear strength parameters determined from biaxial equipment taking account of plane strain conditions for Hostun sand were respectively, c 0 and 46.9) (see Schanz & Alabdullah 2007 and Alabdullah & Schanz 2009 for more details). The results show reasonably good trends between the measured and predicted results. Similar to experimental results, the predicted bearing capacity is first increasing with increasing values of suction. The capillary forces between the grains contribute towards increase in the bearing capacity. The contribution of matric suction towards bearing capacity starts decreasing with further increase in matric suction. The characteristic of the bearing capacity behavior is similar to that of the shear strength of unsaturated soils (Vanapalli 2009). As the soil reaches close to dry conditions, it loses the influence of capillary forces and hence the bearing capacity starts reducing in the residual zone. In summary, there is a good comparison between the measured and predicted bearing capacity values in the saturated (i.e. boundary effect) zone, the transition zone and the beginning of the residual zone (Fig. 5). However, in the residual zone for matric suction values greater than 4 kPa, the predicted bearing capacity is overestimated by the model. 3.5 Prediction of bearing capacity using Kinematic Element Method (KEM) proposed by Gussmann (1986)
The Kinematic Element Method (KEM) proposed by Gussmann (1986) was used to predict the bearing capacity of Hostun sand. Following steps are performed: 1. Design of relevant failure mechanism below footing. 2. Solving the kinematics and statics of the system. 3. Optimization procedure. The shear strength contribution due to matric suction is included in the cohesion term of the soil to take into account of the effect of matric suction on the bearing capacity of unsaturated soil in the KEM approach. In other words, apparent cohesion is used instead of the effective cohesion in this approach. The failure mechanism for bearing capacity was assumed to constitute of bounded rigid elements of straight lines as shown in Fig. 6. Elements of 3, 6 and 12 were used extending optimization procedure below the footing (b/2) to visualize different failure scenarios. The bearing capacity calculations were carried out with the Gussmann (1986) approach using 12 elements as the failure pattern looks similar to expected trends below a strip footing in comparison to 3 and 6 elements.
Bearing capacity (kPa)
800 600 400 200 0 0
Suction (kPa)
Figure 5. Comparison of predicted and experimental results of bearing capacity using Vanapalli & Mohamed (2007).
1198
A comparison is made to check whether the bearing capacity factors generated by the KEM approach are consistent with the results of the bearing capacity factor of N using other methods (i.e. Kumbhojkar 1993 and DIN 4017) for dry soil with different angle of internal friction values. This is the key factor influencing the bearing capacity of dry soil, for which the apparent cohesion value is equal to zero. The footing is directly loaded on surface and hence the influence of overburden pressure is zero. Therefore, there is no influence of the surcharge bearing capacity factor Nq. The bearing capacity factor value, N determined from the KEM approach is shown in Fig. 7 along with other two methods. The results of the bearing capacity factor using three different methods are approximately the same.
For the prediction of bearing capacity of Hostun sand the following parameters are used: Dry unit weight: d 19 kN/m Wet unit weight: b 20 kN/m Angle of internal friction: 46.9 Apparent cohesion: c 2, 3, 4 kN/m (for suction values of 2, 3, 4 kPa) The apparent cohesion, c value was calculated using Eq. 3. c c (ua uw) (sk) tan (3)
The value of the fitting parameter, 1 for non-plastic soils such as Hostun sand used in the present study (Vanapalli & Fredlund, 2000). The results demonstrate that the prediction of bearing capacity for Hostun sand (for both, S 0 and S 1) specimens is in good agreement with the measured values. It also is observed that the bearing capacity of the unsaturated specimen is increasing with increase in matric suction value. However, a smooth nonlinear increase followed with a decrease of bearing capacity is not observed with an increase in the matric suction value as seen from experimental results. 4 SUMMARY AND CONCLUSIONS
Figure 6.
Failure mechanism12 elements.
Bearing Capacity Factor N b
200 160 120 80 40 0 20
DIN 4017 KEM Kumbhojkar (1993)
25
Friction angle ()
30
35
40
45
Figure 7. Variation of bearing capacity factor with angle of internal friction value for different methods.
An experimental program is undertaken to study the bearing capacity behavior of Hostun sand in both saturated, dry and unsaturated conditions using a continuous footing (i.e. strip footing). The bearing capacity studies were undertaken in a specially designed box. The results of the study show that the bearing capacity behavior is significantly influenced by matric suction. The ultimate bearing capacity for unsaturated specimen was found to be 2 to 2.5 times higher in comparison to water saturated and dry specimens. There is a nonlinear increase in bearing capacity with an increase in suction. In the residual zone, the bearing capacity is decreasing with an increase in suction values. The measured results of the bearing capacity were compared with the predicted results using the approache presented by Vanapalli & Mohamed (2007) and KEM developed by Gussmann (1986). The experimental results and the predicted results from both methods are summarized in Fig. 9. The results show acceptable trends between the measured and predicted results. More experimental studies are in progress to investigate the influence of suction on the bearing capacity. These studies will be useful to provide additional credence to the results presented in this study.
1199
1000
X Bearing capacity (kPa)
800 600 400 200 0
Experimental Results Prediction (Gussmann 1986)
Suction (kPa)
Figure 8. Comparison of predicted and experimental results of bearing capacity using KEM (Gussman 1986).
1000
Bearing capacity (kPa)
800 600 400 200 0 0
Prediction (Vanapalli & Mohamed 2007) Experimental Results Prediction (Gussmann 1986)
Suction (kPa)
Figure 9. Comparison of bearing capacity results.
The empirical equation proposed by Vanapalli & Mohamed (2007) needs improvements for predicting bearing capacity values within the residual zone. In the KEM software developed by Gussmann (1986), the negative pore-water pressure that is equal to the capillary pressure or matric suction in a soil has to be included. ACKNOWLEDGMENT The authors would like to thank Prof. Peter Gussmann for providing the KEM software and his valuable help using the KEM software. REFERENCES
Alabdullah, J. & Schanz, T. 2009, Shear strength of unsaturated sand under plane strain conditions, 4th Asia-Pacific Conference on Unsaturated Soils, Newcastle, Australia, pp. 129133. Costa, Y.D., Cintra, J.C. & Zornberg J.C. 2003. Influence of matric suction on the results of plate load tests performed on a lateritic soil deposit. Geotechnical Testing journal, 26(2): 219226. DIN 4017. 2006. BaugrundBerechnung des Grundbruchwiderstandes von Flachgrndungen.
DS 415. 1984. The Danish code of practice for foundation engineering. Danish Society of Civil Engineering. Gussmann, P. 1986. Die Methode der kinematischen Elemente. Baugrundinstitut Stuttgart. Mitteilung Heft 25. Kumbhokjar, A.S. 1993. Numerical evaluation of Terzaghis N. Journal of Geotechnical Engineering, ASCE, 119 (3): 598 607. Lins, Y., Schanz, T. & Vanapalli, S.K. 2009. Bearing capacity and settlement behavior of a strip footing on an unsaturated coarse-grained soil. 4th Asia-Pacific Conference on Unsaturated Soils, Newcastle, Australia, pp. 443448. Meyerhof, G.G. 1951. The ultimate bearing capacity of foundations. Geotechnique 2: 301332. Mohamed, F.M.O. & Vanapalli, S.K. 2006. Laboratory investigations for the measurement of the bearing capacity of an unsaturated coarse-grained soil. In 59th Canadian Geotechnical Conference. BC, Vancouver (http://www.x-cd.com/SeatoSkyOnline/S1/0219-226. pdf) Oh, W.T. & Vanapalli, S.K. 2008. Modelling the stress versus settlement behavior of model footings in saturated and unsaturated sandy soils, In Proceedings of the 12th International Conference of International Association for Computer Methods and Advances in Geomechanics, 16 October, Goa, India, pp: 21262137. Oloo, S.Y., Fredlund, D.G. & Gan, J.K-M. 1997. Bearing capacity of unpaved roads. Canadian Geotechnical Journal, 34(3): 398407. Rojas, J.C., Salinas, L.M. & Seja, C. 2007. Plate-load tests on an unsaturated lean clay. Experimental Unsaturated Soil mechanics: In Proceedings of the 2nd Int. Conf. on Unsaturated Soils, Springer Proceedings in Physics, Weimar (Germany), 445452. Schanz, T. & Alabdullah, J. 2007. Testing unsaturated soil for plane strain conditions: a new double wall biaxial device, In T. Schanz (ed.) Experimental Unsaturated Soil Mechanics, Springer-Verlag. Berlin Heidelberg, pp. 169178. Steensen-Bach, J.O., Foged, N. & Steenfelt. J.S. 1987. Capillary Induced StressesFact or Fiction? In Proceedings of the 9th European Conf. on Soil Mechanics and Foundation Engineering, Budapest (Hungary), 8389. Terzaghi, K. 1943. Theoretical soil mechanics. Wiley. New York. Vanapalli, S.K., Fredlund, D.G. & Pufahl. 1999. The influence of soil structure and stress history on the soil-water characteristics of a compacted till. Gotechnique, 49(2): 143159. Vanapalli, S.K. & Fredlund, D.G. 2000. Comparison of empirical procedures to predict the shear strength of unsaturated soils using the soil-water characteristic curve. Geo-Denver 2000, ASCE Special Publication, 99: 195209. Vanapalli, S.K. & Mohamed, F.M.O. 2007. Bearing capacity of model footings in unsaturated soil. In T. Schanz (ed.) Experimental Unsaturated Soil Mechanics, Springer-Verlag. Berlin Heidelberg, 483493. Vanapalli, S.K. 2009. Shear strength of unsaturated soils and its applications in geotechnical engineering practice. 4th Asia-Pacific Conference on Unsaturated Soils, 2325 November 2009, Newcastle, Australia, pp. 579598.
1200
A simple technique for estimating the 1-D heave in expansive soils

S.K. Vanapalli, L. Lu & W.T. Oh
Civil engineering Department, University of Ottawa, Ottawa, Ontario, Canada
ABSTRACT: Several approaches are available in the literature for calculating the 1-D heave in expansive soils with lightly loaded structures. However, there are limitations because of one or more of the following reasons: i) these approaches are not universally valid as they are proposed using only limited soils data collected locally; ii) they do not use the stress state variables approach that provides a rational basis for interpretation; iii) the various parameters required in these approaches can only be obtained from time consuming laboratory or in-situ tests that are expensive and difficult to be performed. In the present study, a simple technique is proposed to estimate the 1-D heave in expansive soils which requires only the information of plasticity index, Ip and variation in water content with respect to depth in the active zone. The proposed technique was developed using the field measurements data of 1-D heave published in the literature and tested on a case study results summarized by Yoshida et al. (1983) to check its validity. There was good agreement between the measured and the predicted 1- D heave for the case study using the proposed technique. 1 INTRODUCTION Empirical methods use soil classification parameters such as Atterberg limits (Snethen et al. 1977), plasticity index and percent clay (Nayak & Christensen 1971), and percentage swell (Yoder & Witczak 1975). These methods are developed based on limited data collected locally. The swelling pressure is used as a tool to estimate the 1-D heave in expansive soils (Fredlund & Rahardjo 1993). Depending on the loading procedures, several methods are developed such as the free swell tests; the overburden swell tests; and the Constant Volume Swell (CVS) tests using conventional oedometer test methods. The swelling pressure and the 1-D heave in expansive soils can be more reliably measured or calculated using soil suction methods as they are based on the information of the stress state (i.e. suction). In these methods, the influence of suction is taken into account through the use of different parameters such as suction modulus ratio, Cw (Hamberg 1985), suction index, C (Johnson & Snethen 1978), instability index, Ipt (Mitchell & Avalle 1984), swelling index, Cs and corrected pressure, Ps (Fredlund 1983, Fredlund & Rahardjo 1993), or suction compression index, Ch (McKeen 1992). Fredlund (1983) proposed an equation that can be used to calculate 1-D heave in expansive soils using the Constant Volume Swell (CVS) oedometer test results. The two key parameters required in this technique are swelling index, Cs and initial
Expansive soil deposits are extensively found in many arid and semi-arid regions of the world. These soils are conventionally considered as problematic soils since they exhibit high potential of swell and shrinkage characteristics. Lightly loaded structures such as buildings, roads, airports, and pipe lines constructed in or on the expansive soils can undergo severe damages. Therefore, it will be useful to determine the potential heave in expansive soils such that appropriate design and construction measures can be undertaken to minimize the anticipated damages to the structures. In the present study, a simple technique is proposed to estimate the 1-D heave in expansive soils based on the approaches suggested by Fredlund (1983) and Hamberg & Nelson (1984). The proposed method is tested on a case study results summarized by Yoshida et al. (1983) to check its validity. 2 BACKGROUND
There are several procedures used in geotechnical engineering practice to estimate the swelling pressure and the 1-D heave in expansive soils. These techniques can be divided into three main categories based on: (i) empirical methods; (ii) oedometer test methods; and (iii) soil suction methods.
1201
stress state, P0. A realistic estimation of initial stress state, P0 can be obtained based only on the measurement of several indices taking account of the changes in void ratio due to both the net normal stress, ( ua) and matric suction, (ua uw). For the purpose of simplification, the initial stress state, P0 may be assumed to be equal to the corrected swelling pressure, Ps. However, the determination of the corrected swelling pressure, Ps from laboratory test is still elaborate and time consuming. Hamberg & Nelson (1984) presented a different approach that can be used to predict the 1-D heave in expansive soils using the relationship between water content and volume change in the range of shrinkage limit to liquid limit. This method is relatively simpler compared to Fredlund (1983) method; however, the influence of the in-situ stress state is not taken into account in this approach. In the present study, a simple technique is proposed to estimate the 1-D heave in expansive soils for lightly loaded structures. The proposed technique is developed by combining both the Fredlund (1983) and Hamberg & Nelson (1984) approaches to derive a new relationship. Some of the limitations associated with both the approaches are alleviated in the proposed simple technique. 3 1-D HEAVE DETERMINATION METHODS
Uncorrected swelling pressure, Ps e0 Void ratio, e Cs Final stress state, Pf Parallel
Corrected swelling pressure, P's
Adjusted rebound curve
Log ( - u a )
Figure 1. Construction procedure to determine the corrected swelling pressure incorporating for the effect of sampling disturbance (modified from Fredlund 1987).
The 1-D heave in expansive soils can be more reliably determined using Eq. (1) which incorporates the corrected swelling pressure, Ps. H Hi Cs
i 1 i 1 n n
Hi Pf log 1 e0 Ps
(1)
3.1 Fredlund (1983) method Fredlund (1983) suggested that matric suction in an expansive soil specimen can be eliminated by immersing it in water. As a result, total stress changes in the specimen mainly reflect as the swelling behavior of the expansive soil. The CVS oedometer method can be used as a tool to simulate this behavior in a laboratory environment (ASTM D4546-2000). In this test, the specimen in the oedometer is inundated after the application of a token overburden pressure. The overburden pressure is increased restricting the swelling as the specimen tends to swell. The pressure at which no more swelling occurs is called the uncorrected swelling pressure. The specimen is then loaded and unloaded following the conventional consolidation testing procedures to estimate the corrected swelling pressure as shown in Figure 1. The procedure suggested by Fredlund (1987) to estimate the corrected swelling pressure involves the determination of compressibility of the apparatus from the actual deformation measurements and also takes account of sampling disturbance extending the modified graphical method (Casagrande 1936). This procedure is based on the assumption that the in-situ soil suction is transferred onto the effective stress plane.
where Hi thickness of the ith layer, Pf ( y y uwf) final stress state, Ps corrected swelling pressure, Cs swelling index, y total overburden pressure, y change in total stress, uwf final pore-water pressure; and e0 initial void ratio. More details with respect to the determination of the corrected swelling pressure Ps and the 1-D heave calculations are available in Yoshida et al. (1983) and Fredlund & Rahardjo (1993). 3.2 Hamberg and Nelson (1984) method
Hamberg & Nelson (1984) proposed a method to determine the 1-D heave in expansive soils using suction modulus ratio, Cw as given below. H Hi Cw
i 1 i 1 n n
Hi w 1 e0
(2)
where Hi thickness of the ith layer, w change in water content; and e0 initial void ratio. The suction modulus ratio, Cw in Eq. (2) represents the variation of void ratio (i.e. volume in 1-D heave) of soil specimens with respect to water content (Eq. (3); Figure 2). Cw
e
eo
w
(3)
1202
Void ratio
Shrinkage curve S r <100%
ef
The positive (i.e. first part) and negative (i.e. second part) of the equation (i.e. Eq. (4)) represent compression and heave due to overburden and swelling pressure, respectively. In other words, the amount of heave calculated using the second part of Eq. (4) is approximately the same as that of Eq. (2) and can be written as Eq. (5). Cs Hi Hi log( Ps) Cw w 1 e0 1 e0 (5)
e
eo
Cw
Saturation line
The above equation can further be simplified as Cw (6) w Cs A correction parameter, K can be introduced to rewrite the relationship as Eq. (6). log Ps
wf
Shrinkage limit
wo
Water content
log( K Ps)
Cw w Cs
(7)
Figure 2. Procedure for determination of suction modulus ratio, Cw from water content versus void ratio relationship (modified after Hamburg, 1985).
An expression in terms of swelling pressure can be derived from Eq. (7). Ps 10
Cw C w
s
where eo and ef initial and final void ratio of soils, respectively; and w change in water content. The parameter, Cw can be obtained using the Clod test which is the modified form of COLE (coefficient of linear extensibility; Brasher et al. 1966) test. The COLE test is originally developed to determine the heave beneath airfield pavements (McKeen 1981, McKeen & Hamberg 1981). For silty clay soils, the void ratio versus water content relationship shows linear behavior for the water contents greater than the shrinkage limit (Hamberg 1985; Figure 2). The limitations of Eq. (2) are i) the equation does not take into account the effect of applied load and ii) there are difficulties in estimating the Cw values for the water contents close to shrinkage limit due to the nonlinearity in that range. However, experience from several case studies results show that the typical variation of water content in field conditions is in the range which is above the shrinkage limit to the plastic limit or slightly higher. 3.3 The proposed technique
(8)
Substituting Eq. (8) into Eq. (4) yields Hi Hi 10 s H Cs log Pf Cs log K 1 e0 (1 e0 ) Hi or, H Cs log 1 e0 KPf 10
Cw C w s Cw C w
(9)
The derived Eq. (9) can be used for estimating the 1-D heave in natural expansive soils. The required information includes the swelling index, Cs, the suction modulus ratio, Cw and the correction parameter, K. In the following section, empirical relationships that can be used to obtain the three parameters are provided. 4 SIMPLE RELATIONSHIPS FOR ESTIMATING THE PARAMETERS CW, CS AND K
The equation proposed by Fredlund (1983) (i.e. Eq. (1) to determine the 1-D heave in expansive soils can be rewritten as below. H Cs Pf Hi log 1 e0 s P Hi Hi Cs log Pf Cs log P s 1 1 e e 0 0
(4)
Relationships between Cw versus IP and Cs versus IP are developed using the published data on natural expansive soils from various regions of the world. The Cw versus IP relationship in Figure 3 shows that Cw value can be approximated as a constant value of 0.024 for the IP values greater than 30 (Eq. (10)). Expansive soils typically have Ip values greater than 30. Cw 0.024 (for IP 30) (10)
1203
0.04
0.20
Suction modulus, C w
Swelling index, Cs
0.03
0.15
Cs = 0.0193e R = 0.965
2
0.0343(IP)
0.02 Cw = 0.024 for IP > 30 0.01
0.10
Ching & Fredlund 1984 Nelson & Miller 1992 Vu & Fredlund 2004 Clifton et al. 1984
0.05
0.00
20
30
40
60
80
0.00 20
30
40
50
60
70
80
Plasticity index, Ip (%)
Plasticity index, Ip (%)
Figure 3. Values of plasticity index, Ip and suction modulus ratio, Cw from several laboratory tests results (Data for generating this relationship is collected from several publications. These publications are summarized in the References section).
Figure 4. Relationship between the plasticity index, Ip and corrected swelling index, Cs from a several laboratory tests results.
The relationship between the swelling index, Cs and the plasticity index, IP using the published results from the literature is shown in Figure 4. The Cs values were determined as per ASTM D45462000 using undisturbed samples collected from the zone of depth of zero to 2.5 m of the sites, which is typically the active zone in several regions of the world. In addition, the equipment compressibility and sampling disturbance correction was applied following the procedure suggested by Fredlund (1987) for determining the Cs values. The relationship shows that Cs value increases exponentially with increasing IP (Eq. (11)). Cs 0.0193e 0.0343( I P ) (11)
1000
Correction parameter, K
100 10 1 0.1 0.01 0.001 0.0001
K=0.0033e Fredlund 1969 0.64( w) K=0.0039e Hamberg & Nelson 1984 0.64( w) K=0.0035e Osman & Sharief 1987 K=0.0035e0.64(w) Osman & Sharief 1987 0.64( w) K=0.0039e Snethen & Huang 1992
0.64( w)
KI = [-0.0018 ln (IP)+0.01]e KII = 0.0039e

0.64(w) 2
0.64(w)
, R = 0.997
Water content change, w
10
15
The relationship between the correction parameter, K and the water content change, w using the data from five case study results is shown in Figure 5. These case studies are from different regions of the world that include Canada, Sudan and USA (i.e. Fredlund 1969, Hamberg & Nelson 1984, Osman & Sharief 1987, Snethen & Huang 1992). Four of the above five case studies do not have the information of the swelling index, Cs (excluding Fredlund 1969). Therefore, the Cs values were estimated using Eq. (11) based on their IP values. The best-fitting equations shown in Figure 5 indicate that the correction parameter, K versus w relationships can be expressed as Eq. (12). K e ( w ) (12)
Figure 5. Relationship between the correction parameter, K and water content change, w.
of 0.64 for various natural expansive soils with plasticity index, IP values ranging 25 to 45. Therefore, Eq. (12) can be rewritten as below. K I e 0.64 ( w ) (13)
Figure 6 shows the variation of the factor, with the plasticity index, IP of the soils. The factor, increases nonlinearly with increasing IP; the factor, can be expressed as a function of IP as shown in Eq. (14).
0.0018ln( I P ) 0.01
(14)
In addition, the best-fitting equation for all the five case studies (see Figure 5) show that the factor, is independent of the type of soil. In other words, value may be assumed to be constant with a value
Equation (12) can be rewritten as Eq. (15) (hereafter referred to as KI). K I 0.0018ln( I P ) 0.01 e 0.64( w ) (15)
1204
e-4
Fredlund 1969 Hamberg & Nelson 1984 Osman & Sharief 1987 Osman & Sharief 1987 Snethen & Huang 1992
e-5
= -0.0018ln(IP)+0.01
R = 0.967
2
and 0.084 using Eqs (10) (Figure 3) and (Figure 4), respectively. The distribution of initial and final water content with depth along with the saturation water content (i.e. 100% degree of saturation) for the site is shown in Figure 7. The information of water content change, w is derived from this figure to calculate the correction parameter, K (i.e. KI (Eq. (13)) and KII (Eq. (16)). 5.2 Comparison between the measured and the estimated heaves Table 1 and Figure 8 show the comparison between the measured and the estimated 1-D heaves using
Water Content (%)
10 0.0 0.5 1.0 1.5 2.0 2.5 3.0
Initial water content Final water content Water content for S = 100%
e-6 25
30
35
40
45
Plasticity index, IP (%)
Figure 6. Relationship between and plasticity index, IP.
The representative best-fitting equation for all the five case studies data is shown in Eq. (16) (hereafter referred to as KII).
Depth (m)
15
20
25
30
35
40
45
K II 0.0039e 0.64( w )
(16)
In the present study, both KI and KII were used to estimate the 1-D heave of a case study. 5 ANALYSIS OF THE PROPOSED TECHNQUE USING A CASE STUDY RESULTS
The validity of the proposed technique is tested using a case study results published by Yoshida et al. (1983). The average liquid limit, L and plastic limit, P of the expansive soil (i.e. Regina clay) are 77% and 33%, respectively with an average natural water content of 29%. Comparisons are provided between the measured heave and the estimated heave using the proposed method (using KI and KII) and the methods proposed by Fredlund (1983) and Hamberg & Nelson (1984). 5.1 Description of the site The site used in the present study is located in Saskatchewan, Canada (Yoshida et al. 1983), which has lightly loaded industrial building constructed on it. The undisturbed specimens used for the laboratory tests for measuring the corrected Cs values were obtained from the active zone depth (i.e. 0 to 2.3 m). The average plasticity index, IP of the specimen from the clay in the active zone of 0 to 2.3 m depth is 43%. Therefore, the value of is estimated as 0.0032 using Eq. (14). The average value of the initial void ratio, e0 for each layer is 0.962 (Yoshida et al. 1983). The suction modulus ratio, Cw and the swelling index, Cs can be estimated to be 0.024
Figure 7. Variation of in situ water content with respect to the depth. (Yoshida et al. 1983). Table 1. Summary of the case study data and the comparison between the measured and the estimated heave using different methods. Hamberg & Fredlund Nelson Proposed Proposed (KII) (1983) (1984) (KI) (mm) 26 22 19 16 14 11 8 2 118 (mm) 39 46 40 33 21 20 16 11 226 2.13 (mm) 11 19 16 15 14 12 12 11 110 1.04 (mm) 11 18 15 14 12 11 10 10 101 0.95
Depth (m) 0.15 0.45 0.75 1.05 1.35 1.65 1.95 2.25 Total
EH/MH* 1.11
*
Estimated Heave (EH)/Measured Heave (MH) Measured Heave (MH) 106 mm.
1205
Heave (mm)
0.0 0.5 0 50 100 150 200 250
1.0 1.5 2.0 2.5

Estimated (Fredlund 1983) Estimated (Hamberg & Nelson 1984) Estimated (Proposed - KI ) Estimated (Proposed - KII ) Measured (Yoshida et al. 1983)
the heave. In the present study, a simple technique is proposed to estimate the 1-D heave in natural expansive soils by deriving a new relationship from Fredlund (1983) and Hamberg & Nelson (1984) methods alleviating some of the limitations of both these methods. The proposed technique was tested on case study results of Regina clay (Yoshida et al. 1983). The results of the study presented in this paper show that the proposed method can be used reliably in estimating the 1-D heave. The proposed technique is simple and needs only the information of plasticity index, Ip and the variation of natural water content with respect to depth in the active zone of natural expansive soils. REFERENCES
ASTM. 2000. One-dimensional swell or settlement properties of cohesive soils. ASTM D 4546, ASTM Book of Standards, 4.08: 363369. Brasher, B.R., Franzmeier, D.P., Valassis, V. & Davidson, S. E. 1966. Use of saran resin to coat natural soil clods for bulk density and water retention measurements. Soil Science: 101108. Casagrande, A. 1936. The determination of the preconsolidation load and its practical significance. Proc. 1st Int. Conf., Soil Mech. Found. Eng. 3: 6064. Ching, R.K.H. & Fredlund, D.G. 1984. A small Saskatchewan town copes with swelling clay problems. Proc. 5th Int. Conf. on Expansive Soils, 306310. Adelaide, Australia. Clifton, A.W., Yoshida, R.T. & Fredlund, D.G. 1984. Performance of dark hall, Regina, Canada, constructed on a highly swelling clay. Proc. 5th Int. Conf. on Expansive Soils, 197201. Adelaide, Australia. Croney, D. & Coleman, J.D. 1954. Soil structure in relation to soil suction (pF). Journal of Soil Science, 5(1): 163177. Fleureau, J.M., Verbrugge, J.C., Huergo, P.J. & Gomes Correia, A. 2002. Aspects of the behaviour of compacted clayey soils on drying and wetting paths. Canadian Geotechnical Journal 39: 13411357. Fredlund, D.G. 1969. Consolidometer test procedural factors affecting swell properties. Proc. 2nd Int. Conf. on Expansive Soils, 435456. Texas. Fredlund, D.G. 1983. Prediction of ground movements in swelling clays. Proc. 31st Annu. Soil Mech. Found. Eng. Conference., Univ. of Minnesota, MN. Fredlund, D.G. 1987. The prediction and performance of structures on expansive soils. Proc. Int. Symp. on Prediction and Performance in Geotechnical Engineering, 5160. Fredlund, D.G. & Rahardjo, H. 1993. Soil mechanics for unsaturated soils. New York: John Wiley & Sons Inc. Haines, W.B. 1923. The volume changes associated with variations of water content in soil. Jour. Agr. Sci. 13: 296310. Hamberg, D.J. 1985. A simplified method for predicting heave in expansive soils. M.S. thesis, Colorado State University, Fort Collins, CO. Hamberg, D.J. & Nelson, J.D. 1984. Prediction of floor slab heave. Proc. 5th Int. Conf. on Expansive Soils, 137140. Adelaide, South Australia.
Figure 8. Measured and estimated heave using different methods.
three different methods (i.e. Fredlund 1983, Hamberg & Nelson 1984 and the proposed method using KI and KII). The ratios of the estimated total heaves to the measured total heaves were estimated as 1.11, 2.13, 1.04 and 0.95 using Fredlund (1983), Hamberg & Nelson (1984) and the proposed method (using KI and KII), respectively. The results of the analysis show that Hamberg & Nelson (1984) method overestimates the 1-D heave of the case study. This may be attributed to ignoring the effect of overburden pressure as explained in Section 3. On the other hand, the calculated heaves using Fredlund (1983) method and the proposed technique in the present study show good agreement with the measured total heaves. The estimated heave is close to the measured heave using the parameter, KI in comparison to KII. In other words, the total heave can be more reliably estimated when the IP values of the expansive soil at the site is taken into account. However, heave comparisons were reasonable using the parameter, KII (i.e. 0.0039). The technique proposed in this paper is simple and needs only the information of plasticity index, IP and the variation of water content with respect to depth in the active zone, whereas the parameters required for Fredlund (1983) method are elaborate and hence time consuming and expensive. 6 SUMMARY AND CONCLUSION
The 1-D heave in natural expansive soils with lightly loaded structures is commonly determined using the methods proposed by Fredlund (1983) or Hamberg & Nelson (1984). However, determination of the parameters required for using Fredlund (1983) method is expensive and time-consuming. The Hamberg & Nelson (1984) method is relatively easier but it does not take into account of the effect of overburden pressure and hence may over estimate
Depth (m)
1206
Johnson, L.D. & Snethen, D.R. 1978. Prediction of potential heave of swelling soils. Geotechnical Testing Journal, 1(3): 117124. Komornik, A. & David, D. 1969. Prediction of swelling pressure of clays. Journal of the Soil Mechanics and Foundations Division, ASTM, 95: 209225. Likos, W.J., Lu, N. & Sharkey, K.J. 2005. Laboratory characterization of steeply dipping expansive bedrock in the Rocky Mountain Front Range. Journal of Geotechnical and Geoenvironmental Engineering. 131(9): 11621171. McKeen, R.G. 1981. Design of airport pavements for expansive soils. Federal Aviation Agency, US Department of Transportation, Washington, D.C. McKeen, R.G. 1992. A model for predicting expansive soil behavior. Proc. 7th intern. Conf. on Expansive Soils: 16. Texas. McKeen, R.G. & Hamberg, D.J. 1981. Characterization of expansive soils. Trans. Res. Rec. 790, Trans. Res. Board, 7378. Miller, D.J., Durkee, E.B., Chao, K.C. & Nelson, J.D. 1995. Simplified heave prediction for expansive soils. Proc. 1st Int. Conf. on Unsat. Soils, 891897. Balkema, Rotterdam. Mitchell, P.W. & Avalle, D.L. 1984. A technique to predict expansive soils movement. Proc. 5th Int. Conf. on Expansive Soils, 124130. Adelaide, Australia. Nayak, N.V. & Christensen, R.W. 1971. Swell characteristics of compacted expansive soils. Clay and Clay Minerals, 19(4): 251261. Nelson, J.D. & Miller, D.J. 1992. Expansive soils, problems and practice in foundation and pavement engineering. New York: John Wiley & Sons Inc.
Osman, M.A. & Sharief, A.M.E. 1987. Field and laboratory observations of expansive soil heave. Proc. 6th Int. Conf. on Expansive Soils, 105110. New Delhi, India. Perko, H.A., Thompson, R.W. & Nelson, J.D. 2000. Suction compression index based on CLOD test results. Proc. Sessions of Geo-Denver 2000, Geotechnical Spec. Publ. ASCE, 99: 393408. Snethen, D.R. 1980. Characterization of expansive soils using soil suction data. Proc. 4th Int. Conf. on Expansive Soils, 5475. Denver, Colorado. Snethen, D.R. & Huang, G. 1992. Evaluation of soil suction-heave prediction methods. Proc. 7th Int. Conf. on Expansive Soils: 1217. Texas. Snethen, D.R., Johnson, L.D. & Patrick, D.M. 1977. An evaluation of expedient methodology for identification of potentially expansive soils. Soil and Pavements Laboratory, US Army Eng. Waterway Exp. Sta., Vicksburg, MS. Tripathy, S., Subba Rao, K.S. & Fredlund, D.G. 2002. Water contentvoid ratio swellshrink paths of compacted expansive soils. Canadian Geotechnical Journal 39: 938959. Vu, H.Q. & Fredlund, D.G. 2004. The prediction of one-, two-, and three-dimensional heave in expansive soils. Canadian Geotechnical Journal 41(4): 713737. Yoder, E.J. & Witczak, M.W. 1975. Principles of Pavement Design. 2nd Edition. New York: John Wiley & Sons Inc. Yoshida, R.T., Fredlund, D.G. & Hamilton, J.J. 1983. The prediction of total heave of a slab-on-grade floor on Regina clay. Canadian Geotechnical Journal, 20(1): 6981.
1207
A technique for estimating the shaft resistance of test piles in unsaturated soils
S.K. Vanapalli
Civil Engineering Department, University of Ottawa, Ottawa, Ontario, Canada
K.D. Eigenbrod Z.N. Taylan
Civil Engineering Department, Lakehead University, Ottawa, Ontario, Canada Civil Engineering Department, stanbul Technical University, Istanbul, Turkey
C. Catana, W.T. Oh & E. Garven
Civil Engineering Department, University of Ottawa, Ottawa, Ontario, Canada
ABSTRACT: A testing program was undertaken in a laboratory environment to evaluate the shaft resistance of jacked mild steel open end pipe test piles embedded in two different sandy soils (i.e., Soil #1: silty sand and Soil #2: clean concrete sand) under saturated and unsaturated conditions. Based on the results of the study, a technique for estimating the shaft resistance of test piles under unsaturated conditions is presented extending an effective stress analysis approach of the conventional -method. The contribution of matric suction to the total shaft resistance was found to be in the range of 35%40% of the total shaft capacity in silty sand but almost negligible in clean concrete sand. 1 INTRODUCTION are reported in the literature to estimate the shaft resistance of piles in sands due to the contribution of capillary stresses or matric suction. In the study presented in this paper, a laboratory testing program was undertaken to evaluate the shaft resistance of test piles in two different sands (Soil #1, a silty sand & Soil #2, a clean concrete sand) under unsaturated (moist) and saturated (submerged) condition. The objective of this study is to determine the contribution of matric suction towards the total shaft resistance of piles and propose a technique to estimate the shaft resistance in unsaturated soil conditions extending an effective stress analysis approach of the conventional -method. 2 BACKGROUND
Limited number of studies are reported in the literature that consider the influence of matric suction or capillary stresses on the load carrying capacity of shallow and deep foundations (Douthitt et al. 1998, Costa et al. 2003, Georgiadis et al. 2003). More recently, some studies have been undertaken to understand the influence of matric suction on the bearing capacity of shallow foundations in both coarse- and fine-grained soils (Vanapalli & Mohamed 2007, Oh & Vanapalli 2009). The bearing capacity of shallow foundations in unsaturated sands is typically two to four times higher than the saturated bearing capacity. Even low matric suction values in the range of 2 to 6 kPa significantly contribute to the bearing capacity of unsaturated sands (Mohamed & Vanapalli 2006). The design of pile foundations are conventionally based on the principles of saturated soil mechanics or empirical procedures or based on in-situ test results. Typically, pile foundations are designed assuming saturated, dry or submerged conditions. In many cases, pile foundations may be placed under unsaturated conditions. However, to the best of the knowledge of the authors, no studies
This section provides background information of the conventional procedure followed in estimating the pile shaft resistance in saturated sands. The shaft capacity, Qf for cylindrical piles is given by: Q f fs As i 1 K i ( z )tan ( dL )
i n
(1)
1209
where fs skin friction; As surface area of the pile; z vertical effective stress along the pile shaft at depth z; L length of pile; d diameter of pile; angle of friction along the soil/pile interface and Ki coefficient of earth pressure along pile shaft. The above equation can be simplified by replacing the coefficients Ki and tan by a single factor Ki tan and forming the conventional method. Q f fs As
Extending the same philosophy, the contribution of ultimate shaft resistance due to matric suction along the interface of soil and pile material (Hamid and Miller, 2009) can be estimated using Eq. (6) Q( ua uw ) us As (6)
A general expression for estimating shaft resistance of piles can be obtained by substituting Eq. (5) and Eq. (6) in Eq. (3) as given below: Qf(us) Qf [(ua uw) (S) (tan )] dL (7)
i1 ( z )( dL )
i n
(2)
A wide range of recommendations are provided in the literature for values by various investigators (McClelland 1974, Meyerhof 1976, Briaud & Tucker 1997). Significant differences in the recommended values may be attributed to several factors such as in-situ stress conditions, frictional resistance, and compressibility of the soil, pile type, shape and mode of installation. In this paper, the values from the pile test results both under saturated and unsaturated conditions were back calculated. In addition, a simple technique is proposed to estimate the shaft resistance in unsaturated soil conditions using the Soil-Water Characteristic Curve (SWCC) and the effective shear strength parameters. 3 A TECHNIQUE FOR ESTIMATING THE SHAFT RESISTANCE OF PILES IN UNSATURATED SANDS
In Eq. 7, the fitting parameter value equal to 1 can be used for non-plastic soils such as sands (Vanapalli & Fredlund, 2000). Eq. (7) can be used for estimating the variation of shaft resistance of the pile with respect to matric suction, (ua uw) using the SWCC (i.e. relationship between the degree of saturation, S and matric suction). In the present study, SWCCs were measured using pressure plate apparatus. The measured SWCC results were also compared with the estimated SWCC using i) computer software SoilVision 2000 (Fredlund et al. 2002) which uses the grain size analysis data and the volume-mass properties, ii) one point measurement technique proposed by Vanapalli and Catana in 2005. 4 DETAILS OF THE TESTING PROGRAM
A general expression for estimating the shaft resistance of piles in unsaturated sands Qf(us), can be expressed as shown in Eq. (3); Q f ( us ) Q f Q( ua uw ) (3)
The contribution of ultimate shaft resistance due to matric suction Q( ua uw ) , can be estimated extending the approach proposed by Vanapalli et al. (1996) and Fredlund et al. (1996) for predicting shear strength of unsaturated soils using the SWCC and the effective shear strength parameters (Eq. 4). [c (n ua)tan ] [(ua uw) (S) (tan )] (4) where c effective cohesion, angle of internal friction, fitting parameter used for obtaining a best-fit between the measured and predicted values; and S degree of saturation. The second part of Eq. (4), represents the shear strength contribution due to matric suction: us [(ua uw) (S) (tan )] (5)
The geotechnical test pit at Lakehead University with the dimensions of 2.2, 4.4 and 2.5 meters (width, length and height respectively) was used in the research program. The test pit was divided into two smaller units (A and B) using timber lagging at the centre of the pit length (Figure 1). The sample soil was placed in pit A, while pit B was left empty and used only for controlling the suction values in pit A by changing the height of the water table during the tests. 4.1 Soil and pile properties Soil #1 is a well graded silty sand (42% silt 58% sand) and Soil #2 is poorly graded commercially available washed concrete sand (100% sand). Soils were subsequently placed into the test pit A in layers of 150 mm of thickness and compacted with a vibratory plate compactor. Mild steel open end pipe piles of 65 mm outer diameters were used in the present study. The shear strength parameters of the soil-steel interface were determined in cyclic-direct shear tests at constant normal load. The friction angles determined for soil and soil-steel interfaces (i.e., for both Soils #1, Soil #2) are summarized in Table 1 along with the other properties of the soils.
1210
Table 1. Properties of two soils tested. Soil Optimum water content, w Maximum dry unit weight, d Angle of friction. Soil/steel angle of friction, Placed water content, wp Placed dry unit weight, dp Placed total unit weight of unsaturated soil, unsat Degree of compaction Total unit weight of sat. soil sat % kN/m3 deg. deg. % kN/m3 kN/m3 % kN/m3 #1 12.5 18.7 40 36/28 8.5 19.4 21.0 104 22.0 #2 12.5 18.9 40 24/22 6.5 18.9 19.7 100 21.5 Figure 2. Test setup of model pile (a) loading of pile in compression, (b) shaft resistance of pile in tension.
*The effective cohesion is equal to zero for both of the soils.
dimensions: m 0.2
Timber Lagging Pit B PLAN VIEW CROSS SECTION

4.4
2.2
1.3
Pit A
CPT1
CPT2
Wire Piezometer Pile Tensiometers
Commercially available jet fill-type tensiometers, pressure cells and a vibrating wire piezometer were used to collect data during testing. Tensiometers were installed at 0.3, 0.6 and 1.2 meters below the soil surface to measure the matric suction (Fig. 1). A vibrating wire piezometer was placed in Soil #1 with the tip located approximately 1.0 m below the ground surface to measure pore water pressures in the soil during the fluctuations of water levels. Total pressure cells were also installed at four different elevations along the pit wall. The measured horizontal stresses increased almost linearly with depth resulting in earth pressure coefficients, Ki of approximately 0.7 for both soils in terms of total stresses. 4.3 Pile testing The piles were tested in axial compression and tension by loading in increments of approximately 20 N up to the point at which no further load increment could be sustained. Usually a testing set consisted of a test conducted in compression and followed by a tension test. Tests were carried out by varying (raising/ lowering) water levels representing the saturated and unsaturated conditions. Four different tests were conducted for Soil #1 as follows: Test 1: unsaturated, Test 2: saturated, Test 3: unsaturated again and tested after one week, and finally Test 4: the new pile installation. For Soil #2, the pile was repeatedly tested, alternating between tension and compression. This resulted in continuously decreasing pile shaft capacities. Three key tests were conducted using Soil #2. Test 1 was conducted under unsaturated condition. The pile was tested again after a period of 4 days, under unsaturated condition (which is Test 2), and finally the soil was saturated (submerged) and tested (which is Test 3).
Figure 1. study.
2.5
Schematic of the test pit used in the present
4.2
Instrumentation and pile installation
A load frame with 10 tonne capacity was used for pile installation and testing. After the installation of the piles, the soil inside the pile was removed with a hand auger and a void was created below the toe of the pile. The main purpose in creating the void below the pile toe was to eliminate the toe resistance during compressive loading, thus permitting direct measurement of shaft resistance in compression and in tension. Different views of the test setup of the model piles are shown in Figure 2. The test piles were not instrumented.
0.5
1211
4.4
Test results
7 6 5
Legend 1. Cycle 3. Cycle
1
The results of the pile tests that were accompanied by tensiometer readings are summarized for Soil #1 and Soil #2 in Figures 3 to 6. For each test, the maximum initial loads and the maximum loads that were experienced after the last load application are shown for compression and tension. For both soils, the shaft capacities in tension and compression decreased during the repeated loading. In each case, it was apparent that the cavity below the pile toe had collapsed after the water table had been raised to the soil surface. This was confirmed from the observations of soil movement into the hollow pile shaft which contributed to the development of some end bearing. In addition, the soil portions adjacent to the lower pile shaft had loosened up, resulting in a decrease of the shaft capacity. 4.4.1 Soil #1 (silty sand) After raising and lowering the water table, very low shaft capacities were observed in Test 3 when
7
Legend
Load (kN)
4 3
1 1 3.1 2.2 6 1.7 5
C - Compression T - Tension
1 1.5 5 1.0 4 1 1.6 1.0 1 5 1.5 1.1
2 1 0
C C
1.9
T T
A C
B T
K C
Test 1 Moist 1 (Unsaturated)
Test 3 Submerged
3
Placed Dry Unit Weight (18.9 kN/m )
Figure 5.
20
Summary of pile test results for Soil #2.
15
10
6 5
1. Cycle 3. Cycle
1 6 1
1 1
5.1
3 9
Load (kN)
4.7 4.0
Test 1
4.5
Test 2
150 200 250 300 350 400 450
4 3 2 1 0
C C 3.5
4.4
9
3.7
50
100
3.0
C - Compression T - Tension
1 1 5
Time (hours)
0.9 5 0.7
1 5
1.4 0.5
1.0
5
1 5
0.6
0.5 0.4 Z
Figure 6. Variation of tensiometer readings in Soil #2 with time prior to, during and after water sprinkling.
X
T T
A C
K C
Test 2 Submerged (Saturated)
Test 4 New Pile Installation
Placed Dry Unit Weight (19.4 kN/m 3)
Figure 3.
Summary of pile test results for Soil #1.
16
12
Raising Water Table
Lowering Water Table
0.3 m depth 0.6 m depth 1.2 m depth
compared to the Test 1 due to the collapse of the cavity (Fig. 3). For this reason, the pile results Test 2 and Test 3 are considered not fully comparable to Test 1. The pile was removed and subsequently reinstalled about 0.6 m away from the original location (i.e. Test 4). The subsequent load tests after the new pile installation showed slightly higher capacities than Test 1 reflecting the slightly densified conditions due to the effective stress changes experienced due to water table fluctuations. The matric suction values measured during the test at 0.3 m, 0.6 m and 1.2 m are given in Fig. 4. Even under submerged conditions, the matric suction values never reached zero, but stayed for all three instruments in the range of 0.9 to 1.5 kPa.
275
25
50
75
100
125
150
175
200
225
250
Time (hours)
Figure 4. Matric suction changes during raising and lowering the water table level for Soil #1.
4.4.2 Soil #2 (concrete sand) In Test 1, load dropped to 2.2 kN at the end of first cycle and to 1.7 kN after the sixth load application in compression. In tension this values are 3.1 kN and 1.9 kN for respective loading cycles (Fig. 5).
1212
The continuously decreasing shaft capacities can be associated with repeated alteration in loading conditions. After four days, the pile was retested (i.e. Test 2). It was observed that the capacities of the pile in compression and in tension had decreased further. Finally, under submerged condition only the tensile capacities could be recorded due to the same reason as in Soil #1 (the cavity below the pile toe had collapsed when the water table was raised). The matric suction values were measured in Soil #2 before raising and lowering the groundwater level. The matric suction values were decreased at all three levels by adding water to the soil surface using a garden-sprinkler. The largest decrease was observed near the surface, changing from 10 kPa to 3 kPa almost immediately. The smallest change occurred at the lowest tensiometers with 6 hours delay from 18 kPa to 16.2 kPa (Fig. 6). The tensiometers readings increased again after ending the water application to the soil surface. 5 ESTIMATING THE SHAFT RESISTANCE
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.01 0.1 1 10 100 1000
Experimental From SoilVision One point method
Figure 7. Measured and predicted soil-water characteristic curve for Soil #1.
1.0 0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0.01 0.1 1 10 100 1000
Experimental From SoilVision One Point method
The shaft resistances of the tested piles were also calculated using Eq. 7. A value of 0.35 was chosen, which is an average value of various methods summarized in the literature (McClelland 1974, Meyerhof 1976, Briaud & Tucker 1997). Matric suction values used in the calculations are obtained from the tensiometer readings for the related depths. The corresponding degree of saturation values were estimated from the SWCCs for the two soils (Fig. 7 and 8). The computed results (calculated and measured) are summarized in Table 2 and Table 3 for Soil #1 and Soil #2 respectively to provide comparisons. Two key observations can be made from the summarized results. The first observation is related to the contribution of matric suction to the total shaft resistance and the second is related to the difference between the measured and calculated total shaft resistances. The total shaft resistance was found to be 35%40% of the total shaft capacity in the Soil #1 which was silty in nature. However, contribution of matric suction to the total shaft resistance was almost negligible in Soil #2 which was clean concrete sand. Since the matric suction contribution is the same from both the measured and calculated values, it can be concluded that the difference between them can be attributed to the values which are dependent on the stress condition of the soil. For this reason, values were back calculated to understand the influence of unsaturated soil conditions.
Figure 8. Measured and predicted soil-water characteristic curve for Soil #2.
5.1
Back calculated values
The values were back calculated from the measured shaft capacities for three different conditions (i) saturated condition Eq. (8); (ii) unsaturated condition (with matric suction) Eq. (9) and (iii) unsaturated condition without matric suction Eq. (10).
L dL 2 Q f ( us ) Q( ua uw ) L dL 2
Qf
(8)
(9)
Q f ( us ) L dL 2
(10)
The back calculated values of using Eqs. (8), (9) and (10) are summarized in Table 4 using first cycle loads for Soil #1 and Soil #2 for compression only. Appropriate effective unit weights values from
1213
Table 2. Comparison between the measured and calculated shaft resistances for Soil #1 ( 0.35). Shaft length intervals Test # T1 T2 T4 (m) 0.45 0.45 0.60 0.45 0.45 0.60 0.45 0.45 0.60 Degree of saturation (S) 0.780 0.765 0.775 0.985 0.999 0.999 0.810 0.755 0.755 Cal. Q(uauw) kN Qf(us) kN 3.1 0.74 3.11 Meas. Qf(us) kN 4.7 0.9 5.1
Matric suction kPa 13.5 15.0 14.0 1.50 1.00 0.90 11.0 16.0 16.0
} } }
1.84 0 1.82
Table 3. Comparison between the measured and calculated shaft resistances for Soil #2 ( 0.35). Shaft length intervals Test # T1 T2 (m) 0.45 0.45 0.60 0.45 0.45 0.60 Degree of saturation (S) 0.025 0.048 0.050 0.030 0.050 0.060 Cal. Q(uauw) kN Qf(us) kN 1.25 1.24 Meas. Qf(us) kN 2.2 1.5
Matric suction kPa 18.0 12.0 10.0 16.5 10.0 8.00
} }
0.057 0.055
Table 4. The back calculated values from measured shaft capacities for Soil #1 and Soil #2 for compression. Meas. Q(uauw) Soil type #1 #2 Qf(us) Test # T1 T4 T2 T1 T2 kN 4.7 5.1 0.9 2.2 1.5 kN 1.84 1.82 0 0.057 0.055 Back calculated values Eq#8 0.43 Eq#9 0.8 0.9 0.66 0.44 Eq#10 1.30 1.41 0.64 0.43
In Soil #2, which is clean sand with zero fine content, the back calculated values were approximately the same for both saturated and unsaturated conditions. 6 SUMMARY
Table 1 were used in the calculations. The back calculated values (Table 4) for Soil #1 are approximately 35%40% lower if the influence of matric suction is ignored. For saturated condition in Soil #1, value is much lower as it reflects the effect of looser soil conditions near the pile shaft after the collapse of soil in to the cavity below the pile tip.
The shaft resistance of steel pipe test pile that was jacked hydraulically into two different sands was determined under saturated and unsaturated conditions in an instrumented geotechnical test pit. The objective of this study was to determine the contribution of matric suction towards the total shaft resistance of piles and propose a technique to estimate the shaft resistance in unsaturated soil conditions. The contribution of matric suction was found to be between 35%40% of the total shaft capacity of the test piles for silty sand but almost negligible for clean sands. Thus, it can be concluded that in unsaturated soils, with the exception of clean sands and granular soils, the contribution
1214
of matric suction to the total shaft capacity of piles cannot be neglected. In addition, a simple technique was proposed to estimate the shaft resistance in unsaturated soil conditions using the Soil-Water Characteristic Curve (SWCC) and the effective shear strength parameters conventional -method. The results of this study are promising to predict the contribution of matric suction towards shaft resistance using well established -method for sandy soils in the literature. However, more pile load tests need to be performed and evaluated to better understand the contribution of matric suction towards the shaft capacity of piles in different types of sandy soils. ACKNOWLEDGEMENTS This paper is dedicated to late Dr. Eigenbrod from Lakehead University. Authors acknowledge that some of the data presented in this paper were collected by Civil Engineering Students of Lakehead University during a series of undergraduate degree projects.
REFERENCES
Briaud, J.L. & Tucker L. 1997. Design and Construction Guidelines for Downdrag on Uncoated and BitumenCoated Piles, Report 393, National Cooperative Highway Research. Costa, Y.D., Cintra, J.C. & Zornberg, J.G. 2003. Influence of matric suction on the results of plate load tests performed on a leteritic soil deposit. ASTM Geotechnical Testing Journal, 26(2): 219227. Douthitt, B., Houston, W. & Houston, S. 1998. Effect of wetting on pile friction. Proceedings of the Second International Conference on Unsaturated Soils, Beijing, 2730 August 1998, 1: 219224, International Academic Publishers. Fredlund, M.D. & Wilson, G.W. 2002. Use of the grainsize distribution for estimation of the soil-water characteristic curve. Canadian Geotechnical Journal 39(5): 11031117. Fredlund, D.G., Xing, A., Fredlund, M.D. & Barbour, S.L. 1996. The relationship of the unsaturated soil shear
strength to the soil-water characteristic curve. Canadian Geotechnical Journal 33: 440448. Georgiadis, K., Potts, D.M. & Zdravkovic, L. 2003. The influence of partial soil saturation pile behaviour. Geotechnique 53(1): 1125. Hamid, T. & Miller G.A. 2009. Shear strength of unsaturated soil interfaces. Canadian Geotechnical Journal 46(5): 595606. McClelland, B. 1974. Design of deep penetration piles for ocean structures. Journal of Geotechnical Enineering Division ASCE 100(GT 7): 705747. Meyerhoff, G.G. 1976. Bearing capacity and settlement of pile foundations., Journal of Geotechnical Enineering Division ASCE 102(GT3): 195228. Mohamed, F.M.O. & Vanapalli, S.K. 2009. Experimental investigations for determining the bearing capacity of unsaturated soils from cone penetration tests. 62nd Canadian Geotechnical Conference, Halifax, 2024 September 2009: 205213. Mohamed, F.M.O. & Vanapalli, S.K. 2006. Laboratory investigations for the measurement of the bearing capacity of an unsaturated coarse-grained soil. 59th Canadian Geotechnical Conference, Vancouver. 14 October 2006: 219226. Oh, W.T. & Vanapalli, S.K. 2009. A simple method to estimate the bearing capacity of unsaturated finegrained soils, 62nd Canadian Geotechnical Conference, Halifax, 2024 September 2009: 234241. SoilVision. 2000. Version 3.04. Software Package for Modelling the Engineering Behavior of Unsaturated Soils. Soil Vision Systems Ltd. Vanapalli, S.K. & Mohamed, F.M.O. 2007. Bearing capacity of model footings in unsaturated soils. Experimental Unsaturated Soils Mechanics, Ed. Schanz, T. Springer Proceedings in Physics 112: 483494. Vanapalli, S.K. & Catana, M.C. 2005. Estimation of the soil-water characteristic curve of coarse grained soils using one point measurement and simple properties. International Symposium on Advanced Experimental Unsaturated Soil Mechanics, Trento, 2729 June 2005: 401407, Taylor and Francis. Vanapalli, S.K. & Fredlund, D.G. 2000. Comparison of empirical procedures to predict the shear strength of unsaturated soils using the soil-water characteristic curve. The Proceedings of Unsaturated Soil sessions of Geo-Denver 2000, Denver, 58 August 2000, 99: 195209, American Society of Civil Engineers, Special Publication. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction. Canadian Geotechnical Journal 33(3): 379392.
1215
Bearing capacity of an unsaturated sand from model footing tests

S.K. Vanapalli, R. Sun & X. Li
Civil Engineering Department, University of Ottawa, Ottawa, ON, Canada
ABSTRACT: A series of model footings tests were conducted on compacted sand using a specially designed equipment to investigate the variation of bearing capacity with respect to matric suction. The bearing capacity values were determined using three different sizes of square model footings (i.e. 20 mm 20 mm; 25 mm 25 mm; 37.5 mm 37.5 mm) over a suction range of 0 to 12 kPa. The experimental results show that there is a non-linear increase in the bearing capacity of the tested sand with respect to matric suction. The variation of the bearing capacity with respect to matric suction was also predicted using the Soil-Water Characteristic Curve (SWCC) and the effective shear strength parameters for one of the model footings. Discussions are offered with respect to the bearing capacity of sand including the influence of footing size on the bearing capacity test results in this paper. 1 INTRODUCTION the Soil-Water Characteristic Curve (SWCC) and the saturated shear strength parameters. In this paper, experimental studies related to the variation of bearing capacity with respect to matric suction determined in a laboratory environment using different sizes of model footing tests on compacted unsaturated sand using specially designed equipment are summarized. In addition, an attempt is also made to predict the variation of the bearing capacity with respect to matric suction for a model footing using the approach presented by Vanapalli & Mohamed (2007). 2 BACKGROUND
The conventional soil mechanics principles are used in engineering practice for saturated sands, silts and clays, and dry sands. However, soils could either be saturated with water or have other fluids in the voids (e.g. air) together with water. There are numerous soils; for example, expansive, residual, compacted or collapsible soils that are commonly encountered in engineering practice in a state of unsaturated condition. About 33% of the earths surface is considered either arid or semi-arid and the soils in these regions are typically in a state of unsaturated condition. The mechanics of saturated soils are not valid in the interpretation of the engineering behavior of unsaturated soils. Shallow foundations are usually located above the ground water table where the soil is typically in a state of unsaturated condition. Bearing capacity is one of the key parameters required in the design of shallow foundations. Significant research has been undertaken to study the bearing capacity of saturated soils. However, the influence of matric suction, (ua uw) which is defined as the difference between the pore air pressure ua and pore water pressure uw is conventionally not taken into account in the bearing capacity of soils. Several studies have shown that the bearing capacity significantly increases due to contribution of matric suction (Broms 1964, SteensenBach et al. 1987, Fredlund & Rahardjo 1993, Schnaid et al. 1995, Miller & Muraleetharan 1998, Oloo et al. 1997, Costa et al. 2003, Yongfu 2004). More recently, some studies are undertaken to predict the bearing capacity of unsaturated sand and silty soils (Vanapalli & Mohamed 2007, Vanapalli et al. 2007, Oh & Vanapalli 2008) using
Terzaghi (1943) suggested Equation 1 to estimate the ultimate bearing capacity of saturated soils: qu cNc qNq 0.5BN (1)
where qu ultimate bearing capacity; c effective cohesion; q effective overburden pressure; and B width of the footing. The bearing capacity factors associated with cohesion, surcharge and unit weight are Nc, Nq and Nr respectively which are a function of angle of internal friction, . Eq. 1 is commonly used in the estimation of bearing capacity of soils in engineering practice for general shear failure conditions. Vesi (1973) proposed an improved bearing capacity approach based on Terzaghis theory by introducing new bearing capacity factors and shape factors as follows: qult cNcvc qNqvq 0.5BNv (2)
1217
where Ncv, Nqv and Nv are the bearing capacity factors associated with cohesion, surcharge and unit weight respectively. c 1 (B/L)(Nq/Nc); q 1 (B/L)tan ; c 10.4(B/L). Mohamed & Vanapalli (2006) carried out model footing tests in a specially designed bearing capacity tank (University of Ottawa Bearing Capacity Equipment, UOBCE, 900 900 750 mm), which has provisions to simulate both saturated and unsaturated conditions in the tank. Two different sizes of model footings (i.e. 100 mm 100 mm and 150 mm 150 mm) were used to determine the variation of bearing capacity with respect to matric suction of compacted sand. Based on these tests and using other model footing studies results from the literature, Vanapalli & Mohamed (2007) proposed an equation to both interpret and predict the bearing capacity of surface footings as follows:
qu c ua uw b 1 S N c 1.0

Nq B B 0.5 B N 1.0 0.4 (3) L Nc L
tan u
uw AVR S tan
where (ua uw)AVR [(ua uw)1 (ua uw)2]/2, (ua uw)1 is the matric suction close to the surface of the footing and (ua uw)2 is the matric suction at the depth of 1.5 B (B is the width of the footing); (ua uw)b air entry value, kPa, S is the degree of saturation, L is the length of footing, is soil unit weight, Nc and Nq are bearing capacity factors from Terzaghi (1943), N is bearing capacity factor from Kumbhokjar (1993). Vanapalli & Mohamed (2007) also provided a relationship (Figure 1) between the fitting parameter, and the plasticity index, Ip that can be used in the prediction of the bearing capacity of different finegrained soils. Equation (3) is valid for interpreting the bearing capacity of unsaturated soils assuming
8 7
Botkin Pit Silt
drained loading conditions. There is a smooth transition between the bearing capacity of unsaturated and saturated soils. In this approach, the footing failure is assumed to be consistent with the general shear failure conditions as defined by Terzaghi (1943). Oh & Vanapalli (2008) conducted finite element analysis using SIGMA/W of GEO-SLOPE to model the stress versus settlement behavior of foundations in unsaturated sands simulating the testing conditions summarized in Vanapalli & Mohamed (2007). The results of the modeling studies show that the conventional stress distribution theory concepts extending elastic analysis can be used to estimate the stresses below foundation (Poulos & Davis 1974). In addition, the assumption of calculating the average matric suction value from 0 to 1.5 B depth region is reasonable as it provided good comparison between the measured and predicted bearing capacity values for sands in unsaturated conditions. These results also show that the bearing capacity of model footings of 50 mm width or lower can be conducted using relatively smaller size tanks of 300 mm diameter. In other words, the results using tank sizes greater than 300 mm also provide similar results when the model footings sizes are less than 50 mm. In the present study, several square model footings (i.e. 20 mm 20 mm; 25 mm 25 mm; 37.5 mm 37.5 mm) were used to determine the variation of the bearing capacity of sand with respect to matric suction in specially designed equipment. Also, comparisons are provided between the measured bearing capacity values for 37.5 mm 37.5 mm model footing tests and predicted bearing capacity values using Equation 3. 3 EQUIPMENTS AND METHODOLOGY
3.1 Bearing capacity test system Figure 2 shows the bearing capacity test system which includes a soil container and other accessories. The bearing capacity of model footings placed on the surface of the compacted soil specimen in the soil container under both saturated and unsaturated conditions was determined by loading them to failure conditions using a robust loading frame. 3.2 Soil container
Fitting parameter,
6 5 4 3 2 1 0 0 5 10 15
3 different sands
Glacial Till
= -0.0027(Ip2 )+0.3987(Ip)+1
20
Plasticity index, Ip
Figure 1. Relationship between the bearing capacity fitting parameter, and the plasticity index, Ip (Vanapalli and Mohamed 2007).
The soil container has different elements to facilitate saturation and desaturation of the soil sample in the container. Figure 3 shows the schematic of the soil container with various elements. Figure 4 shows different elements of the soil container. Figure 4a shows the base plate with circular grooves to facilitate drainage of water both into and out of the container. The circular grooves
1218
Figure 2. Bearing capacity test system with tall soil container.
Figure 3. Schematic of the soil container with various elements.
Figure 4. Various components of the soil container.
of the bottom plate are in turn covered using a stainless steel lid (Figure 4b) of 260 mm in diameter and 1.25 mm in thickness with small holes (2 mm in diameter). This plate facilitates gradual (i.e. steady state) movement of water into and out of the soil container. On the lid a geotextile and a filter paper are provided (Figure 4c) to prevent the soil particles
clogging into the circular grooves of the base plate and valves. The water table in the soil container can be lowered to a maximum depth of 700 mm below soil surface. The matric suction values below the footing can be varied by adjusting the depth of water table below the footing. The bearing
1219
capacity of the model footings was determined under different matric suction values following the procedure described in later sections. 3.3 Suction profile set
The relationship between matric suction and depth of ground water table was measured using a specially designed unit which is referred to as suction profile set in this paper. The suction profile set consists of a long transparent acrylic soil column, a water container, thin plastic drain tube, valves and a pulley block. Suction profile set was fixed on the left post of bearing capacity frame (see Figure 2, item 11). The relationship between matric suction and depth of ground water table was pre-calibrated using this unit in order to achieve the required matric suction below model footing in the soil container of bearing capacity test system by simply adjusting the depth of the water table and allowing it to reach equilibrium conditions.
This technique facilitates as an additional check to estimate the matric suction if the tensiometers are not working. Also, the placement of several tensiometers in the soil container can be reduced to measure the matric suction below the ground water table. This technique also facilitates in the reduction of the errors associated with measurement of low matric suction values using tensiometers due to possible disturbance of the soil below the footing during the loading stage. 4 TEST RESULTS AND ANALYSIS
Tensiometer
Depth of water table
Unsaturated soil
Saturated soil Water Container
Bearing capacity tests were conducted on compacted sand under both saturated and unsaturated conditions using the bearing capacity test system in the present research program. The grain size distribution of sand is plotted in Figure 6. The effective shear strength parameters of sand were determined using direct shear apparatus (i.e. c and were equal to 2.1 kPa and 35.3, respectively). The measured dry density value of the sand from the container was 14.7 kN/m3. The bearing capacity of model footings was determined over a matric suction range of 0 to 12 kPa. This range of suction extends over the boundary effect, transition and residual stages (Vanapalli et al. 1996) for the sand studied in the present research program. Figure 7 shows the SWCC of sand which was measured using the Tempe cell. Three model footings of different sizes (i.e. 20 mm 20 mm; 25 mm 25 mm; 37.5 mm 37.5 mm) were used to study the influence of footing size on the bearing capacity of unsaturated sand. The details of the test program are summarized in Table 1. The matric suction values in the soil container were controlled by adjusting the water head and also confirmed by tensiometers which were inserted into the soil for several series of test results.
100
80 Percent passing (%)
60
Soil Column
40
20
0 0.01
0.1
1 Particle size (mm)
10
100
Figure 5.
Schematic diagram of suction profile set.
Figure 6. Grain size distribution of sand used for the present study.
1220
100
800
12 kPa 10 kPa
80
Applied pressure, p (kPa)
600
8 kPa 6 kPa
4 kPa
60
400
40
2 kPa Saturated 2 kPa 4 kPa 6 kPa 8 kPa 10 kPa 12 kPa
20
200
Saturated
0 0.1
1 Matric suction, (ua- uw) (kPa)
10
10
12
14
16
18
Settlement, S (mm)
Figure 7.
Measured SWCC of sand.
Table 1. Combination of footing size and matric suction for bearing capacity test of sand. Footing size (mm) 20 20 25 25 37.5 37.5 Matric suction (kPa) Saturated, 2, 4, 6, 8, 10, 12, 14, 18 Saturated, 2, 4, 6, 8, 10, 12, 14, 18 Saturated, 2, 4, 6, 8, 10, 12
Figure 8. Applied pressure versus settlement for footing of 37.5 mm 37.5 mm.
1000 Ultimate Bearing Capacity, q (kPa)
800
600
400
Footing 37.5 mm x 37.5 mm Footing 25 mm x 25 mm Footing 20 mm x 20 mm
The compacted sand in the container was first saturated by controlling the water head until the water table reaches the soil surface. For the tests in unsaturated conditions, the average matric suction values within the stress bulb corresponding to the water table depths of 240, 440 and 640 mm were 2, 4 and 6 kPa, respectively. To obtain the matric suction of 8 kPa and higher, the water table in the container was lowered beneath the bottom of soil container and let the water drain out freely. For matric suction values of 2, 4, and 6 kPa, a time period of 24 hours was required to achieve these values after lowering the water table. However, more time was needed for the matric suction of 8 kPa and higher (i.e. one to three days). Different suction values were observed from tensiometers that were placed close to the top surface of the soil in certain cases. These differences may be attributed to evaporation losses and measurement of matric suction in the sheared zone close to the model footing. Figure 8 shows the applied pressure versus settlement behavior obtained from bearing capacity tests performed with the model footing of 37.5 mm 37.5 mm. From the test results, general shear failure were generally observed. However, well defined general shear failure was not observed when the degree of saturation was higher and matric suction values below the footing were lower (i.e. 2 kPa and 4 kPa; see Figure 7 and 8). On the other hand, well defined general shear failure characteristics were always observed when matric suction values
200
10
12
14
16
18
20
Suction, ua-uw (kPa)
Figure 9. Relationship of bearing capacity versus matric suction.
were higher. This behavior may in part be attributed to using relatively small size footings. The variation of the bearing capacity with respect to different matric suction values for three model footings are shown in Figure 9. The results demonstrate that there is a significant increase in the bearing capacity due to the contribution of matric suction in the range 0 to 6 kPa. For the matric suction values higher than 6 kPa, the bearing capacity is relatively constant. The degree of saturation is low for suction values greater than 6 kPa and hence the contribution of matric towards bearing capacity (i.e. (ua uw)AVRS in Equation 3) is small. In addition, the corresponding density, value decreases as water content reduces due to desaturation at higher suction values. Figure 9 also shows that the size of model footings (i.e. 20 mm 20 mm; 25 mm 25 mm; 37.5 mm 37.5 mm) almost has limited influence on the bearing capacity under various matric suction values. This is because the model footings sizes used in the study fall in a small range.
1221
800 Footing 37.5 mm x 37.5 mm Ultimate Bearing Capacity, q (kPa) 700 600 500 400 300 200 100 Measured values Predicted curve
REFERENCES
Broms, B.B. 1964. The effect of degree of saturation on the bearing capacity of flexible pavements. Highway Research Record 71: 114. Costa, Y.D., Cintra, J.C. & Zornberg, J.G. 2003. Influence of matric suction on the results of plate load tests performed on a lateritic soil deposit. Geotechnical Testing Journal 26(2): 219226. Fredlund, D.G. & Rahardjo, H. 1993. Soil Mechanics for Unsaturated Soils. New York: John Wiley & Sons, INC. Kumbhojkar, A.S. 1993. Numerical evaluation of Terzaghis Nr. J. Geotech. Eng., ASCE, 119(3), 598. Miller, G.A. & Muraleetharan, K.K. 1998. In situ testing in unsaturated soil. Proceedings of the 2nd International Conference on Unsaturated Soils 1: 416421. Beijing, China. Mohamed, F.M.O & Vanapalli, S.K. 2006. Laboratory investigations for the measurement of the bearing capacity of an unsaturated coarse-grained soil. Proceedings of the 59th Canadian Geotechnical Conference. Vancouver, British Columbia, Canada. Oh, W.T. & Vanapalli, S.K. 2008. Modelling the stress versus settlement behavior of model footings in saturated and unsaturated sandy soils. In Proceedings of the 12th International Conference of International Association for Computer Methods and Advances in Geomechanics (IACMAG), Goa, India. Oloo, S.Y., Fredlund, D.G. & Gan, J.K-M. 1997. Bearing capacity of unpaved roads. Canadian Geotechnical Journal 34: 398407. Poulos, H.D. & Davis, E.H. 1974. Elastic solutions for soil and rock mechanics. John Wiley and Sons, New York. Schnaid, F., Consoli, N.C., Cudmani, R.O. & Militistsky, J. 1995. Load-settlement response of shallow foundations in structured unsaturated soils. Proceedings of the 1st International Conference of Unsaturated Soils: 9991004. Paris, France. Steensen-Bach, J.O., Foged, N. & Steenfelt, J.S. 1987. Capillary induced stressesfact or fiction? 9th ECSMFE, Groundwater Effects in Geotechnical Engineering: 8389. Dublin. Terzaghi, K. 1943. Theoretical Soil Mechanics. New York: John Wiley and Sons, INC. Vanapalli, S.K., Fredlund, D.G., Pufahl, D.E. & Clifton, A.W. 1996. Model for the prediction of shear strength with respect to soil suction. Can. Geotech. J. 33(3): 379392. Vanapalli, S.K. & Mohamed, F.M.O. 2007. Bearing capacity of model footings in unsaturated soils, Experimental Unsaturated Soil mechanics. Proceedings of the 2nd International Conference on Unsaturated Soils: 483493. Weimar: Springer. Vanapalli, S.K., Oh, W.T. & Puppala, A.J. 2007. Determination of the Bearing Capacity of Unsaturated Soils Under Undrained Loading Conditions. Proc. of 60th Canadian Geotechnical Conference & 8th Joint CGS/IAH-CNC Groundwater Conference, T6-A, Ottawa, Canada. Vesic, A.S. 1973. Analysis of ultimate loads of shallow foundations. Journal of the Soil Mechanics and Foundation Division, ASCE 99(SM1): 4573. Yongfu Xu 2004. Bearing capacity of unsaturated expansive soils. Geotechnical and Geological Engineering 22: 611625.
10
12
Suction, ua-uw (kPa)
Figure 10. Comparison of measured and predicted bearing capacity.
Equation 3 is used in the prediction procedure to provide comparisons with the measured values of bearing capacity for a typical model footing (i.e. 37.5 mm 37.5 mm) as similar trends in experimental results were observed with other two smaller size footings. Figure 10 summarizes the bearing capacity results from experimental studies and predicted values. There is reasonably good agreement between the measured and predicted bearing capacity. Problems associated with the estimation of higher suction values have contributed to differences in predicted and measured bearing capacity values. More studies are in progress to better understand the problems and revise the protocols for future studies. 5 SUMMARY AND CONCLUSION
Model footing tests were conducted to determine the bearing capacity of saturated and unsaturated soil under drained loading conditions using a strain rate of 1.143 mm/min. The measured bearing capacity values of footings (i.e. 20 mm 20 mm; 25 mm 25 mm; 37.5 mm 37.5 mm) are close to each other as the model footings sizes range fall in a narrow band. The nonlinear increase of the measured bearing capacity values was observed for all the model tests. The equation for predicting bearing capacity of unsaturated soils using the saturated shear strength parameters, c and and the SWCC is promising and be used by the practicing engineers. ACKNOWLEDGEMENTS The authors would like to thank Geotechnical Lab Technician Kulan Ambalavanar and Student Workshop Technician John Perrins for their time and efforts in the fabrication of the equipments used in this study.
1222
Interpretation of the bearing capacity of unsaturated soils extending the effective and the total stress approaches
Sai K. Vanapalli & Won T. Oh
Civil Engineering Department, University of Ottawa, ON, Canada
ABSTRACT: The model footing and in-situ plate load tests results available in the literature were reanalyzed extending three different approaches to interpret the bearing capacity of unsaturated fine-grained soils; i) the effective stress approach by Terzaghi (1943) ii) the total stress approach by Skempton (1948) and iii) the reduction factors approach by Terzaghi (1943). The analysis of the results show that the bearing capacity values estimated extending the total stress approach using the unconfined compressive strength results for unsaturated fine-grained soils are close to the measured results in comparison to the other two methods. The analysis of the bearing capacity results on coarse-grained soils support using the effective stress approach by Terzaghi (1943). 1 INTRODUCTION conditions of pore-air and pore-water during loading stages can influence the bearing capacity of coarse- and fine-grained unsaturated soils. Our present understanding of the bearing capacity of unsaturated soils is limited as only few studies are reported in the literature with respect to the bearing capacity of unsaturated fine-grained soils. In the study presented in this paper, model footing and in-situ plate load tests results on unsaturated fine-grained soils available in the literature (Costa et al. 2003, Rojas et al. 2007, Vanapalli et al. 2007) are reanalyzed extending both the ESA and the TSA. Based on analysis of the results, some recommendations are made with respect to the interpretation of the bearing capacity of both coarse- and fine-grained unsaturated soils. 2 BACKGROUND
Bearing capacity of shallow foundations in saturated soils are conventionally interpreted using either the effective (i.e. c, ; Terzaghi 1943) or the total stress (i.e. u 0; Skempton 1948) approach depending on drainage conditions during the loading stages. However, there are limited studies reported in the literature with respect to the interpretation of the bearing capacity of unsaturated soils. Terzaghi (1943) bearing capacity theory (i.e. Effective Stress Approach; hereafter referred to as ESA) was extended by Vanapalli & Mohamed (2007) to interpret the variation of bearing capacity with respect to matric suction for three different coarse-grained soils. Reasonably good comparisons were observed between the measured and the predicted bearing capacity values. Schnaid et al. (1995) and Costa et al. (2003) carried out in-situ plate load tests to study the effect of matric suction on the bearing capacity of unsaturated fine-grained soils. They estimated the bearing capacity values extending the ESA; however, the bearing capacity values were overestimated when compared with the measured values. Recently, Vanapalli et al. (2007) conducted model footing (B L 50 50 mm) tests on a compacted glacial till. They suggested that the bearing capacity of unsaturated fine-grained soils can be reliably interpreted extending Skempton (1948) bearing capacity theory (i.e. total stress approach; hereafter referred to as TSA). This approach needs the unconfined compressive strength results of unsaturated fine-grained soils. Several parameters such as type of soil (i.e. coarse- or fine-grained soils) and drainage
2.1 Bearing capacity of unsaturated soils extending Effective Stress Approach (i.e. ESA) Oloo et al. (1997) proposed a method to estimate the bearing capacity of a surface footing on unsaturated soils extending the ESA as given below: qult ( unsat ) {c (ua uw )b tan [(ua uw ) (ua uw )b ]tan b } 1 B N 2 (1)
where qult(unsat) ultimate bearing capacity of unsaturated soil, c, effective cohesion and internal friction angle, respectively, (ua uw)b air-entry value, (ua uw) matric suction, B width of footing,
1223
soil unit weight, N bearing capacity factor; and b internal friction angle indicating the rate of increase in shear strength related to the suction. Eq. (1) is based on the assumption that the shear strength failure envelope of unsaturated soils is bilinear for simplification purposes (Gan et al. 1988). In other words, the bearing capacity of unsaturated soils with respect to suction linearly increases with an angle of up to the air-entry value and then with a constant angle of b (i.e. ) for suction values greater than the air-entry value, (ua uw)b (Figure 1). The equation proposed by Oloo et al. (1997) (i.e. Eq. (1)) has limitations with respect to interpreting the bearing capacity behavior of unsaturated soils over a large suction range. This is because the bearing capacity of unsaturated soils is nonlinear for suction values greater than the air-entry value, (ua uw)b both for fine- and coarse-grained soils (see curves (ii) and (iii) respectively in Figure 1). In case of fine-grained soils, the bearing capacity increases as the suction increases and likely converges to a certain value (curve (ii) in Figure 1). On the other hand, for coarse-grained soils, the net contribution of matric suction towards bearing capacity decreases as the matric suction approaches the residual suction value. The b values become negative with further increment of suction values (see curve (iii) in Figure 1; Oh et al. 2009). This implies that the bearing capacity values will be overestimated when interpreted using constant b value (i.e. Eq. (1)) for high suction values. Schnaid et al. (1995) carried out in-situ plate (0.3, 0.45, 0.6, 0.7 and 1 m) load tests for unsaturated fine-grained soils. The bearing capacity values interpreted extending the ESA was 4 to 6 times greater than the measured values. The same trend was also observed for the in-situ plate (dia. 0.8 m) load tests results by Costa et al. (2003). The discrepancy between the measured and the estimated bearing capacity values extending the ESA can be attributed to two key reasons. Firstly, the drainage conditions of pore-air and pore-water during loading stages in unsaturated
fine-grained soils are not well defined (i.e. it may not be representing fully drained condition). Secondly, the bearing capacity equation originally proposed by Terzaghi (1943) was based on the General Shear Failure (hereafter referred to as GSF) criteria assuming drained loading conditions. In most cases, well-defined GSF modes are not observed for both model footing and in-situ plate load tests from the stress versus settlement relationships in unsaturated fine-grained soils. The punching shear failure dominates in unsaturated fine-grained soils in many cases (Oloo 1994, Schnaid et al. 1995, Costa et al. 2003, Rojas et al. 2007, Vanapalli et al. 2007). Schnaid et al. (1995) and Costa et al. (2003) also estimated the bearing capacity values with reduced effective shear strength parameters (i.e. two-third of the initial values; Eqs. and ), which is the conventional approach for interpreting local shear failure conditions (Terzaghi 1943). c* 0.67c tan* 0.67 tan (2) (3)
Bearing capacity
' > b
b
(iii) coarse-grained soils
where c*, * modified effective cohesion and effective internal friction angle, respectively. Reasonably good agreement between the measured and the estimated bearing capacity values was observed using the reduction factors approach for the results presented by Schnaid et al. (1997). However, the estimated bearing capacity values were still higher than the measured values by 3 to 5 times for the results by Costa et al. (2003). Oloo (1994) also extended a similar approach earlier to interpret the model footing (width and dia. 30 mm) tests results for two different compacted unsaturated fine-grained soils (i.e. silt and till). The analysis results showed that the reduction factors approach could be applied for the glacial till, but not for the silt. The studies by Oloo (1994), Schnaid et al. (1995), and Costa et al. (2003) indirectly suggest that using the reduced shear strength parameters should not be generalized or extended for all types of unsaturated fine-grained soils and suction values. Extending the concept in Eq. (1), Vanapalli & Mohamed (2007) proposed a model to predict the nonlinear variation of bearing capacity of unsaturated soils with respect to matric suction for surface footings as follows: qult ( unsat ) [c (ua uw )b (1 S )tan (ua uw )AVR S tan ] Ncc 0.5B N
' = b
(ua - aw)b
(ii) fine-grained soils (i) bilinear
Suction (ua - uw)
(4)
Figure 1. Variation of bearing capacity with respect to suction for different types of soils.
where (ua uw)AVR average matric suction value, S degree of saturation, fitting parameter with
1224
respect to bearing capacity, which is a function of IP, Nc bearing capacity factor from Terzaghi (1943), N bearing capacity factor from Kumbhokjar (1993); and c, shape factors from Vesi (1973). Vanapalli & Mohamed (2007) suggested a relationship between the fitting parameter, in Eq. (4) and plasticity index, IP after analyzing five sets of data (i.e. three sands (Mohamed & Vanapalli 2006, Steensen-Bach et al. 1987) and silt and till (Oloo 1994)) (Figure 2). 0.0031(IP)2 0.3988(IP) 1 (5)
140
Ultimate bearing capacity
120 100 80 60 40
Measured (SM) Predicted (SM) (= 3.5) Measured (QM) Predicted (QM) (= 3.5)
(a)
Costa et al. 2003 Dia. = 800 (mm) IP = 8 (%)
10
15
20
25
30
35
Matric suction, (ua - uw) (kPa)

1000 800 600 400 200 0
Rojas et al. 2007 Dia. = 310 (mm) Ip = 12 (%) Measured Predicted ( = 3.5) (b)
Two sets of in-situ plate load tests results (Costa et al. 2003, Rojas et al. 2007) are analyzed in the present study using Eq. (4) (Figure 3(a) and (b)). The terms, SM and QM in Figure 3(a) stand for slow and quick load increment, respectively. The fitting parameter values for both the test results (i.e. SM and QM) were estimated as 3.5 and plotted in Figure 2 along with the results from the previous studies. The results in Figure 2 show that equal to 1 can be used for coarse-grained soils regardless of the type of sand (i.e. Unimin sand, Lund sand, and Sollerod sand). The good agreement between the measured and the predicted bearing capacity of unsaturated coarsegrained soils can be attributed to the fact that both the pore-air and the pore-water are under drained conditions during loading stages and the failure mode was following the GSF criteria. In other words, drainage conditions and failure mode in unsaturated coarsegrained soils was consistent with the Terzaghi (1943) bearing capacity theory (i.e. ESA). The parameter was found to be increasing nonlinearly with increasing IP from the study by Vanapalli & Mohamed (2007). However, two data sets plotted in the present study show that values are constant beyond an IP value of 8. The different behavioral trend in the fitting parameter, can be explained offering the following two reasons. Firstly,
8 7
= -0.0031(I P)2+0.3988(I P)+1
Ultimate bearing capacity (kPa)
10
20
30
40
50
60
70
Figure 3. Comparison between the measured and the estimated bearing capacity values using the effective stress approach.
Fitting parameter,
the two hollow square points () in Figure 2 were obtained using the result of only one data point, which means values may not represent the variation of bearing capacity over a wide range of suction. Secondly, the values obtained from both the previous (Vanapalli & Mohamed 2007) and present study may be considered to be reasonable as the parameter may also be influenced by the drainage condition (or rate of loading). In other words, it is likely that different rate of loading can lead to different values of even at the same IP value. Hence, more studies are necessary to investigate the effect of rate of loading on the fitting parameter, for unsaturated fine-grained soils. 2.2 Bearing capacity of unsaturated soils extending total stress approach
6 5 4 3 2 1 0 0 5
Vanapalli & Mohamed 2007
Present study (Costa et al. 2003) Present study (Rojas et al. 2007) 3 different sands
10
15
20
Plasticity index, Ip
Figure 2. Relationship between the plasticity index, IP and the fitting parameter, .
Figure 4 shows loading of a footing located on an unsaturated fine-grained soil. Typically, the slip surfaces below footings are not extended to the ground surface but instead restrict to vertical planes when a load is applied. This characteristic behavior indicates that the bearing capacity of the unsaturated fine-grained soils is governed by the compressibility of the soil block, A-A-B-B below a footing (Figure 4). In this case, the soil around the soil block acts as confining pressure and the
1225
Load
1000
Fitting parameter,
A Footing
A'
100
8e 08 2.1 9988 . = 2 =0 R
0.0
90
) 3(I P
10 =9
B'
Figure 4. Failure mechanism in unsaturated finegrained soils below a footing (Oh & Vanapalli 2009).
10
20
30
40
50
60
Plasticity index, IP
bearing capacity of an unsaturated fine-grained soil can be represented as a function of compressive strength of the soil block. Vanapalli et al. (2007) extended this concept and proposed a model to estimate the bearing capacity of unsaturated fine-grained soils using the equation below: qult(unsat) cu(unsat) NCWCW (6) where qult(unsat) ultimate bearing capacity for an unsaturated soil, cu(unsat) shear strength from unconfined compressive strength for an unsaturated soil, NCW bearing capacity factor with respect to constant water content condition; and CW shape factor with respect to constant water content condition. The form of Eq. (6) is the same as Skempton (1948) equation that is used for interpreting the bearing capacity of saturated soils under undrained loading conditions. The subscript, CW in Eq. (6) is used to show that typically the pore-air is under drained condition and the pore-water is under undrained condition during loading stages in unsaturated finegrained soils (i.e. constant water content condition). Vanapalli et al. (2007) carried out model footing (B L 50 50 mm) tests in unsaturated finegrained soils for five different suction values to study the validity of Eq. (6). Based on the comparison between the measured and the predicted bearing capacity values, they suggested that the bearing capacity factor, Nc proposed by Skempton (1948) and the shape factor, c [ 1 0.2(B/L)] proposed by Meyerhof (1963) and Vesi (1973) for saturated fine-grained soils under u 0 condition can also be used for unsaturated fine-grained soils instead of NCW and CW (Eq. (7)). B qult ( unsat ) 5.14 cu ( unsat ) 1 0.2 L
Figure 5. Relationship between the fitting parameter and the plasticity index, IP (Oh & Vanapalli 2009).
soils. Based on this concept, Oh & Vanapalli (2009) proposed a model to estimate the variation of unconfined compressive strength with respect to suction using the unconfined compressive strength under saturated condition and the SWCC as given below: (u uw ) (S ) / cu ( unsat ) cu ( sat ) 1 a ( Pa / 100 ) (8)
where cu(sat), cu(unsat) unconfined compressive strength under saturated and unsaturated condition, respectively, S degree of saturation; and , fitting parameters. Oh & Vanapalli (2009) analyzed six sets of unconfined compression tests results for unsaturated fine-grained soils and proposed a relationship between the fitting parameter, and plasticity index, IP (Eq.(9); Figure 5).
9 2.1088 e 0.0903( I P )
for 8.0 I P 15.5 for 15.5 I P 60.0
(9)
In addition to , the fitting parameter, equal to 1 for coarse-grained soils and 2 for fine-grained soils are required. 3 ANALYSIS OF THE BEARING CAPACITY TESTS RESULTS
(7)
Eq. (7) indicates that the bearing capacity of unsaturated fine-grained soils can be estimated extending the TSA (i.e. Skempton (1948) bearing capacity theory) using only unconfined compressive strength values for unsaturated fine-grained
In the present study, the model footing and in-situ plate load tests results available in the literature are reanalyzed using three different approaches; i) the effective stress (ESA; Eq. (4)), ii) the total stress (TSA; Eq. (7)), and iii) the reduction factor approach. (Eqs. (2) and (3) along with Eq. (4)). Figure 6 shows the SWCCs for the three soils used in the analyses. The in-situ plate load tests data by Costa et al. (2003) are analyzed and shown in Figure 7. Two different fitting parameter values; namely 3.5 (present study) and 4.39 (Eq. (5), Figure 2) were
1226
100
Table 1. Fitting parameters and their values used in the analyses. ESA IP
Costa et al. 2003 Rojas et al. 2007 Vanapalli et al. 2007
80 60 40 20 0 0.01
TSA 2 2 2 9 9 9
3.5, 4.0 3.5, 5.34 3.5, 6.44
Costa et al. (2003) Rojas et al. (2007) Vanapall et al. (2007)

100 1000 10000
1000 800 600 400 200 0
8 12 15.5
0.1
10
Suction, (ua - uw) (kPa)
Figure 6. Soil-water characteristic curves used for the present study.

Measured Predicted (ESA, Eq. (4), = 3.5) Predicted (ESA, Eq. (4), = 4.0) Predicted (punching failure, = 3.5) Predicted (TSA, Eq. (7)) Costa et al. 2003 Dia. = 800 (mm) IP = 8 (%)
Measured Predicted (ESA, Eq. (4), = 3.5 Predicted (ESA, Eq. (4), = 5.34 Predicted (punching failure, = 3.5) Predicted (TSA, Eq. (7)) Rojas et al. 2007 Dia. = 310 (mm) IP = 12 (%)
150
100
10
20
30
40
50
60
70

50
Figure 8. Comparison between the measured and the predicted bearing capacity values using different approaches (Rojas et al. 2007).
0 5 10 15 20 25 30 35
Measured Predicted (ESA, Eq. (4), = 3.5) Predicted (ESA, Eq. (4), = 6.44 Predicted (punching failure, = 3.5) Predicted (TSA, Eq. (7))
Figure 7. Comparison between the measured and the predicted bearing capacity values using different approaches (Costa et al. 2003).
500 400 300 200 100 0
used to estimate the bearing capacity values extending the ESA (i.e. Eq. (4)). The Unconfined Compressive Strength (hereafter referred to as UCS) for each suction value required for estimating the bearing capacity values using the TSA was obtained using Eq. (8) and the SWCC (Figure 6). The UCS under saturated condition (i.e. cu(sat)) was back-calculated from the plate load tests results using Skempton (1948) equation since the information was not available in the literature (Costa et al. 2003). The fitting parameters, equal to 2 (i.e. finegrained soil) and equal to 9 (Eq. (9)) were used in the analysis. The parameters and their values used in the analyses are summarized in Table 1. The analysis of results summarized in Figure 7 indicate that the bearing capacity values estimated extending the ESA with the fitting parameter, equal to 3.5 (present study) and the TSA can provide reasonably good agreement when compared with the measured values. However, the bearing capacity values estimated using equal to 4 (Vanapalli & Mohamed 2007) and reduction factors approach underestimated the bearing capacity values. Figures 8 and 9 show the comparisons between the measured and the predicted bearing capacity
Vanapalli et al. 2007 BxL = 50 x 50 (mm) IP = 15.5 (%)
50
100
150
200
250
Figure 9. Comparison between the measured and the predicted bearing capacity values using different approaches (Vanapalli et al. 2007).
values for the data by Rojas et al. (2007) and Vanapalli et al. (2007), respectively. The saturated UCS, cu(sat) for analyzing the data by Rojas et al. (2007) was also back-calculated using Skempton (1948) equation and estimated to be equal to 20 kPa. This UCS is approximately the same as the value obtained using cone penetration tests for the Lean clay close to the ground surface under the saturated condition (Rollins et al. 2009). In both cases, the bearing capacity values predicted using the TSA (i.e. Eq. (7)) provided most reasonable comparisons, while those obtained extending reduction factor approach were significantly underestimated.
1227
SUMMARY AND DISCUSSION
Vanapalli & Mohamed (2007) proposed a model to predict the variation of bearing capacity of unsaturated soils extending Terzaghi (1943) bearing capacity theory (i.e. Effective Stress Approach; ESA). The model requires shear strength parameters under saturated condition (i.e. c, ), SWCC, and the fitting parameter, . The comparison between the measured and the predicted bearing capacity values using the proposed method showed that different values of are required for fine-grained soils depending on IP values, whereas constant value (i.e. 1) can be used for all unsaturated coarse-grained soils (i.e. soils with IP 0) . The study presented in this paper shows that the parameter, can be influenced by the drainage condition for unsaturated fine-grained soils. In other words, the values of for the same unsaturated fine-grained soils can be different depending on the drainage conditions. This behavior can be attributed to the fact that, achieving fully drained conditions along with the GSF criteria is less likely for unsaturated fine-grained soils. Therefore, it may not be reasonable to estimate the bearing capacity of unsaturated soils extending the ESA. More studies on the bearing capacity of unsaturated fine-grained soils are necessary under drained loading conditions (i.e. using significantly slow rate of loading or strain rate) to check if unique fitting parameter, as a function of plasticity index, IP is possible. The bearing capacity results analyzed using the ESA were also interpreted extending Skempton (1948) bearing capacity theory (i.e. Total Stress Approach; TSA) using the model proposed by Vanapalli et al. (2007). The proposed model similar in form as Skempton (1948) equation and requires only unconfined compressive strength results of unsaturated fine-grained soils. The comparison between the measured and the predicted bearing capacity values using the proposed model (i.e. TSA) provided good agreement. This can be attributed to the fact that the drainage conditions (i.e. constant water content condition) for both unconfined compression and in-situ plate load (or model footing) tests in addition to using identical strain rates. The studies presented and discussed in this paper support extending the TSA rather than the ESA in the interpretation of the bearing capacity of unsaturated fine-grained soils. REFERENCES
Costa, Y.D., Cintra, J.C. & Zornberg, J.C. 2003. Influence of matric suction on the results of plate load tests performed on a lateritic soil deposit. Geotechnical Testing journal 26(2): 219226.
Gan, J.K-M., Fredlund, D.G. & Rahardjo, H. 1988. Determination of the shear strength parameters of an unsaturated soil using the direct shear test. Canadian Geotechnical Journal 25: 500510. Kumbhokjar, A.S. 1993. Numerical evaluation of Terzaghis N. Journal of Geotechnical Engineering 119(3): 598 607. Meyerhof, G.G. 1963. Some recent research on the bearing capacity of foundations. Canadian Geotechnical Journal 1(1): 1626. Mohamed, F.M.O. & Vanapalli, S.K. 2006. Laboratory investigations for the measurement of the bearing capacity of an unsaturated coarse-grained soil. Proc. 59th Canadian Geotechnical Conference, 14 October 2006, Vancouver, BC. Oh, W.T. & Vanapalli, S.K. 2009. A simple method to estimate the bearing capacity of unsaturated soils. Proc. 62nd Canadian Geotechnical Conference, 2024 October, Halifax: 234241. Oh, W.T., Vanapalli, S.K. & Puppala, A.J. 2009. Semiempirical model for the prediction of modulus of elasticity for unsaturated soils. Canadian Geotechnical Journal 46(8): 903914. Oloo, S.Y. 1994. A bearing capacity approach to the design of low-volume traffics roads. PhD thesis, University of Saskatchewan, Saskatoon, Canada. Oloo, S.Y., Fredlund, D.G. & Gan, J.K-M. 1997. Bearing capacity of unpaved roads. Canadian Geotechnical Journal 34(3): 398407. Rojas, J.C., Salinas, L.M. & Seja, C. 2007. Plate-load tests on an unsaturated lean clay. Experimental Unsaturated Soil Mechanics, Springer Proceedings in Physics 112: 445452. Rollins, K.M., Adsero, M.E. & Brown, D.A. 2009. Jet grouting to increase lateral resistance of pile group in soft caly. In M.H. Hussein, M.G. Iskander. & D.F. Laefer (ed.), Contemporary Topics in Ground Modification, Problem Soils, and Geo-Support; Proc. Int. Foundations Congress and Equipment Expo, ASCE, Orlando, 1519 March, 2009: 265271. Schnaid, F., Consoli, N.C., Cumdani, R. & Milititsky, J. 1995. Load-settlement response of shallow foundations in structured unsaturated soils. Proc. 1st International Conference on Unsaturated Soils, 68 September, Paris: 9991004. Skempton, A.W. 1948. The u 0 analysis for stability and its theoretical basis. Proc. 2nd International Conference of Soil Mechanics and Foundation Engineering, 1: 7277. Steensen-Bach, J.O., Foged, N. & Steenfelt. J.S. 1987. Capillary induced stresses Fact or fiction? Proc. 9th European conference on soil mechanics and foundation engineering, Budapest, Hungary: 8389. Terzaghi, K. 1943. Theoretical Soil Mechanics. John Wiley & Sons, New York, NY, USA. Vanapalli, S.K. & Mohamed, F.M.O. 2007. Bearing capacity of model footings in unsaturated soils. Proc. Intern. Conf. From Experimental Evidence towards Numerical Modeling of Unsaturated Soil, Weimar, Germany, Unsaturated Soils: Experimental Studies: 483493. Vanapalli, S.K., Oh, W.T. & Puppala, A.J. 2007. Determination of the bearing capacity of unsaturated soils under undrained loading conditions. Proc. 60th Canadian Geotechnical Conference, 2124 October, Ottawa: 10021009. Vesi, A.B. 1973. Analysis of ultimate loads of shallow foundations. Journal of the Soil Mechanics and Foundation Division, ASCE 99(SM1): 4573.
1228
Three dimensional numerical simulation of a residential building on expansive soil subject to a leaking underground
X. Zhang
Department of Civil & Environmental Engineering, University of Alaska Fairbanks, Fairbanks, Alaska, USA
ABSTRACT: At present most researchers use a two-step superposition method to simulate the foundation on shrink-swell soils. In the first step, the differential movements of the ground soil surfaces with an ideally flexible weightless slab are calculated. In the second step, soil-structure interaction is then simulated when the foundation is imposed on the surface which has been distorted by the moisture changes. This approach can cause unrealistic results due to the simplified assumptions. In this paper, a three dimensional non-linear fully coupled soil-structure interaction system is used to simulate the effects of leaking underground water pipe on the behavior of a residential building on expansive soils. The soil, structure, and the soil-structure interaction are simultaneously simulated in one single program without above assumptions. A leaking pipe is located near the center of the house where a bathroom is located. The simulation results are discussed. 1 INTRODUCTION Bulut 2001; Vu and Fredlund 2003; Wray 2004, 2005). In the first step, the differential movements of the soil are simulated without considering the load transferred from the slab. In the second step, a slab is forced to be fully contact with the distorted ground soil surface in order to simulate the soil structure interaction. The following limitations exist in the current research are: most of the simulations are one or two dimensional; coupling effects between water flow and deformation are neglected; influence of weather is not accurately considered, constant material properties parameters are used while it is well-known that the unsaturated soil behaviour is notoriously highly nonlinear. In the second step, usually Soil is considered as solid with no water; Winkler or elastic half space foundations are used; no separation or frictional behaviour is allowed; only one or two factors are considered simultaneously. The most significant limitation is that the soil and foundation have never been investigated in a unified system. Zhang (2004) developed a complete system for three dimensional simulations of the behaviour of residential buildings on expansive soils. This study utilized simple and readily available historic weather data such as daily temperature, solar radiation, relative humidity, wind speed, and rainfall as the input. Accurate three dimensional predictions are obtained by integrating a number of different analytical and numerical techniques: different simulation methods for different boundary conditions such as tree, grass, and bare soils, coupled hydro-mechanical stress analysis to describe deformation of saturated-unsaturated soils, jointed elements simulation of soil-structure
Expansive soils are extensively found all over the world. Every year the loss caused by expansive soils is huge. Jones and Holtz (1973) estimated that the annual cost of expansive soil damage in the U.S. is $2.2 billion, which exceeds that caused by earthquakes, hurricanes, and flood combined in an average year. Krohn and Slosson (1980) estimated that the annual cost of expansive soil damage in the US to be $7 billion in 1980. Krohn and Slosson further estimated that damages to single-family and commercial buildings accounted for nearly onethird of the total amount of damage resulting from expansive soils. A damage survey conducted solely in Dallas County, Texas, identified 8,470 residential foundation failures in only one year, 98% of which occurred in expansive soils (Wray 1995). Slab foundations are commonly used to withstand possible damages to residential building built on expansive soils caused by seasonal variations in soil moisture conditions. The current design methods for a slab-on-ground resting on expansive soils were developed for climate-controlled situations (El-Garhy and Wray 2006). However, in addition to seasonal climate variation, there are many other environmental factors resulting in moisture variations and subsequent volume change of expansive soils. For example, leaking underground water facility is often blamed for soil heaving beneath the slab foundation and associated distress of the building superstructure. At present, most researchers used a two-step approach to simulate the behaviour of foundation on expansive soils (Wray 1978; Gay 1993; PTI 1996;
1229
interaction, analysis of structure stress by general shell elements, and assessment of structure damage by the smeared cracking model. The system overcomes the limitations in the existing methods and can be used to simulate the behavior of residential building on expansive soils in a single program. This paper presents some preliminary results for a three dimensional numerical simulation of a residential building on expansive soil subject to a leaking underground. Detailed descriptions for the proposed system can be found in Zhang (2004) and Zhang and Briaud (2006). 2 THREE DIMENSIONAL NUMERICAL SIMULATION OF A RESIDENTIAL BUILDING ON EXPANSIVE SOIL SUBJECT TO A LEAKING UNDERGROUND Model used in the simulation
2.1
Fig. 1a shows the simulated domain with a residential building. Fig. 1b shows layout of the residential building. In Figure 1b, RM1 through RM4 represents four different bedrooms. D1 through D8 represents doors for different rooms with heights of 2.0 m. D1, D2, D3 and D7 have widths of 2.0 m while D4, D5, D6, and D8 have widths of 1.0 m. WD1 through WD10 represents windows for different rooms. All of them have the same size of 2.0 m (width) by
1.5 m (height). The foundation of the residential building consists of three 8 m 8 m 0.4 m slabs poured together. This EI values of the slabs were approximately 8.5 105 kN m2, equivalent to a slab on grade with stiffening beam in both directions, a slab thickness of 0.10 m (4 inches), the depth of the stiffening beam of 0.75 m (30 inches), the beam width of 0.305 m (12 inches), and the beam spacing of 4.572 m (15 feet). This simplified building was used to illustrate application of the proposed system. More complicated and realistic structure can be used such as a building with beams and roof. Both the slabs and walls were assumed to be made of plain concrete. The mass density for concrete was taken as 2.4 103 kg/m3 and the load applied to the structure (including slabs and walls) are self-weight simulated using gravimetric body force. As a result, the applied loads depend upon the geometry of the structure and were not uniform. The Youngs Modulus of the concrete footing was assumed to be E 2 107 kPa. Considering that the strain in the slabs and the walls were small while the rotation was large, the section Poissons ratio was defined as zero 0.0 (ABAQUS/Standard 2002). Shell elements were used for the slab and walls and the shell thickness remained constant. The soil domain has dimensions of 48 m 48 m 6 m and includes the same two soil layers as those in a site at Arlington, Texas. The first layer was a dark gray silty clay layer with a thickness of 1.8 m with a grass root zone in the top 0.47 m. The second layer was a brown silty clay layer with a thickness of 4.2 m. All the soil properties are function of both mechanical stress and matric suction. Detailed material information can be found in Zhang (2004). The soil surface surrounding the concrete slab was covered by Johnsongrass (Sorghum halepense). The FAO-56 PM method was used to determine the evapotranspiration at the ground soil surface. The initial conditions of the soil domain was obtained by running the transient coupled hydromechanical analysis with a time step of one day/ simulation for a period of 10 years using the weather conditions at the Arlington site (Texas) between August 1, 1999 and July 31, 2001. It is assumed that there is a leaking pipe right below the slab of RM3 and caused suction of a 4 m 4 m area as indicated by the shaded area in Figure 1b to be 10 kPa for 100 days. 3 PRESENTATION OF NUMERICAL SIMULATION RESULTS AND DISCUSSION
Figure 1. Simulated domain. (a) 3D view, (b) Layout of the residential building.
Figure 2 shows the suction distribution before and after the leaking. Figure 2a shows the suction of the simulated domain with the residential building
1230
being removed before the leaking occurrence. Figure 2b shows the suction profile at the same time (half of the soil domain is removed to show the suction profile in the vertical direction). It can be seen that before leakage, the suction underneath the slab is lower than that outside of the footing. The suction reduction below the foundation is similar to an island (Figure 2b). The soil below the center of the foundation has the lowest suction, which is due to the fact that the evapotranspiration was interrupted by the coverage of the foundation. The suction increased gradually towards the edge of the slabs where the influence of the climatic condition becomes dominant. Figures 2c and 2d show the
suction distribution at 100 days after the leakage. It can be seen that the suction in RM3 was significantly reduced (Figure 2c) and the suction reduction island enlarged (Figure 2d) due to the leakage. Comparing Figures 2a and 2b, it can also be seen that there is clear soil heaving in the leakage area. Figures 3a and 3b show the distributions of contact pressure between the soil and the slab before and 100 days after the leaking, respectively. As shown in Figure 3a, before the leakage, the contact pressure between the soil and slabs are not uniform, which is attributed to the non-uniform distribution of suction as shown in Figures 2a and 2b. Such a distribution is considered to be reasonable since construction of a residential building on the ground destroy the original equilibrium formed between the climate and the ground. The coverage provided by the foundation interrupts the pre-existing evapotranspiration process and causes moisture accumulation and subsequent heave below the foundation. The maximum heave is in the center where the influence of the climate (mainly evapotranspiration) is the smallest. Accordingly, the maximum contact pressure is in the center of the slab, which is 27.5 kPa. The contact pressure close to the edge is much smaller due to the drier soils. At the corner close to WD 6, the contact pressure is zero, indicating that the soil does not provide any support to the slab.
Figure 2. Suction distribution. (a) Initial Condition before leakage (View from the Top with the Building Removed), (b) Initial Condition before leakage (Side Profile with Half of the Soil Domain Removed) (c) 100 days after Leakage (View from the Top with the Building Removed), (d) 100 days after Leakage (Side Profile with Half of the Soil Domain Removed).
Figure 3. Contact pressure between the soil and the slab (View from the Top with the Building Removed). (a) Initial Condition before leakage, (b) 100 days after Leakage.
1231
Figure 3b shows the distributions of contact pressure between the soil and the slab after the leaking. Due to the soil heaving, the high contact pressure area become smaller, but the maximum contact pressure increased to 30.6 kPa to maintain the force equilibrium in the vertical direction. Figures 4a and 4b show openings between the soil and the slab before and 100 days after the leaking, respectively. As can be seen in Figure 4a, before the leakage, the soil and the slabs were basically in good contact, which is represented by the blue color. At the corner of RM4 close to WD6, there is a maximum separation between the soil and the slab, which is 2.45 cm. At the other corners, the separations are about 1.03 cm. At the double entrance between D1 and D3, the soil and the slab are in good contact. As can be seen in Figure 4b, 100 days after the leakage, the separation at the corner of RM4 close to WD6 increases to 3.13 cm. At the other corners, the separations increase to 1.83 cm. At the double entrance between D1 and D3, the soil and the slab are still in good contact. Due to the leakage and subsequent soil heaving, openings between the soil and the slabs increase.
Figures 5a and 54b show Von Mises stresses in the residential building (soil removed) before and 100 days after the leaking, respectively. Von Mises stress is a stress-invariant used in yield criteria, which is Equal to 1 2
1 2 2 1 3 2 2 3 2
where 1, 2, 3 the major, medium, and minor principal stresses, respectively. It is calculated independently of the coordinate reference system and does not carry directional stress information such as normal and shear stresses, but carries enough information to identify hot-spots where failure might occur. As can be seen in Figure 5a, the Von Mises stress concentrations are always located at places when there are rapid changes in shapes or connections such as corners of the doors and window. Due to the simulated case is more similar to the center lift case, the maximum Von Mises stresses are between D6 and D8 at the top of the walls due to the bending, which is 4.95 103 kPa. As shown in Figure 5b, 100 days after the leakage, the Von Mises stress concentrations in the residential building increase at all places. The maximum Von Mises
Figure 4. Openings between the Soil and the Slab (View from the Top with the Building Removed). (a) Initial Condition before Leakage, (b) 100 days after Leakage.
Figure 5. Von Mises stresses in the Rresidential Bbuilding (View from the tTop with soil removed). (a) Initial Condition before Leakage, (b) 100 days after Leakage.
1232
stresses increase to 5.238 103 kPa, 6% higher than that without leakage. It is worth noting that the assumed leakage does not represent the most severe condition according to the method to apply leakage: only the surface nodes are assumed to be at a suction of 10 kPa. Because the simulated soils are highly expansive soils with very low permeability, the resulted soil heaving is small. In real situations, water leakage usually has more pronounced effect on the distress in the superstructure. 4 CONCLUSIONS
In this paper, a newly-proposed complete system is used to simulate the influencing of a leaking underground pipe below a slab foundation. The numerical analysis indicated that the presence of a leaking underground water pipe can cause soil heaving and severe distress in the superstructure. The resulting damage is expected to be larger than that caused by the seasonal climate changes. The paper also indicates that the proposed system can simulate the soil-structure interaction of residential building on expansive soils in a comprehensive way. REFERENCES
ABAQUS/Standard Users Manual. 2002. Vol. I, II, and III, Version 6.3, Hibbit, Karlsson and Sorenson Inc., 1080 Main Street, Pawtucket, RI02860-4847, U.S.A. Building Research Advisory Board (BRAB). 1968. Criteria for selection and design of residential slabson-ground, National Academy of Sciences, Building Research Advisory Board. Bulut, R. 2001. Finite element method analysis of slabs on elastic half space shrink-swell soil foundations. Ph.D. Dissertation, Department of Civil Engineering, Texas A&M University, College Station, TX. El-Garhy, B.M. & Wray, W.K. 2006. Investigation of a leaking water line beneath residential foundation on expansive soils. 4th International Conference on Unsaturated Soils, April 2006, Carefree, AZ. Geotechnical Special Publication NO. 147, Volume 1, 367378. Fredlund, D.G. & Rahardjo, H. (1993). Soil mechanics for unsaturated soils. John Wiley and Sons, New York. Fredlund, M.D., Stianson, J.R., Fredlund, D.G., Vu, H. & Thode, R.C. (2006). Numerical Modeling of Slab-OnGrade Foundations. 4th International Conference on Unsaturated Soils, April 2006, Carefree, AZ. Geotechnical Special Publication NO. 147, Volume II, 21202132. Gay, D.A. (1993). Development of a predictive model for pavement roughness on shrink-swell clay. Ph.D. Dissertation, Department of Civil Engineering, Texas A&M University, College Station, TX. Grove, A.T. 1980. Geomorphic evolution of the Sahara and the Nile. In M.A.J. Williams & H. Faure (eds), The Sahara and the Nile: 2135. Rotterdam: Balkema.
Jappelli, R. & Marconi, N. 1997. Recommendations and prejudices in the realm of foundation engineering in Italy: A historical review. In Carlo Viggiani (ed.), Geotechnical engineering for the preservation of monuments and historical sites; Proc. intern. symp., Napoli, 34 October 1996. Rotterdam: Balkema. Johnson, H.L. 1965. Artistic development in autistic children. Child Development 65(1): 1316. Jones, D.E. & Jones, K.A. 1987. Treating expansive soils. Civil Engineering, ASCE, New York, 57(8), 6265. Krohn, J.P. & Slosson, J.E. 1980. Assessment of shrinkswell soils in the United States. Proc. Fourth International Conference on Shrink-swell Soils, Denver, Colorado, 1, 596608. Li, J. 2006. Two dimensional simulation of a stiffened slab on expansive soil subject to a leaking underground water pipe. 4th International Conference on Unsaturated Soils, April 2006, Carefree, AZ. Geotechnical Special Publication NO. 147, Volume II, Pp. 20982120. Post-Tensioning Institute (PTI). 1996. Design and construction of post-tensioned slabs-on-ground, 2nd Ed., Phoenix, Arizona. Post-Tensioning Institute (PTI). 2004. Design and construction of post-tensioned slabs-on-ground, 3rd Ed., Phoenix, Arizona. Standards Association of Australia, AS 2870 Standard House 1996. Residential slabs and footings, Sydney NSW, Australia. Wire Reinforcement Institute 1996. Design of slab-onground foundations. Technical Report, Wire Reinforcement Institute, Hartford, CT 06103. Wray, W.K. 1995. So your home is built on expansive soils: a discussion of how expansive soils affect buildings. American Society of Civil Engineers, New York. Wray, W.K. 1997. Using soil suction to estimate differential soil shrink or heave. Unsaturated Soil Engineering Practice, Geotechnical Special Publication NO. 68, 88112. Wray, W.K., EI-Garhy, B.M. & Youssef, A.A. (2005). Three-dimensional model for moisture and volume changes prediction in shrink-swell soils. Journal of Geotechnical and Geoenvironmental Engineering, ASCE 131(3), 311324. Zhang, X. 2004. Consolidation theories for saturatedunsaturated soils and numerical simulations of residential buildings on shrink-swell soils. Ph.D. Dissertation, Department of Civil Engineering, Texas A&M University, College Station, TX. Zhang, X. & Briaud, J.L. 2006. Coupled Hydro-Mechanical Stress Soil-Structure Interaction of Foundation on Saturated and Unsaturated Soils. 4th International Conference on Unsaturated Soils, April 2006, Carefree, AZ. Geotechnical Special Publication NO. 147, 2, 21382149. Zhang, X. & Briaud, J.L. 2008. Improved approach to construct constitutive surfaces for stable-structured soils covering both saturated and unsaturated conditions. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 134(6), 876882. Zhang, X., Lytton, R.L. & Briaud, J.L. 2005. Coupled consolidation theory for saturated-unsaturated soils. Biot 2005 Centennial Conference, University of Oklahoma, Norman.
1233
Slopes
Stochastic modelling of unsaturated slope stability

Patrick Arnold
School of Mechanical, Aerospace and Civil Engineering, The University of Manchester, Manchester, UK
Michael A. Hicks
Faculty of Civil Engineering and Geosciences, Delft University of Technology, Delft, The Netherlands
ABSTRACT: This paper presents a stochastic study of slope stability in unsaturated soils for steady state flow conditions. In order to account for material non-linearity and the spatial variability of soil properties, an elasto-viscoplastic finite element program has been combined with random field theory within a Monte Carlo framework. The results of multiple realisations have been evaluated in terms of reliability. Focusing on the influence of spatial variability of matric suction, the results highlight the importance of spatial correlation length and degree of anisotropy in the assessment of unsaturated slope stability. 1 INTRODUCTION of spatial variability of matric suction has been investigated. The paper uses similar methodology to previous investigations on saturated slopes. For example, undrained clay slopes have been analysed by Paice and Griffiths (1997), Hicks and Samy (2002a, 2002b, 2002c, 2004), Griffiths and Fenton (2004), Spencer and Hicks (2007) and Hicks et al. (2008), whereas liquefaction studies for sand slopes have been conducted by Hicks and Onisiphorou (2005). Meanwhile, drained slope studies have been carried out by Griffiths et al. (2007, 2009). 2 CONSTITUTIVE MODEL FORMULATION
Analysing the stability of soil slopes is one of the oldest tasks in geotechnical engineering and there are numerous approaches available. But how close to real behaviour are the designs of new slopes and the assessments of existing slopes? There is clearly a degree of uncertainty associated with every geotechnical design. Sources of uncertainty may include slope geometry, quality of soil data, complexity of the material model, and so on. However, this paper focuses on the uncertainty due tothe inherent spatial variability of soils. In general, soil property data from in situ or laboratory testing may be limited, so that a degree of uncertainty exists about what is in the ground. In engineering practice, this uncertainty leads to the use of cautious estimates of soil properties for characterising soil deposits. However, approximate probabilistic methods such as the first-order reliability method, combined with limit equilibrium or finite element analysis, are gaining increasing popularity as an objective means to quantifying the effect of the uncertainty. The main drawback of these simpler approaches, however, is that the spatial nature of the soil variability is either accounted for in a simplistic way or possibly not at all. In nature, soil deposits are generally laid down in distinct geological zones or layers, with weak layers being possible areas of shear failure. However, failure may also propagate through local weak zones within so-called uniform deposits. In this paper, the stability of an unsaturated soil slope is analysed, accounting for the spatial variability of material properties. In particular, the influence
Numerous models are available for describing the mechanical and hydraulic behaviour of unsaturated soils, including some advanced formulations. However, the main purpose of this study is to provide a preliminary investigation into the influence of spatialvariability of soil properties on unsaturated slope failure. Hence a rather simple formulation has been applied. 2.1 Shear strength The mechanical constitutive formulation is based on Bishops effective stress and a linear elastic, perfectly plastic Mohr-Coulomb soil model. The failure criterion extended for unsaturated conditions is f c ( ua) tan (ua uw) tan (1)
1237
where f is the soil shear strength, c is the effective cohesion, is the effective angle of friction, is the total stress, ua is the pore air pressure, uw is the pore water pressure, ( ua) is the net stress and (ua uw) is the matric suction. The effective stress parameter can be expressed as N (Vanapalli et al. 1996; Fredlund et al. 1996), where N is the normalised water content (which is equal to the degree of saturation S), and is a fitting parameter. 2.2 Water retention curve
Table 1. E v c ksat n
Model parameters. Soil unit weight Youngs modulus Poissons ratio Dilation angle Effective friction angle Effective cohesion Fitting parameter Sat. hydr. conductivity Fitting parameter Fitting parameter 20 5 104 0.3 0 25 10 1.0 8 106 0.05 4.0 [kN/m3] [kPa] [] [] [] [kPa] [] [m/s] [kPa1] []
The water retention curve has been modeled using the relationship proposed by van Genuchten (1980),
w r
s r
1 ( s )n
(2)
where w is the actual volumetric water content, s and r represent the volumetric water contents at the saturated and residual conditions respectively, s is the matric suction, and , n and m are fitting parameters. By using the statistical pore-size distribution relationship, m 1 1/n, proposed by Mualem (1976) and by assuming r 0, the above equation simplifies to 1 S n 1 s
1
1 n
instance, during or antecedent to a rainfall event. In the present study, the flow has been idealised as a vertical steady state infiltration process. For a homogeneous soil, this assumption is reasonable and has been applied in many studies. However, for heterogeneous soils the assumption of one-dimensional flow may not be appropriate. This is because water will tend to flow in a rather tortuous manner as it seeks to follow the path of least resistance. Applying Darcys law, the suction profile for a steady state vertical infiltration may be computed by, zi 1 zi
i 1
(3)
d 1 [ q / ki 1( )]
(6)
2.3 Hydraulic conductivity function The hydraulic conductivity k(s) at a given suction level is k(s) ksatkr(s) (4)
where z is the elevation above the ground water table, and k() are the suction head and the hydraulic conductivity at the corresponding elevations, and q is the flow rate. For this preliminary investigation, a constant flow rate q has been assumed for both the homogeneous and heterogeneous cases. 2.5 Summary of material parameters
where the relative coefficient of permeability kr(s) has been modelled using the function proposed by van Genuchten (1980),
1
Table 1 summarises the material parameters and lists typical mean values for a sandy silt. 3 STOCHASTIC MODELLING PROCESS
k ( s ) ksat
( s )n 1 1 ( s )n [1 ( s )n ] is the
( n1) 2n
(1 n ) n

(5) hydraulic
and where ksat conductivity. 2.4
saturated
Flow formulation
The spatial variability of a soil property X is generally represented by an assumed probability distribution and by the point and spatial statistics of X. The point statistics are the mean and standard deviation, X and X respectively, which may be combined to give the coefficient of variation, V
Transient seepage processes can have a significant influence on unsaturated slope stability; for
X X
(7)
1238
The spatial statistics are the scales of fluctuation, , in the coordinate directions. Vanmarcke (1983) defined as the distance over which a soil property is strongly correlated. Hence, a small scale of fluctuation indicates a weakly correlated deposit, whereas a large scale of fluctuation indicates a more homogeneous soil. In this paper the random property fields have been generated using an algorithm based on Local Average Subdivision (Fenton and Vanmarcke 1990). For 2D problems, this involves generating a random field for a square domain through a process of uniform subdivision. The final field comprises square cells, with each cell having a unique value of X that is correlated with neighbouring cells. Spatially varying property distributions for general problems such as slopes are easily managed by mapping the cell values from the random field onto the structure being analysed. Hicks and Samy (2002b, 2004) described the process for generating anisotropic random fields; that is, fields in which the horizontal and vertical scales of fluctuation, h and v, are not equal. In this case, the degree of anisotropy of the heterogeneity is given by h/v. Note that, for a given set of statistics, there are an infinite number of possible random fields. Of course, they will all look similar, but they will differ with respect to the distribution of strong and weak zones and each, when used in a finiteelement analysis, will lead to a different finite element solution. Hence it is necessary to carry out multiple realisations of the problem, as part of a Monte Carlo simulation, to ensure a converged solution. 4 DEFINITION OF RANDOM PROPERTIES
Although it is possible to define all properties as spatially variable, the current investigation has been simplified by considering only the spatial variability of ksat, and n. The effective strength parameters c and have been considered constant in order to focus on the influence of spatial variability of the suction stress. Table 2 lists the distributions and point statistics selected for the analysis. Wierenga et al. (1991), Russo and Bouton (1992), Schaap and Leij (1998) and Zhu et al.
Table 2. Assumed distributions and point statistics. V ksat n 0.9 0.5 0.05 Distribution log-normal log-normal normal
(2007) investigated the influence of variability of the van Genucthen parameters on the flow properties. Schaap and Leij (1998) concluded that, of the threeparameters, n can be established with greatest certainty. Moreover, Zhu and Mohanty (2003) concluded that the impact of the variability in is larger than that of n, and that it may therefore be reasonable to treat n as deterministic. Hills et al. (1992) showed that should be treated as variable, but suggested that reliable results may be obtained by varying either or n and holding the other parameter constant. Zhang (2002) stated that the sensitivity of unsaturated flow to log is much higher than to log ksat. It has been found that ksat is best described using a log-normal distribution (e.g. Russo and Bouton 1992; Fenton and Vanmarcke 1998; Zhu and Mohanty 2003). The variability of is also best treated as log-normal (e.g. Russo and Bouton 1992; Schaap and Leij 1998; Zhu and Mohanty 2003), whereas, for n, both the normal distribution (Russo and Bouton 1992) and the log-normal distribution (Schaap and Leij 1998) have been recommended. Results by (Russo and Bouton 1992) suggested a moderate correlation between ksat and the parameters describing the retention curve, but only a very small correlation between n and . Zhu and Mohanty (2003) pointed out that cross-correlation between ksat and increases the hydraulic conductivity. In this investigation the ksat, n and fields are assumed to be uncorrelated. The scarcity of information available on the in situ variability of soil property data complicates the definition of a correct statistical correlation between the variables, although the implementation of a cross-correlation structure between c, and s(ksat, , n) is under investigation. Since the scale of fluctuation mainly depends on the geological deposition process, the spatial correlation structure will be similar for different material parameters. Hence it is reasonable to use the same correlation length, k n , hereafter sat referred to as . 5 NUMERICAL SIMULATION
Figure 1 shows a 45 slope of height H 5 m, which has been analysed in plane strain using an elastoviscoplastic finite element program based on Smith and Griffiths (2004). The domain has been discretised using 3010 eight-node quadrilateral finite elements, each of dimensions 0.25 0.25 m and using reduced 2 2 Gaussian integration. The slope has been cut from a heterogeneous 10 m deep soil layer and is founded on a firm rock base. For small scales of fluctuation, a high level of discretisation is required to avoid excessive
1239
10m
5m
10m 0.0 ZGW
5m
5m
Figure 2. Figure 1. Problem domain, boundary conditions and finite element discretisation.
Example field of for v 1.0 m and 6.
averaging of material properties. However, computation times may be reduced while maintaining sufficient accuracy, by assigning random field cell values to the element Gauss points (Hicks and Samy 2002b, 2004). Hence, the spatially varying soil profiles have been modelled using 12040 random field cells of size 0.125 0.125 m. The material properties for the soil layer are defined in Tables 1 and 2. The aim of the analysis has been to investigate the influence of the horizontal and vertical scales of fluctuation on the relationship betweenreliability and global factor of safety F, in which F is defined relative to the mean property values. For any realisation based on a given set of statistics, three random fields are generated; these being for ksat, and n. For example, Figure 2 shows a typical random field of for v 1.0 m and 6, in which the darker zones represent lower values of . The three random fields are used to back-figure a random field of suction stress, and it is this field that is mapped onto the problem domain for the finite element analysis. Specifically, for each realisation 200 matric suction profiles are back-figured, one for each random field column mapped onto the problem domain. This has been achieved by the stepwise integration of equation 6, starting from theground water level. The incremental increase or decrease in matric suction is computed, based on the values of q, ksat, and n at the current location in the column. The suction stress (s) profile is computed simultaneously to the matric suction profile using equation 3, whereafter the average is taken over the height of each random field cell. Once the 2D suction stress field has been completed, it is mapped onto the Gaussian integration points of the finite element mesh. Thus the distribution of shear strength is known throughout the mesh via equation 1. Figure 3 shows the random field of suction stress for (a) no flow and (b) steady inflow, based on the distribution of shown in Figure 2. These results are for the horizontal water table level at a depth of 6 m below the slope crest. In these figures,
Figure 3. inflow.
Fields of s for (a) no flow and (b) steady
the darker zones indicate higher suction stresses. The relative increase in suction stresses for the chosen rate of infiltration is clearly seen. For a given set of statistics, multiple realisations of the problem are analysed. For each realisation, gravitational loading is applied to the slope to generate the in situ stresses and the strength reduction method is utilised to determine the point of failure. This method is commonly used for saturated problems, but it has also been applied to unsaturated slopes (e.g. Cai and Ugai 2004; Griffiths and Lu 2005; Li et al. 2006). The critical condition for a particular realisation is obtained by reducing the shear strength gradually until slope failure occurs. For a given factor of safety, the reliability of the slope is given by R 1 Nf N (8)
where N is the total number of realisations and Nf is the number of realisations in which the slope fails. For this investigation 500 realisations where sufficient to give a converged solution. 6 RESULTS
Figures 47 summarise the results for when the ground water level is 6.0 m below the crest of
1240
10 9 8 7 6 5 4 3 2 1 0 0
elevation above GW level [m]
2.1 2 1.9 1.8 1.7 1.5 1.4 10 1.3 0 15 0.5 1 zGW/H 1.5 2
Q = 0.0 Q = -0.07 saturated
1 0.9 0.8 0.7 0.6 0.5

=1 =3 =6 = 24 = 1000 v/H = 0.2 v/H = 0.2 v/H = 0.2 v/H = 0.2 v/H = 0.2
q = 0.0 m/s zGW = -6.0 m
Q = 0.0 Q = -0.07
1.6
s [kPa]
Figure 4. (a) Suction stress profile; (b) Influence of ground water table on factor of safety.
0.4 0.3 0.2 0.1 0 0.8 0.85 0.9 0.95 1 F
1.05
1.1
1.15
1.2
1 0.9 0.8 0.7 0.6 0.5 0.3 0.2 0.1 0 0.8 0.85 0.9 0.95 1 F 1.05
= 1 /H = 0.1 = 1 /H = 0.2 = 1 /H = 0.4 = 1 /H = 1.0
q = 0.0 m/s zGW = -6.0 m
Figure 6. Reliability versus global factor of safety for anisotropic heterogeneity and no flow.
1 0.9 0.8 0.7 0.6 0.5 0.3 0.2 0.1 0 0.8 0.85 0.9 0.95 1 F
=1 =3 =6 = 24 = 1000 v/H = 0.2 v/H = 0.2 v/H = 0.2 v/H = 0.2 v/H = 0.2
0.4
q = -5.610-7m/s zGW = -6.0 m Q
1.1
1.15
1.2
Figure 5. Reliability versus global factor of safety for isotropic heterogeneity and no flow.
0.4
1.05
1.1
1.15
1.2
the slope. Two cases have been considered; no flow and steady state infiltration representing a medium rainfall event of q 48 mm/day. This equates to q 5.6107 m/s and a normalised flow rate based on the mean saturated permeability of Q qksat 007 . Figure 4(a) shows the suction stress profile for the assumed mean material properties and both flow cases, whereas Figure 4(b) shows the effect of water table level on F. For a homogeneous slope and a 6.0 m deep ground water table, the safety factors are 1.68 and 1.77 for the no flow and infiltration cases respectively, compared with F 1.39 for saturated conditions. Figure 5 shows the influence of scale of fluctuation on the relationship between reliability and global factor of safety, for the case of isotropic heterogeneity and no flow. For larger values of /H there is a wider distribution of responses. This is because there is a wider range of possible random fields, due to each field having a more uniform appearance. Conversely, as /H becomes smaller the variability across each field becomes greater; this leads to greater averaging of material properties within each realisation, and a narrower response distribution with a mean tending towards F 1.0 as becomes very small.
Figure 7. Reliability versus global factor of safety for anisotropic heterogeneity and constant infiltration.
Previous studies for saturated slopes have shown that failure mechanisms are attracted to weaker zones. It is therefore interesting to note that, at F 1.0 based on the mean properties, R 50% for most values of /H. This is due to the non-linearity between and s which is implicit in equation 3. Note that, for this analysis, the spatial variability of ksat has no influence since q 0.0. Moreover, the variation in is much greater than that for n. Similar observations may be made for the anisotropic results in Figure 6. Once again, the response distribution becomes wider as the degree of spatial correlation increases, in this case due to an increase in . Figure 7 shows equivalent results for the case of anisotropic heterogeneity and steady state infiltration. In contrast to Figures 5 and 6, R is now less than 50% for F 1.0, as is generally found for saturated slopes. It is thought that this is due to the mean suction stress profile for all realisations being closer to the suction stress profile based on the mean material properties, for the chosen infiltration rate of q 5.55 107 m/s. This is in contrast to the q 0.0 analysis, in which the mean suction
1241
stress profile was significantly more tensile then the profile based on mean properties. 7 CONCLUSIONS
This paper has illustrated a stochastic approach for accounting for uncertainties associated with the spatial variability of material properties in unsaturated soils. So far, only the spatial variability of suction stress has been considered. However, theinfluence of spatial variability has been clearly demonstrated for a simple problem involving both no-flow and steady state infiltration boundary conditions. In particular, the scale of fluctuation and degree of anisotropy of the heterogeneity are seen to be important statistical parameters. The influences of 2D flow, spatial variation of c and , and cross-correlation between parameters are currently under investigation. REFERENCES
Cai, F. and K. Ugai (2004). Numerical analysis of rainfall effects on slope stability. International Journal of Geomechanics 4(2), 6978. Fenton, G.A. and E.H. Vanmarcke (1990). Simulation of random fields via local average subdivision. Journal of Engineering Mechanics, ASCE 116(8), 17331749. Fenton, G.A. and E.H. Vanmarcke (1998). Spatial variation in liquefaction risk. Gotechnique 46(6), 819831. Fredlund, D.G., A. Xing, M.D. Fredlund, and S. L. Barbour (1996). The relationship of the unsaturated soil shear strength to the soil-water characteristic curve. Canadian Geotechnical Journal 33(3), 440448. Griffiths, D.V. and G.A. Fenton (2004). Probabilistic slope stability analysis by finite elements. Journal of Geotechnical and Geoenvironmental Engineering 130(5), 507518. Griffiths, D.V., G.A. Fenton, and M.D. Denavit (2007). Traditional and advanced probabilistic slope stability analysis. In Probabilistic Applications in Geotechnical Engineering (GSP 170), ASCE, pp. 110. Griffiths, D.V., J. Huang, and G. A. Fenton (2009). Influence of spatial variability on slope reliability using 2-D random fields. Journal of Geotechnical and Geoenvironmental Engineering 135(10), 13671378. Griffiths, D.V. and N. Lu (2005). Unsaturated slope stability analysis with steady infiltration or evaporation using elasto-plastic finite elements. International Journal for Numerical and Analytical Methods in Geomechanics 29(3), 249267. Hicks, M.A., J. Chen, and W.A. Spencer (2008). Influence of spatial variability on 3D slope failures. In Proc. of the 6th Int. Conf. in Computer Simulation, Risk Analysis and Hazard Mitigation, Cephalonia, Greece, pp. 335342. Hicks, M.A. and C. Onisiphorou (2005). Stochastic evaluation of static liquefaction in a predominantly dilative sand fill. Gotechnique 55(2), 123133. Hicks, M.A. and K. Samy (2002a). Influence of anisotropic spatial variability on slope reliability. In Proc. of
the 8th Int. Symp. on Numerical Models in Geomechanics, Rome, Italy, pp. 535539. Hicks, M.A. and K. Samy (2002b). Influence of heterogeneity on undrained clay slope stability. Quarterly Journal of Engineering Geology and Hydrogeology 35(1), 4149. Hicks, M.A. and K. Samy (2002c). Reliability-based characteristic values: A stochastic approach to Eurocode 7. Ground Engineering 35(12), 3034. Hicks, M.A. and K. Samy (2004). Stochastic evaluation of heterogeneous slope stability. Italian Geotechnical Journal 38(2), 5466. Hills, R.G., D.B. Hudson, and P.J. Wierenga (1992). Spatial variability at the Las Cruces Trench site. In M.T. van Genuchten, F.J. Leij, and L.J. Lund (Eds.), Indirect Methods for Estimating the Hydraulic Properties of Unsaturated Soils, University of California, Riverside, USA, pp. 529538. Li, R., Y. Yu, L. Deng, and G. Li (2006). Stability analysis of unsaturated soil slope by 3-D strength reduction FEM. In Unsaturated Soil, Seepage, and Environmental Geotechnics (GSP 148), ASCE, pp. 6269. Mualem, Y. (1976). A new model for predicting the hydraulic conductivity of unsaturated porous media. Water Resources Research 12(3), 513522. Paice, G.M. and D.V. Griffiths (1997). Reliability of an undrained clay slope formed from spatially random soil. In Proc. of the 9th Int. Conf. on Computer Methods and Advances in Geomechanics, Volume 1, Wuhan, China, pp. 543548. Russo, D. and M. Bouton (1992). Statistical analysis of spatial variability in unsaturated flow parameters. Water Resources Research 28(7), 19111925. Schaap, M.G. and F.J. Leij (1998). Database-related accuracy and uncertainty of pedotransfer functions. Soil Science Society of America Journal 163(10), 765779. Smith, I.M. and D.V. Griffiths (2004). Programming the Finite Element Method. Chichester: John Wiley & Sons. Spencer, W.A. and M.A. Hicks (2007). A 3D finite element study of slope reliability. In Proc. of the 10th Int. Symp. on Numerical Models in Geomechanics, Rhodos, Greece, pp. 539543. van Genuchten, M.T. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal 44, 892898. Vanapalli, S.K., D.G. Fredlund, D.E. Pufahl, and A.W. Clifton (1996). Model for the prediction of shear strength with respect to soil suction. Canadian Geotechnical Journal 33(3), 379392. Vanmarcke, E.H. (1983). Random Fields: Analysis and Synthesis. Cambridge, Massachusetts: MIT Press. Wierenga, P.J., R.G. Hills, and D.B. Hudson (1991). The Las Cruces Trench site: Characterization, experimental results, and one-dimensional flow predictions. Water Resources Research 27(10), 26952705. Zhang, D. (2002). Stochastic Methods for Flow in Porous Media. San Diego: Academic Press. Zhu, J. and B.P. Mohanty (2003). Effective hydraulic parameters for steady state vertical flow in heterogeneous soils. Water Resources Research 39(8), 1227, doi:10.1029/2002WR001831, 12.112.12. Zhu, J., M.H. Young, and M.T. van Genuchten (2007). Upscaling schemes and relationships for the Gardner and van Genuchten hydraulic functions for heterogeneous soils. Vados Zone Journal 6(1), 186195.
1242
Influence of seasonable water pressure variations on unsaturated slopes stability

F.A.J. Franch
Franch Engenharia e Consultoria de Fundaes, So Paulo, SP, Brasil Universidade de So Paulo, So Paulo, SP, Brasil
M.M. Futai
Universidade de So Paulo, So Paulo, SP, Brasil
ABSTRACT: Unsaturated slopes fail during rainy periods especially because of the increase in the water pressure in the soil, which causes the reduction in the effective stress on slopes and the reduction in the shear strength of unsaturated soils. A research conducted on an unsaturated slope in So Paulo, SP, Brazil, measured on a daily basis the precipitations, the groundwater level and the water pressure up to 3.0 m depth, during eight months. Laboratory and in situ tests provided information about the characteristics of the soils composing the experimental slope. Analyses of the slope stability were performed for the dates when the maximum and minimum water pressures were recorded during the monitoring period. The results of the monitoring were satisfactory considering that the tensiometers installed in the experimental area presented readings coherent with the precipitations. The results of the stability analysis have shown variations on the landslide factor of safety as an effect of seasonal variations of the water pressure. 1 INTRODUCTION pressures were not measured at the experimental field: a slope where tensiometers were installed to read the water pressure (uw). Considering that the air pressures were not taken into account, equation (1) may be rewritten as follows: c tg (uw) tg b (2)
The concepts of unsaturated soils are already sufficiently grounded, however, few applications or practical studies have used them. This paper presents a case that used the concepts of the soil mechanics at all research stages. According to Fredlund (1979), the behavior of unsaturated soils may be expressed by three combinations of variables of stress state: ( uw), ( ua), (ua uw). The author elected the combination using the variables ( ua) and (ua uw) to represent the shear strength of unsaturated soils, because it allows representing the effects of the normal stress change, disregarding the air pressure (ua), separately from the effects caused by the suction change (ua uw). Equation (1) represents the proposal of Fredlund et al. (1978) for the strength of unsaturated soils: c ( ua ) tg (ua uw ) tg b (1)
where is shear strength; c is effective cohesion; is normal stress; ua is air pressure in the soil; is effective friction angle; uw is water pressure in the soil; and b is the angle that indicates the coefficient of the increased shear strength due to the suction increase. In this research the air pressure (ua) in equation (1) was not measured in the laboratory tests to set the strength parameters. This is due to the practical impossibility of measuring it given the types of tests developed. Likewise, the air
Equation (2) is considered consistent, as it disregards the air pressure to set the parameters for soil strength in the laboratory and in actual field situations. On the other hand, it is known that stress trajectories, whether in field or in the laboratory, are not exactly the same and may generate different air pressures in both situations, resulting in distinct shear strengths. This discrepancy occurs only when the interstitial air is trapped in bubbles, in which case the saturation level is higher than about 85%. Where the saturation levels are lower than that value, the air is free in the soil and its pressure is the atmosphere pressure, i.e., equal to zero, and may be disregarded. On the opposite hand, where the water pressure is positive and higher than a certain value, all the air is dissolved in the water, and equation (2) is also valid. The air pressures in field develop slightly differently from how they develop in the laboratory, from a same increase in the normal stress and same water pressure value; and this difference in behavior contributes to change in a non-significant way the shear strength of the soils.
1243
This paper presents the results of the water pressure and groundwater level monitoring at the experimental field. It also shows stability analyses of the monitored slope, which reveal the effect of seasonable variations of water pressure on the slope stability. 2 EXPERIMENTAL FIELD
Experimental Field
For this study to be carried out, an experimental field was implemented in So Mateus district, eastern region of So Paulo/SP. The area is owned by the Municipal Government of So Paulo and is considered an area of geological/geotechnical risk. Within the past ten years (in 1998 and in 2004), there were two landslides in the slope where the experimental field was implemented. The average precipitation in the region is 1.450 mm per year, slightly higher than in So Paulo downtown, because it is near Serra do Mar. The experimental field was installed in a slope measuring 150 m in length approximately. Within the experimental field, the slope is approximately 14 m high, with 28 degrees of inclination, 16 m wide, and 30 m long. The vegetation that covered the southern portion of the experimental field was cleared and a cement/sand mortar layer was applied to the soil surface whereas the grass in the northern portion was maintained, as Figure 1 shows. The geological/geotechnical profile of the experimental slope shows a red and white clay landfill layer, which is approximately 3.0 m thick that overlaps in part the reddish brown micaceous silty residual soil. There is a clayey layer with vegetal detritus, which is about 20 cm thick, between the clayey and silty horizons, and allows the conclusion that the clayey layer is a landfill, as Figure 2 shows. The instrumentation used for the daily monitoring consisted of a water level meter built at the slope foot, which was 4 m deep and indicated a groundwater level at a 3.7 m depth in January 2005; a pluviometer installed 50 meters away from the experimental field; and 25 tensiometers installed in two sections perpendicular to the level curves, of which one section in the slope portion with vegetation and the other in the portion covered with mortar. The results of the water pressure measured in the mortar-covered section will not be shown in this paper because its focus is the influence of the infiltration into the natural ground. The description of the used tensiometers is presented by Franch et al. (2005). The sections are located at the center of the vegetation-covered section and the mortar-covered section, in five elevations (lines 1 to 5) and in three depths: 0.5 m, 1.5 m, and 3.0 m, as Figure 2 shows. The section with vegetation had 12 tensiometers.
Figure 1.
837 834
Experimental field view.

835.35
tensiometer Elevation (m) h=15,0 831 828 825 820.35 822 819 816 813 -817.45 (07/01/2005) waterlevel meter h=4,0m
clayey embankment
silty residual soil clay with vegetal detritus 30 40 45
10
20 Distance (m)
Figure 2. Geological/geotechnical profile and instrumentation used in the slope section coated with vegetation: tensiometers and water level meter.
MATERIALS AND METHODS
Tests were performed for the characterization and determination of hydraulic and strength parameters using undisturbed soil samples from the clayey landfill and the silty residual soil. The characterization of the soils and the hydraulic parameters are presented in Franch (2008), and are not presented in this paper. Following are the results of the direct shear and unconfined compression tests performed, which will be used in the stability analyses. This study involved strength tests performed in saturated and unsaturated samples of soils to obtain the parameters: c, , and b. The tests were conducted in body proofs with different suctions. In direct shear tests the suctions (initial and final) were measured by using the filtering paper method, and in unconfined compression tests the water pressures were measured by using a high-capacity tensiometer during the tests. 3.1 Clayey landfill The strength parameters obtained for the flooded landfill in direct shear tests were: effective friction
1244
angle of 31 and effective cohesion of 15 kPa. The strength envelopes obtained for unsaturated clay (uw 50, 130 and 200 kPa) were coherent among them resulting in increased cohesions with increased suction, except one air-dried sample with 1% residual humidity, which presented reduced cohesion in relation to the water pressure sample of 200 kPa, as shown in Figure 3a. The b value obtained for these tests was 36, considering a water pressure variation range from 0 to 60 kPa, a range in which landslides are likely to occur. For water pressure values lower than 60 kPa the value obtained for b was 22. The interpretation of the results of unconfined compression in unsaturated samples was based on the effective friction angle and cohesion adjustment so that the envelope will touch the Mhrs circle. The results achieved in the unconfined compression tests with water pressure equal to 37, 81, 161, and 242 kPa were coherent with those achieved in the direct shear tests, resulting in b value equal to 38 for the water pressure variation range from 0 to 60 kPa, and 27 for water pressures lower than 60 kPa. The b values obtained in both types of tests (36 and 38) were unexpectedly higher than the value (31). The literature (Campos, 1997; Futai et al, 2004) shows results of tests with controlled suction that revealed results similar to those presented in this paper. It was expected that b would be equal to , for suction values lower than the air entry value, which in the case of the tested clay is 150 kPa, and b values lower than for suctions greater than the air entry value.
3.2
Residual silty soil
The strength parameters obtained for the flooded residual soil in direct shear tests were: effective friction angle of 35 and effective cohesion of 19 kPa. The strength envelopes obtained for unsaturated silt (uw 80 and 130 kPa) were coherent with each other, resulting in increased cohesions with reduction in water pressures, including one airdried sample. The b value obtained for these tests was 35, considering water pressures between 0 and 60 kPa, and for pressures lower that this value, the b value obtained was 21. The results obtained in unsaturated unconfined compression tests and analyzed in the same way as in the tests performed in the clayey landfill presented increased cohesion with reduction in the tested water pressure (uw 27, 48, 99 e 457 kPa), except the test with water pressure equal to 457 kPa, which presented decreased cohesion in relation to the test with water pressure equal to 99 kPa. The b value obtained for these tests was 11, disregarding the result of the sample with water pressure equal to 457 kPa. For water pressures between 0 and 60 kPa, the b value is quite different from that obtained in the direct shear test (35). For pressures below 60 kPa, the b value obtained was 19. The strength envelopes of unsaturated silt in unconfined compression tests were drawn based on an estimated saturated strength envelope, with the following effective geotechnical parameters: c 0 and 22, defined based on conservative values indicated in the literature (Metro 1980). The effective parameters obtained in flooded direct shear tests, c 19 kPa and 35, could not be used, as explained below.
835.35
Antecedent Precipitation: 3 days - 3 mm 7 days - 38 mm 14 days - 99 mm 28 days - 199 mm
0 0
E F H I M
837
J L
+10 clayey +20 embakment +10 831 Elevation (m)
-10
C
-5 -15
-5 825
G -10
821.15
D A B
-15
-15
817.67
micaceus silty residual soil
LEGEND limit of the instrumentation location of tensiometer 819 contour of same water pressure -5 water pressure in kPa contact clay-silt 30 40 813
0
Figure 3.
10
20
Distance (m)
Distribution of water pressures in the slope on February 1st.
1245
The silt presents a well-defined foliation plan that makes an angle of about 45 with the horizontal. Considering that ordinarily foliation plans present a lower shear strength than the other rupture plans in a given soil, the conclusion was that the results of the direct shear tests indicated the shear strength of the silt in an imposed plan, different from the foliation plan, and the results of the unconfined compression tests indicated the shear strength of the silt in the foliation plan. Accordingly, it was verified that the soil anisotropy influences its strength and, therefore, the experimental slope stability, but this was not the purpose of this study. 3.3 Summary of the obtained parameters
be used in the stability calculations. The b values indicated in the table relate to water pressure levels close to the values recorded in the monitoring, between 0 and 60 kPa. The values adopted for the silt in the stability calculations are values among those obtained (or adopted) for the two types of tests, considering that the ruptures in soils with foliation plan usually occur partly in such plan and partly in plans different from it. The natural specific weight of clay is 20.5 kN/m3 and of silt is 17.8 kN/m3. 4 DISTRIBUTION OF WATER PRESSURES IN THE SLOPE
The strength parameters of soils in the experimental slope, obtained in direct shear and unconfined compression tests, are shown in Table 1 along with the values adopted for each soil and which will
Table 1. slope. Soil Clay Strength parameters of soils in the experimental Strength parameters c (kPa) () b () C.D. b () C.S. c (kPa) () b () C.D. b () C.S.* Obtained 15 31 36 38 19 35 35 11 Adopted 15 31 36 5 25 11
Silt
* Value calculated based on the adopted null cohesion values and 22, based on literature data (Metro, 1980).
In general the tensiometers installed in the experimental area presented readings coherent with the precipitations, recording increased water pressure in rainy days or the days subsequent to rainy days, and decreased or stable water pressures in drought periods. The readings of the 12 tensiometers installed in the experimental slope section covered with vegetation were used to produce boundary iso-values of water pressures on February 1st and August 31, as shown in Figures 3 and 4. See Franch (2008) for further monitoring details and how this figure was prepared. During rainy periods, as on February 1st, a suspended groundwater table was detected several times at the top of the slope, on clayey soil, as illustrated in Figure 3. This can be explained by the anisotropy related to the permeability of that soil and the clayey layer with vegetal detritus between the clayey landfill and the silty residual soil, which would have caused the water retention in the clayey layer.
835.35
J L H M
Antecedent Precipitation: 3 days - 0 mm 7 days - 0 mm 14 days - 0 mm 28 days - 9 mm

E F G
837
-60 clayey
embakment
831 Elevation (m)
-10 825
-20 LEGEND limit of the instrumentation 819 location of tensiometer contour of same water pressure -5 water pressure in kPa contact clay-silt
821.15
-20 -30
-50 -40
A B
-60
C D
-30
micaceus silty residual soil
-30
817.42
0
Figure 4.
10
20
30 Distance (m)
40
813
Distribution of water pressures in the slope on August 31.
1246
It is worth noting that in the yearly drought period the lower value of water pressure measured was 60 kPa in the clayey layer. As the air entry suction value of this clay is 150 kPa, it is always saturated. In the rainy period, the water pressure in the clay may be positive and curiously it overlaps the silt that remains unsaturated. On August 31st, the yearly drought period, there is a marked difference in the water pressure value thresholds, which are significantly lower that in the rainy period, as Figure 4 shows. The effect of the evapotranspiration caused by the vegetation in this slope section manifests itself in a decrease in water pressures corresponding to up to a 2.5 m depth in the clayey and a 3.0 m depth in the silty soil. 5 STABILITY ANALYSIS
1.50
r = 35.2
24 22 20 18 16 14 12 10 8 6 4 2
Solo residual silte areno-argiloso Aterro argila areno-siltosa
0 kPa
0 kPa
+10 kPa +20 kPa +20 kPa +10 kPa -5 kPa -10 kPa
0 kPa
-15 kPa -15 kPa -10 kPa -5 kPa
0 0.0 2.5 5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0
Distance (m)
Figure 5. Result of the stability analysis in the vegetation-covered slope section on February 1.
1.63
Below are the results of the stability analysis of the slope monitored in these tests for the purposes of determining the effect of seasonable variations of the water pressure on the slope stability. This effect is shown through the comparison of the safety factors of landslide on the dates the highest (February 1rainy period) and the lowest (August 31drought period) water pressures were recorded during the monitoring period, as shown in Figures 5 and 6, respectively. The water pressure distributions on these dates are represented by lines which represent the boundary iso-values of water pressure in the slope. The water pressure values below the 3.0 m depth are estimated considering that the field readings were made up to that depth. The air pressures in the soil were not considered in this research, as described above. The strength envelopes supplied by equation (2) were used in the stability calculations. According to the manual of the SLOPE/W (GEO-SLOPE 1998) software, used in the stability calculations, the water pressure iso-value lines must begin and end at the end of the work area, therefore the field readings were adjusted. The water level position in the ground along the slope was estimated based on the water level measured at the slope foot, which was instrumentalized through the water level meter. The estimate of the water pressure iso-value lines below the 3.0 m depth considered the contact between the clayey and the silty soil. It was found that below such contact, that is, in the silty soil, the water pressures drop because of the hydraulic characteristics of the soils. The comparison of the safety factors obtained for the dates on which the highest and the lowest water pressures were measured in the slope revealed that in the yearly rainy period the landslide safety of the slope is 7% lower than in the drought period, with the safety factor changing from 1.62 a 1.50 (Spencers Method), as is shown in Table 2.
r = 33,0
-60kPa -50kPa -40kPa -30kPa -20kPa -10kP P 0kPa 0kPa -10kPa -20kPa -30kPa -40kPa -30kPa -20kPa -10kPa 0kPa
Solo residual silte areno-argiloso
5.0 7.5 10.0 12.5 15.0 17.5 20.0 22.5 25.0 27.5 30.0 32.5 35.0 37.5 40.0 42.5 45.0
Distance (m)
Figure 6. Result of the stability analysis in the vegetation-covered slope section on August 31. Table 2. Comparison of the safety factors. Safety factor Date Feb 1 Aug 31 Spencer 1.50 1.62 Bishop 1.51 1.62 Janbu 1.40 1.51 Safety variation 7%
The safety factors obtained by using the Modified Bishop Method and the Janbu Method indicated reductions in the slope safety by 7% and 8%, respectively. It should be considered that the precipitations during the monitoring period were 20% lower that the historical averages for
1247
the monitoring months (January to September), of Posto Jd. Centenrio located in River Aricanduva Basin, near the experimental field. It is possible that heavier precipitations would cause greater reductions in the slope safety. Curiously the layer of clayey landfill retains a great part of the water so that this causes the water pressures in the residual soil to vary little, as detected during the monitoring. On the other hand, there could be a landslide in the clayey layer should the positive water pressures increase significantly. 6 CONCLUSIONS
retains the water, and the water pressures in the residual soil layer vary little. REFERENCES
CAMPOS, T.M.P. 1997. Resistncia ao cisalhamento de solos no saturados (Shear strength of unsaturated soils). III Simpsio Brasileiro sobre Solos Nosaturados (III Brazilian Symposium on Unsaturated Soils), vol. 2, 399417. FRANCH, F.A.J., Futai, M.M. and Marinho, F.A.M. 2005. Medio da suco com uso de tensmetro em um talude no saturado (Suction measurement using tensiometer in a non saturated slope). IV Conferncia Brasileira sobre Estabilidade de Encostas (IV Brazilian Conference on Embankment Stability), Vol. 1. FRANCH, F.A.J. 2008. Influncia do tipo de revestimento superficial no fluxo no saturado e sua influncia na estabilidade de taludes (Influence of the surface coating type on unsaturated flow and its influence on slope stability). Dissertation for Masters Degree from Polytechnic School of the University of So Paulo, So Paulo. FREDLUND, D.G., MORGENSTERN, N.R. e WIDGER, R.A. 1978. The shear strength of unsaturated soil. Canadian Geotechnical Journal, Vol. 15, n. 3, pp. 313321. FREDLUND, D.G. 1979. Appropriate concepts and technology for unsaturated soils. Canadian Geotechnical Journal, Vol. 16, pp. 121139. FUTAI, M. M., ALMEIDA, M. S. S. and LACERDA, W. A. 2004. Resistncia ao cisalhamento de solos tropicais no saturados (Shear strength of unsaturated tropical soils). 5th Brazilian Symposium on Unsaturated Soils. Volume I, pp. 4354, So Carlos SP. GEO-SLOPE International Ltd. SLOPE/W for slope stability analysis 1998. Users Guide, Version 4, Calgary, Alberta, Canada. METRO COMPANHIA DO METROPOLITANO DE SO PAULO 1980. Technical documentation NC-03/80.
This paper presented the application of the concepts of unsaturated soils to the stability of embankments. The results of unsaturated strength tests of two soils, a clayey landfill and a residual soil, were shown. The tensiometers were installed in the slope to allow generating two patterns of water pressure iso-values, one for the drought period and the other for the rainy period. It was found that in the rainy period positive water pressures were recorded in the layer of the clayey landfill. The slope stability analyses were performed considering the strength envelopes of unsaturated soils and the water pressure distribution patterns measured in field. It was found that there is a less significant decrease in the landslide factor of safety when comparing the slope in the rainy period and in the drought period. This may be explained considering that: (a) the monitoring was carried out during a period of an anomalous year with less than the average precipitations; and (b) the clayey layer
1248
Shear and soil-water retention behaviour of a variably saturated residual soil and its implication on slope stability
A. Jotisankasa & J. Tapparnich
Geotechnical Engineering Research and Development Center (GERD), Department of Civil Engineering, Faculty of Engineering, Kasetsart University, Thailand
ABSTRACT: Numerous shallow slope failures took place in residual soils derived from sedimentary rock formation of Uttaradit province, Northern Thailand in 2005, due to prolonged and intense rainfall. Shear behaviour as well as water retention behaviour of this material has thus been investigated in details in order to investigate the slope failure mechanism. Fully saturated Consolidated-Drained (CD) as well as suction-monitored direct shear tests have been performed on undisturbed samples collected from depths of 0.31 m. A miniature tensiometer has been used for suction measurement during these tests. In addition, influence of number of drying/wetting cycles on saturated shear strength is investigated. The results from a simple infinite slope analysis suggest that the major slope destabilization mechanism is a combination of material degradation and pore water pressure increase. 1 1.1 BACKGROUND 2006 Uttaradit landslide
In May 2123, 2006, a heavy rainfall of about 400 mm triggered flash flood and hundreds of shallow landslide in northern provinces of Thailand, including Uttaradit. Nearly a hundred people were reported dead or missing. Considerable damage was also incurred to infrastructure such as bridges, roads, drainage systems, and agricultural lands. Although the hourly rainfall data during the incident is not available, the local communities reportedly observed an unprecedented heavy rainfall over a period of two to three hours prior to landslide (ADPC, 2006). The mass movements were categorized into four types: shallow earth slips, gully erosion, failures of cut slopes, as well as erosioninduced bank failure. The predominant category (about 80% of all failures) is the shallow slip type which occurred within the residual soil mantle to a bedrock depth of 40100 cm (Fig. 1). The topography of these areas consists of hilly terrains with prevailing gradients ranging from about 25 to 45 deg. Deforestation is believed to be another contributing factor on the shallow landslides in Uttaradit (ADPC, 2006). In addition, slope cutting for the purposes of settlement expansion and road construction is another cause of slope failure in the area. The geology of the landslide vicinity is mainly of Carboniferous/Permian sedimentary rocks including siltstone, mudstone, and shale, which show distinct fold structure. It has been shown that such argillaceous materials are likely to degrade with time after exposure to weather
Figure 1. Shallow landslides in Uttaradit in 2006.
or wet/dry cycles (e.g. Sabatini et al., 2002 and Alonso & Pineda, 2007). Recently cut slopes that at first appear stable will eventually exhibit surficial sloughing after some years if unprotected. 1.2 Shallow slope failure mechanism Such problems of shallow landslide have received considerable attentions by the research communities worldwide (e.g. Vaughan, 1985, Cho & Lee, 2002, Ng & Shi, 2003, Collins & Znidarcic, 2004, and Rahardjo et al., 2007). The slopes involved in shallow failure are in general within the vadose zone, which exhibit soil suction, or negative pore water pressure, most time of the year. Prolonged rainfall and subsequent infiltration diminished soil suction to nearly zero at a critical depth of slope and often become a triggering mechanism of shallow slope failure. Even though many fast-moving shallow landslide and debris flow are finally triggered by positive pore water pressure rise or static liquefaction (e.g. Vaughan, 1985, and Johnson & Sitar, 1990) the incorporation of unsaturated soil properties into slope stability and infiltration analysis has been shown to yield a better representation of actual soil
1249
slope behavior. As evidenced by Collins & Znidarcic (2004) and Godt et al., (2009), slopes with a relatively steep gradients (e.g. 45 deg.) might fail when the pore water pressure becomes close to zero and not yet highly positive. The main objective of this study is therefore to investigate the shear and soil-water retention behaviour of the residual soils from the Uttaradit landslide area both in saturated and unsaturated conditions. A typical slope within the area has also been instrumented in order to obtain a better picture of the pore water pressure variation in the field. In addition, the influence of wetting/drying cycles on shear behaviour of the soil will be studied, together with their implications on slope stability. 2 2.1 INSTRUMENTED SLOPE Site description
Figure 2. Photo of the studied site and its cross-section.
Top Soil Residual Soil (VI) Completely Weathered Rock (V)

Figure 3. Typical soil profile of the studied area, based on Little (1969) interpretation.
The studied site is situated near the foothill in the agricultural area with a variety of fruit plantations and an average slope angle of about 26 deg. The character of this site is considered representative of the major landuse type in the hilly area of Uttaradit. Figure 2 shows the photo of the site and lay out of instruments. Five stations of tensiometers, and inclinometers were installed at five elevations. The tensiometers have been developed by Jotisankasa et al. (2007) and manufactured at Kasetsart University using commercial miniature MEMs pressure sensors and standard 1bar AEV porous stone. The devices, capable of measuring pore water pressure ranging from 80 to 600 kPa, were installed at depths of 0.5 m and 1 m. One automatic tipping bucket raingauge was also installed at the toe of slope. Inclinometers installed were based on MEMs accelerometer attached to a PVC tube buried to a depth of relatively competent bedrock at about 1 m. The readings from the sensors were recorded with a logging system that can wirelessly transfer the data to mobile phones. In addition, a test pit was also excavated to a depth of about 1 m for visual inspection (Fig. 3), thin-wall tube sampling, and double-ring infiltration tests. Falling head tests were performed within lined boreholes to determine the so-called saturated permeability at depths of about 2 and 3 m. The interpretation of the falling head test was based on Hvorslev (1951) and Garga & Blight (1997). As shown in Figure 4 and Table 1, at greater depth, the soil becomes coarser, denser and of greater value of specific gravity. The deeper completely weathered rock was with more visible feature of cementation. The upper material likely to be involved in slope failure is classified as low plasticity silt. Figure 5 shows the decrease in permeability with depth. The value of permeability reduces, by about three to four orders of magnitude, from the upper 1 m of the soil to the bedrock at the depth of 23 m.

Figure 4. Table 1.
Grain size distribution curves. Basic soil properties and classification. Atterbergs limits, % LL 48.2 40.5 40.8 39.9 PL 32.4 29.4 29.6 28.1 PI 15.7 11.1 11.2 11.8 Gs 2.616 2.724 2.728 2.746
Depth (metre) Top soil 0.3 0.8 1.0
Void ratio, e 1.090 0.05 0.963 0.05 0.703 0.04 0.736 0.07
USCS ML ML ML SM
1250

which the pore water pressure reduced during the dry season (from 8/11/2008 onwards) appeared to be higher as pore water pressure decreased. The inclinometer reading did not indicate any significant movement of the slope during the measurement. This observation agrees with the negative pore water pressure of the soil which gave rise to stability of the slope. 3 3.1 SHEAR AND WATER RETENTION BEHAVIOUR Saturated shear behaviour
Figure 5.
10
Variation of permeability with depth.

a)
Pore water pressure, kPa
0 -10 -20 -30 -40 -50 -60 -70 10 0 -10 -20 -30 -40 -50 -60 -70 -80 80 Depth 0.5 m
Level 2 Level 4 Level 5
Pore water pressure, kPa
b)
Depth 1 m
Level 3 Level 5
A series of multistage and single-stage direct shear tests on saturated and unsaturated undisturbed specimens about 63 mm in diameter was performed to investigate shear behaviour. A conventional direct shear box was used to determine the effective strength parameter (c, ) in slow consolidated-drained multistage shearing tests, conducted at three normal stresses of 16, 32, 64 kPa. The samples tested were taken from the agricultural area at ground surface (indicated as 0 metre) and at depths of 0.3, 0.7, 0.8, and 1 metre. The soil at 1 m depth has the highest effective cohesion intercept due to its cementation: a feature expectable of a structured soil or completely weathered rock. 3.2 Unsaturated shear behavior
70
c)
Daily rainfall, mm
60 50 40 30 20 10 0
1/9/2008 1/10/2008 1/11/2008 1/12/2008 1/1/2009
Date
Figure 6. rainfall.
Variation of pore water pressure and daily
2.2
Monitoring results
Large scatter in the permeability data is evident especially for the ground at depth less than 1 m, which is probably attributed to the heterogeneity of the residual soil. As also noticed by Vaughan (1985) and Gerscovich et al. (2006), the accentuated variability in permeability can lead to strongly non-uniform distribution of pore water pressure and sometimes highly positive pore water due to impeded seepage.
The suction-monitored direct shear tests on the unsaturated specimens were conducted at a constant normal stress of 16 kPa. This value of normal stress corresponds to a soil thickness of about 0.8 to 1 metre, which is the expected failure surface. The apparatus used was a conventional direct shear box, modified in order that the KU-tensiometer can be inserted through the top cap and monitor the soil suction during shear, as shown in Figure 8 (Jotisankasa and Mairaing, 2009). All shearing tests were done with samples maintained in the constant water content condition and at a shearing rate of 0.05 mm/min. Examples of test results are illustrated in Figure 9. The insets in the figure refer to the different values of initial suction in the tests. The range of samples

Figure 6 shows the variation of pore water pressures and daily rainfall with time. Both at the depth of 0.5 and 1 metres, the pore water pressure remained negative for most time of the year. Despite the rainfall of about 68 mm on the day of 3/11/2008, the pore water pressure remained only about zero kPa, and not much greater than that. The rate at

Figure 7. Effective failure envelops of saturated samples.
1251
initial suction was chosen to reflect the actual variation in the field as shown in Figure 6. The samples with higher suctions exhibit higher strength and tend to dilate more. During the initial stage of shearing, the soil volume and the suction slightly decreased. As the samples started to dilate and reaching the peak strength, the suction also appeared to increase accordingly. The sample with suction of zero was tested in a multistage manner and therefore no results are available for large horizontal displacement. Figure 10 shows the variation of shear strength and suction for the samples at depth of 0.7 m. The fitting equation used for the unsaturated shear strength is that of Fredlund & Rahardjo (1993) as follows: c ( ua) tan (ua uw) tan b (1)
where c effective cohesion intercept, normal total stress, ua pore air pressure (for the direct shear tests carried out at atmospheric pressure, ua equals zero), uw pore water pressure, is the effective angle of shearing resistance, and b the angle of shearing resistance with respect to suction. Table 2 summarizes all the shear strength parameters. The variation between the shear strength and suction appears to be nonlinear. It is appreciated that various mathematical expressions have been proposed to fit this non-linear trend (e.g. Vanapalli et al., 1996), though none of them will be presented here for brevity of the paper. 3.3 Soil-water retention curves
Figure 8.
Suction-monitored direct shear box.

60
Shear stress, kPa
50 40 30 20 10 0 1.0
.8 .6 .4 .2 0.0 -.2 -.4 100
80 60 40 20 0 0 1 2 3 4 5
Matric suction,
a)
kPa kPa kPa kPa kPa
The volumetric water content-suction relationships have also been determined for all samples, as shown in Figure 11. Suction measurements were made using the miniature tensiometers and relative humidity sensor (Jotisankasas et al., 2007). The test procedure involved gradually wetting and drying the sample while measuring the suction, weight and dimensions at each stage. During the wetting path, once the soil suction fall below 1 kPa, the specimen was submerged under water with a nominal vertical overburden stress of 1 kPa for a period of at least 5 days. The values of water content at suction of 0.1 kPa, shown in Figure 11, was in fact arbitrarily chosen to indicate the water content of soaked samples, s, in the logarithmic suction plot. This soil water retention curve was also used to predict the failure envelope, using the simplified approach by Jotisankasa & Mairaing (2009) (Eq. 2),
80
Vertical displacement, mm
Shear strength, kPa
b)
70 60 50 40 30 20 10 0 0 10 Experimental results Jotisankasa&Mairaing, (2009) 20 30 40 50 60 70 80
= 0 kPa = 2 kPa = 10 kPa = 30 kPa = 57 kPa
Matric suction, (ua-uw), kPa
c)
kPa kPa kPa kPa kPa
Matric suction(ua-uw), kPa
Figure 10. Shear strength versus suction of sample at depth of 0.7 m. Table 2. Shear strength parameters. c kPa 22.8 4.6 15.9 13.7 37.2 deg. 17.6 40.4 36.6 32.0 24.5 b deg. 27.7
Depth (metre)
6 7 8
Horizontal displacement, mm
Figure 9. Shearing behaviour of the samples at depth of 0.7 m with different suctions in suction-monitored direct shear box tested at a normal stress of 16 kPa.
Top soil 0.3 0.7 0.8 1.0
1252
Volumetric water content, %
.6
80
Shear strength, kPa
.5 .4 .3 .2 .1 0.0 .1 0 m-wetting curve 0 m-drying curve 1 m-wetting curve 1 m-drying curve 1 10 100 1000
70 60 50 40 30 20 10 0 0 10 20 30 40 50 60 70 80
Zero cycle 5 cycles 10 cycles
Suction, kPa
Normal stress, kPa
Figure 11.
Soil-water retention curves.
Figure 12. envelope.
Influence of drying/wetting cycles on failure
[c ( n ua )tan ] (ua uw ) 33 (tan ) s (2) where s saturated volumetric water content, and 33 is the volumetric water content at 33 kPa suction or at the nominal field capacity. As shown in Figure 10, Jotisankasa & Mairaing (2009) approach could predict the lower bound of the shear strength satisfactorily for suction less than about 30 kPa. As suggested by Jotisankasa & Mairaing (2009), this approach is suitable to be applied for approximation of unsaturated shear strength for very large areas where only available data are the field capacity of the soils and not the complete soil-water retention curves. 3.4 Influence of wetting/drying cycles
account the saturated/unsaturated shear strength. Equation (4) was used to calculate factor of safety, F, for both saturated and unsaturated case, F c ( z cos 2 ) tan uw tan z sin cos (4)
In order to clarify the expected degradable properties of the material, a series of shearing tests were carried out on samples taken at depth of 1 metre which followed varying cycles of wetting/drying. For each cycle, the sample was soaked in water for a couple of days before oven-dried at 105C for a day. This was believed to represent the extreme case of temperature and moisture content change that soil can undergo in the field. It is noted that the soil at 1 m depth in the field (zero wet/dry cycle) behaves as structured soil exhibiting some effective cohesions. The effective cohesion of sample subjected to 5 cycles of wetting/ drying was reduced to nearly zero by destructuration process (Fig. 12). With greater cycles, the failure envelope still remained relatively unchanged. The sample must have been degraded to the extent that nearly all the cementation structure has been destroyed after no more than 5 cycles. 4 IMPLICATIONS ON SLOPE STABILITY
where, if uw 0, and b if uw 0. Hypothetical slopes with depth of failure, z 0.5 and 1 m, and slope gradient, 25, 45, and 60 degrees, were analyzed. These slopes represent the typical range of terrains in Uttaradit area that suffered from landslide. The soil properties used were those from Table 2 and Figure 12. A constant value of b 27.7 degree is assumed for all materials representing the average value in the range of suction between 0 and 20 kPa. The variation between the factor of safety and pore water pressure were then plotted in order to find out the triggering pore water pressure for different scenarios. It can be seen in Figure 13 that the pore water pressure which triggers the slope failure (at F 1) is of positive values for both slope gradients. As for the influence of drying/wetting cycles (Figure 14), for steep slope ( 60 deg.) with nearly zero suction, only five cycles (or less) of extreme wetting/drying cycles might cause some instability of the slope. The soil suction thus provides a crucial stabilizing effect for the degraded slope (destructured/cementation destroyed) with gradient more than about 60 deg. 5 CONCLUSIONS
In order to put all these observed behaviour in the context of slope stability, some simple infinite slope analyses were carried out which take into
This paper describes some important soil behaviour affecting mechanism of rainfall-induced shallow slope failures in residual soils derived from sedimentary rock formation of Uttaradit province. The shear strength of the soil increases with depth, whereby the soil at 1 m depth has the highest effective cohesion intercept due to its cementation. This cohesion however can be destroyed by a couple of cycles of extreme wetting/drying. Unsaturated shear strength has also been determined using the
1253
7 6 Factor of Safety 5 4 3 2 1 0 -30 c' = 4.6 kPa ' = 40.4, b = 27.7 H = 0.5 m t = 19.2 kN/m3 -20
slope angle 25 deg. slope angle 45 deg.
a)
-10 0 10 pore water pressure, u, kPa
20
30
5 Factor of Safety 4 3 2 1 0 -30 c' = 13.7 kPa ' = 32.0, b = 27.7 H=1m t = 19.3 kN/m3 -20
slope angle 25 deg. slope angle 45 deg.
b)
-10 0 10 pore water pressure, u, kPa
20
30
Figure 13. Variation between factor of safety and pore water pressure for different slopes (zero wetting/drying cycles).
7 6 Factor of Safety 5 4 3 2 1 0 -30 -20 -10 0 10 pore water pressure, u, kPa 20 30 zero cycle 5 cycles 10 cycles
= 27.7 H=1m t = 19.3 kN/m3 Slope gradient 60 deg.
Figure 14. Variation between factor of safety and pore water pressure for slope with varying drying/wetting cycles.
suction-monitored direct shear tests. The variation between the shear strength and suction appears to be nonlinear. The results from a simple infinite slope analysis suggest that the major slope destabilization mechanism is a combination of material degradation and pore water pressure increase. ACKNOWLEDGEMENTS This study was financially supported by the National Research Council of Thailand (NRCT). The landslide photo shown is from the archive of Geotechnical Engineering Research and Development Center of Kasetsart University, Thailand. The assistance of Mr. Wasan Suttinun, and Mr. Narin Hunsachainun during the installation of instruments at the studied site is thankfully acknowledged. REFERENCES
ADPC-Asian Disaster Preparedness Center (2006). Rapid assessment, Flashflood and landslide disaster in the provinces of uttaradit and sukhothai, northern Thailand, May 2006. (www.adpc.net/enewsjuly/Uttaradit_ rapidassessment.pdf)
Alonso, E.E. and Pineda, J.A. (2007). Degradation of argillaceous rocks: a challenge for unsaturated geomechanics. Proc. 3rd Asian Conference on Unsaturated Soils (Unsat-Asia 2007), Nanjing, China, 2123 April 2007, 326. Cho, S.E. and Lee, S.R. (2002). Evaluation of Surficial Stability for Homogeneous Slopes Considering Rainfall Characteristics. Journal of Geotechnical and Geoenvironmental Engineering. ASCE. Vol. 128, No. 9, September 1, 756763. Collins, B.D, and Znidarcic, D. (2004). Stability analyses of rainfall-induced landslides. Journal of Geotechnical and Geoenvironmental Engineering. ASCE. April 2004, Vol. 130, No. 4, 362372. Garga, V.K. and Blight, G.E. (1997). Permeability. Mechanics of residual soils. (ed. G.E. Blight), A.A. Balkema, Rotterdam, 7994. Gerscovich, D.M.S., Vargas, E.A., de Campos, T.M.P. (2006) On the evaluation of unsaturated flow in a natural slope in Rio de Janeiro, Brazil. Engineering Geology, 88 (2006) 2340, Elsevier B.V. Godt, J.W., Baum, R.L. and Lu, N. (2009). Landsliding in partially saturated materials. Geophysical Research Letters, Vol. 36, 15, L02403, doi: 10.1029/2008 GL035996. Hvorslev, M.J. (1951). Time-lag and soil permeability in groundwater observations. Bulletin No. 36, US Waterways Experiment Station. Vicksburg, Miss., USA. Johnson, K.A. and Sitar, N. (1990) Hydrological conditions leading to debris-flow initiation. Canadian Geotechnical Journal, 27: 789801. Jotisankasa, A., Porlila, W., Soralump, S. and Mairiang W. (2007). Development of a low cost miniature tensiometer and its applications. Proc. 3rd Asian Conference on Unsaturated Soils (Unsat-Asia 2007), Nanjing, China, 2123 April 1007, 475480. Jotisankasa, A. and Mairaing, W. (2009). Suctionmonitored direct shear testing of residual soils from landslide-prone areas, Journal of Geotechnical and Geoenvironmental Engineering, ASCE. Ng, C.W.W. and Shi, Q. (2003). The stability of unsaturated soil slopes subjected to rainfall infiltration. Numerical Analysis and Modelling in Geomechanics (John W. Bull) London: Spon Press, 101128. Sabatini, P.J., Bachus, R.C., Mayne, P.W., Schneider, J.A. and Zettler, T.E. (2002). Shales and degradable materials. Evaluation of Soil and Rock Properties, Report no. FHWA-IF-02-034, Geotechnical Engineering Circular No. 5, 243. Rahardjo, H., Ong, T.H., Rezaur, R.B. and Leong, E.C. (2007). Factors controlling instability of homogeneous soil slopes under rainfall. Journal of Geotechnical and Geoenvironmental Engineering. ASCE. Vol. 133, No. 12, December 1, 15321543. Vanapalli, S.K., Fredlund, D.G., Pufhal D.E. and Clifton, A.W. (1996). Model for the prediction of shear strength with respect to soil suction. Can. Geot. J. 33, 379392. Vaughan, P.R. (1985). Pore pressures due to infiltration into partly saturated slopes. Proc. 1st International Conference on Geomechanics in Tropical Lateritic and Saprolitic soils. Brazil, Vol. 2, 6171.
1254
Back-analysis of infiltration characteristic of a saprolitic hillslope by considering permeability heterogeneity

A.K. Leung & C.W.W. Ng
Department of Civil and Environmental Engineering, Hong Kong University of Science and Technology, HKSAR, Hong Kong
ABSTRACT: In situ geological features always induce permeability heterogeneity. To account for the hydraulically heterogeneity of a porous material implicitly, the two parameters of a permeability tensor, namely principal permeability ratio, kx/ky, and principal flow orientation, , may be used. In particular, the effect of on Pore-Water Pressure (PWP) distribution is sometimes overlooked and rarely discussed. In an attempt to investigate how permeability heterogeneity influences the transient seepage in a saprolitic hillslope, a case history is back-analysed by carrying out a series of parametric study. Both kx/ky, and are examined in this study. The measured and computed PWP profiles are compared, particularly focusing on the magnitude of PWPs as well as the change of groundwater table. By considering the in situ geological features, proper set of parameters for each material is deduced to capture the in situ infiltration characteristic of the hillslope. 1 INTRODUCTION the direction of principal permeability, ki (i x, y), coincides the global axis i. After matrix operation, the components of [K], kij, can be expressed as (SEEP/W 1998): kxx kx cos 2 ky sin2 kxy kyx kx cos sin ky cos sin kyy ky cos 2 kx sin2 where is the angle between the global and local axis. A positive indicates that the dip direction of ki is the same as that of the slope, and vice versa. The diagonal components of [K] are the principal permeabilities along the local axes. On the contrary, the off diagonal components are the associated permeabilities when the direction of ki does not coincide with the global axes. The influence of kx/ky on the pore-water pressure response in a soil slope has been extensively investigated (Iverson 1990; Reid 1997; Ng & Shi 1998). However, the effect of on transient seepage on a hillslope has rarely been reported. In this study, the IP test is back-analysed. A series of seepage analyses is conducted using different sets of [K], where both kx/ky and are taken into account. By considering the in situ geological features, a set of kx/ky and for each material is deduced, aiming to capture the overall seepage characteristic of the hillslope. 2 BRIEF DESCRIPTIONS OF THE IP TEST (2)
Stress-Dependent Soil-Water Characteristic Curve (SDSWCC) (Ng & Pang 2000) and permeability function are the vital hydraulic properties of an unsaturated soil to characterise transient seepage in saturated/unsaturated soil system. Field determinations of these properties are always valuable due to its representativeness of the real conditions. An Instantaneous Profile (IP) test was recently conducted at a saprolitic hillslope in Hong Kong to estimate the in situ SDSWCCs and permeability functions (Ng et al. 2010). Unfortunately, geological features that are frequently found in the hillslope (Corley et al. 1999; Aydin et al. 2000) may sometimes hinder the interpretation of the field data and hence the estimation of the in situ hydraulic properties. The presence of stratification and discontinuities often induces permeability heterogeneity of a soil matrix. Water may preferentially flow at certain orientation at a higher permeability. To account for the hydraulic heterogeneity of a porous medium implicitly, a second-order permeability tensor, [K], may be used (Bear 1972; Oda 1985; Dmitriev et al. 2003). The [K] in the local coordinate system (x, y) can be mathematically expressed as follows.
K C T K C C T
kxx kyx kxy kyy
kx 0
0 ky
C
(1)
where [C] is the direction cosine; [K] is the permeability tensor in global coordinate system (x, y), where
Figure 1 shows the schematic setup of the IP test at the toe of the hillslope. The slope gradient is
1255
3500
3000 Test ring, diameter = 3m Top Soil Colluvium CDT 50 JFT_2 JFT_3
500
3.1 Transient water flow in unsaturated soils Extensive experimental investigations have been verified that Darcys law is applicable for water flow through unsaturated soils (Buckingham 1907). Considering the continuity equation and Darcys law of a soil element, the governing equation for two-dimensional water flow in a hydraulically heterogeneous porous medium can be expressed as follows (SEEP/W 1998): hw h h h kyx w kxy w kyy w kxx x x y y x y hw 0 q t mw w t (3) where hw is the total head; q(t) is the external unit flux; w is the unit weight of water; and mw is the slope of a SDSWCC. In this analysis, the volume change of the soil during drainage or ponding processes is ignored. 3.2 Geometry and input parameters
2900
45 o JFT_1 TDR_1 JFT_4 JFT_5 TDR_2 JFT_9 JFT_10 TDR_4
TDR Depth: TDR_1 - 0.835m TDR_2 - 1.853m TDR_3 - 2.500m TDR_4 - 3.585m
2000
Polythene sheeting Compacted Fill
JFT_6 JFT_7 JFT Depth: JFT_8 JFT_1 - 0.36m TDR_3

JFT_2 - 0.77m JFT_3 - 0.95m JFT_4 - 1.17m JFT_5 - 1.54m JFT_6 - 1.85m JFT_7 - 2.13m JFT_8 - 2.43m JFT_9 - 2.60m JFT_10 - 2.99m
1200
All TDRs are located at 150mm from the center of the test ring. All JFTs are located at 600mm from the center of the test ring. All dimensions are in mm, except otherwise stated. GWT: below platform 10m
Figure 1. The test set up and the instrumentation plan (Ng et al. 2010).
Table 1. Stage
IP test schedule. Phase Date Duration 4 days 24 days 4 days 16 days
1 (Day 04) 1st Ponding 2 (Day 428) 1st Drainage 3 (Day 2832) 2nd Ponding
26 Oct 2007 630 Oct 2007 30 Oct 3 Nov 2007 4 (Day 3236) 2nd Drainage 319 Nov 2007
200
approximately 28 in average. A 3 m deep, 1.2 m wide trench was excavated at upslope side of the test plot to install a 0.06 mm think polythene sheeting. It acted as a cut-off sheet to minimise the recharge from upslope. After the installation of the polythene sheeting, the trench was backfilled. Subsequently, a circular steel test ring (3 m in diameter) was embedded 100 mm into the flattened plot (3.5 m 3.5 m) to retain water upon the ponding processes. Ten jet-fill tensiometers and four time domain reflectometry moisture probes were installed to respectively measure the negative Pore-Water Pressure (PWP) and to deduce the volumetric water content through a soil-specific calibration curve. The test schedule is summarised in Table 1. For each ponding phase, a water level of about 0.1 m was applied on the ground surface inside the ring and the water level was checked and refilled to the same level every 12 hours. For each drainage phase, the test plot was allowed to dry under natural evaporation. 3 BACK-ANALYSIS OF THE IP TEST
3.2.1 Finite element mesh Figure 2 shows the simplified finite element mesh of the saprolitic hillslope as shown in Figure 1. The hillslope is assumed to be inclined at 30 to the horizontal. A 1 m-thick colluvium is overlaid at the ground surface while the underlying CDT is 3 m thick. To incorporate the stress effect on the hydraulic properties, the CDT is divided into two layers (CDT1 and CDT2) according to their approximate stress levels. The bedrock is situated at 4 m depth. The 3 m-depth trench for polythene sheeting (PS) installation is also simulated at upslope side. It is done by switching the origin material properties to the ones of PS when the IP test commences.
The inevitable in situ geological features may sometimes hinder the interpretation of field-measured data. Numerical back-analysis of the IP test is perhaps one of the feasible methods to further improve the understandings of the monitoring results. In the following numerical simulations, a series of seepage analysis is conducted using SEEP/W (1998).
Figure 2. Finite element mesh of the idealised hillslope.
1256
3.2.2 Boundary conditions To resemble the initial PWP profiles and the groundwater level as observed in the field, the following boundary conditions are applied. A constant total head of 24 m and a reviewed zero flux (SEEP/W, 1998) is respectively specified on the right boundary EF and the left boundary AB. The bedrock is assumed to be impermeable and a zero flux bottom boundary is applied on boundary AF. Prior to the start of the IP test, a small rainfall intensity of 0.1 mm/day is first applied on the boundary BCDE under a steady-state condition. Subsequently, a transient seepage analysis is conducted by specifying the recorded rainfall intensity happened from 130 September 2007 on boundary BCDE. It is assumed that the infiltration rate is 60% of the rainfall intensity (Ng et al. 2001). The procedures of the numerical simulation are the same as those summarised in Table 1. Each 4-day ponding phase is simulated as follows. A pressure head of 0.1 m is applied on the boundary BCDE. The pressure head then declines linearly to 0 m after 12 hours. This pattern is repeated until the end of the forth day. For each drainage phase, a zero flux is imposed on the boundary BCDE to simulate the internal drainage. Natural evaporation is not considered in this study for simplicity. 3.2.3 Soil parameters Two vital hydraulic properties required to solve equation (3) are the SWCC/SDSWCC and the permeability function. For infiltration problems, it is more appropriate to use wetting SDSWCCs and permeability functions instead of drying ones (Ng & Pang 2000). The wetting SDSWCCs of colluvium at 0 kPa and that of CDT at 40 and 80 kPa at low suctions (500 kPa) were measured using a modified one-dimensional stress-controllable volumetric pressure plate extractor (Ng & Pang 2000). Each laboratory-measured SDSWCC was then fitted and extrapolated using Fredlund & Xing (1994). As illustrated by Leung & Ng (2009), about 1% of water content by mass was retained at total suction of 1000 MPa for CDT material at zero vertical stress. It may be reasonable to postulate that specimen having higher stress levels has greater water retention capability at high suctions (500 kPa), as observed at low suctions. The extrapolated SDSWCC for each material at high suctions is thus modified accordingly. Figure 3 shows the fullsuction range wetting SDSWCC for each material. On the other hand, the wetting permeability function for each material is predicted using Fredlund et al. (1994) and is shown in Figure 4. For simplicity, the hysteretic behaviour of the SDSWCCs and permeability functions is not considered in this study. Regarding the PS, it is assumed that it does not have water retention capability, i.e., w 0, while
40 Volumetric water content (%) 35 30 25 20 15 10 5 0 0.1 Col-N-0-W CDT-N-40-W CDT-N-80-W 1 10 100 1000 Matrix suction (kPa) 10000 100000 1000000
Figure 3. Full-suction range wetting SDSWCCs for colluvium at 0 kPa and CDT at 40 and 80 kPa.
1.E-06 1.E-07 1.E-08 Permeability (m/s) 1.E-09 1.E-10 1.E-11 1.E-12 1.E-13 1.E-14 0.1 1 10 Matrix suction (kPa) 100 1000 Col-N-0-W CDT-N-40-W CDT-N-80-W
Figure 4. Predicted wetting permeability functions for colluvium at 0 kPa and CDT at 40 and 80 kPa. Table 2. Summary of numerical analyses. Col-N-0-W Case 1 2 3 4 5 6 k-ratio 1 5 10 10 10 10 0 0 0 30 45 0 CDT-N-40-W k-ratio 1 5 10 10 10 10 0 0 0 30 45 30 CDT-N-80-W k-ratio 1 5 10 10 10 1 0 0 0 30 45 0
the permeability with order of 34 m/s is constant for entire suction range. 3.3 Analysis procedures
In this study, the two parameters of [K], namely principal permeability ratio, k-ratio (kx/ky), and principal flow orientation, , are examined (see equation (1)). The numerical simulations conducted are summarised in Table 2. Case 13 and Case 35 respectively aims to illustrate the effect of k-ratio and on PWP
1257
distribution. In addition, the two parameters of each material in Case 6 are chosen to best fit the field measurements. For each case, both k-ratio and remain constant throughout the IP test simulation. 4 4.1 INTERPRETED RESULTS Effect of principal permeability ratio, k-ratio
Physically, the parameter k-ratio represents the relative principal permeability between the global axes. Parametric study from Case 1 to 3 aims to investigate the effect of k-ratio on the flow characteristic in the hillslope. The direction of ki for all materials is considered to coincide the global axis, i.e., 0. Figure 5 shows the comparisons between measured and computed PWP profiles across the section A-A (see Figure 2). The hydrostatic line is also shown for ease of reference, assuming the groundwater table is located at about 3 m depth. It is noted that the computed PWPs on Day 0 are resulted from transient seepage after the September rainfall. Moreover, comparison on Day 32, i.e., end of the second ponding phase, is not shown since the computed results are extremely similar to those on Day 4. As shown in Figure 5 (a) and (b), the computed PWPs above 2.5 m depth on Day 0 and 4 are consistently larger than the hydrostatic pressure, indicating a net downward flux. Moreover, the computed PWPs at colluvium appear to be comparable and
-20 0 -10 0 10 20
-20 0
-10
10
20
(a)
Antecedent rainfall
colluvium CDT
(b)
1st
ponding
colluvium CDT
3
Day 0_field Day 0_Case 1 Day 0_Case 2 Day 0_Case 3 Hydrostatic line
Depth (m)
4
-60 0
4
-20 0 20
-40
-20 0
-10
10
20
(c)
1
colluvium CDT
-20 -10 0 0
1st drainage
1
(d)
2nd drainage
colluvium CDT
2
1
are insensitive to the k-ratio. It is probably because the permeability of colluvium at suction 10 kPa (about 109 m/s) is much higher than the average rainfall infiltration rate of the antecedent rainfalls (about 1011 m/s). This hence leads to fast percolation of water in the relative permeable colluvium stratum. Besides, in the CDT stratum, the computed PWPs are apparently the same at depths 1 m to 1.5 m on both days. The predicted GroundWater Table (GWT) level decreases by about 40% when k-ratio increases from 0 to 10. Due to the larger horizontal principal permeability, lateral flows towards the downhill direction become more likely to occur and hence limits the rise of GWT. In comparison with the field-measured data, the simulated results appear to have fair agreement on Day 0 and 4 when the permeability is considered to be isotropic. Figure 5 (c) and (d) respectively depicts the comparison between measured and computed PWP profiles on Day 28 and 48. As drainage takes place, the water seeps under gravity action. The PWP profiles on both days gradually move towards the hydrostatic line as the moisture continuously loses. Since the evaporation process is not considered in this study, it is not surprised that net downward flux is resulted upon the two drainage phases. Nevertheless, the comparable measured and computed PWPs at colluvium on Day 48 indicates that the effect of evaporation in the field was not pronounced after a 16-day drainage period (see Figure 5 (d)). On the other hand, the GWT rises dramatically under isotropic permeability condition, which is significantly over-predicted by more than 1 m (30%) from the actual case. Similar to those observed on Day 0 and 4, the GWT level in the CDT stratum decreases significantly as k-ratio increases, probably due to the increase amount of lateral water flow. Despite the good agreement between measured and computed PWPs in Case 3 on Day 28 and 48, the large k-ratio gives unsatisfactory predictions on Day 0 and 4, where both PWPs and GWT level are significantly under-predicted. The comparisons indicate that the hillslope hydraulic behaviour cannot be captured by solely examining one single parameter, k-ratio. The unsatisfactory predictions may probably due to the unrealistic presumption that is equal to 0. In other words, substantial lateral water flow along some inclined preferential paths might happen in the field. 4.2 Effect of principal flow orientation, The principal flow orientation, , physically represents the average preferential flow path orientation. To examine the effect of on the PWP distribution, another three cases of seepage analysis (Case 3 to 5) are conducted using different values
3
Figure 5. Comparisons of measured and computed PWP profiles due to k-ratio on (a) Day 0 (b) Day 4 (c) Day 28 and (d) Day 48 at section A-A.
1258
of while the k-ratio is constant for all materials (see Table 2). The comparison between measured and computed PWP profiles for Case 3 to 5 is shown in Figure 6. On Day 0, it can be seen that the computed PWPs at colluvium become less when the increases to 45 clockwise (see Figure 6 (a)). It is probably because the increase of results in greater amount of lateral flow and greater moisture loss. The suction hence increases subsequently (see Figure 3). On the other hand, the computed PWP profiles are fairly comparable at the CDT stratum. The magnitude of PWPs is uniform between 1 m to 2.5 m depths at about 4 kPa. Moreover, the increase of from 0 to 45 appears to raise the GWT level by about 40%. Since the k-ratio is larger than 1, i.e., kx is larger than ky, the increase of positive in clockwise direction basically allows more downward seepage to occur under the action of gravity. The continuous accumulation of groundwater hence leads to the rise of GWT. When compared to the field-measured PWP profiles, it is observed that the larger the the closer the agreement. Similar observations are also found at the end of 1st ponding on Day 4 (see Figure 6 (b)). The effect of on the PWP distribution on Day 28 and 48 appears to be similar to that on Day 0 and 4 (see Figure 6 (c) and (d)). However, the rise of GWT level as a result of the increase of is found to be less significant after the two drainage phases (see Figure 6 (a) and (b)). For the
same increment of , the GWT level raise by only about 30%. This reduced amount of GWT rise is probably because the rate of moisture loss (or suction recovery) upon internal drainage attains its steady-state condition after the prolonged drying period. After two ponding-drainage phases, the position of GWT appears to be more or less the same for all three cases. Although the computed PWPs at colluvium are substantially underpredicted by about 10 kPa when compared to the field data, close agreement are generally found on Day 28 and 48, especially at low value of . Given the same k-ratio, the simulated results from Case 3 to 5 demonstrate that different values of can significantly affect the PWP distribution as well as the level of GWT. Therefore, the effect of on transient seepage in a hillslope should not be overlooked. In comparisons with the fieldmeasured PWP profiles, none of the above five cases can satisfactorily mimic the overall flow characteristic in the hillslope by considering either k-ratio or alone. In other words, combinations between these two parameters for each material should be sought to better capture the hydraulic heterogeneity in the field. 4.3 Deduction of hydraulic properties for the field measurement Understanding the geological model of the hillslope may be useful to deduce a reasonable set of k-ratio and for each material. Colluvial material, which is resulted from past mass wasting process, is often found to overly in layers on hillslope surface. Large contrast between the principal permeabilities is thereby expected. Upon continuous wetting-drying cycles, geological process like leaching gradually decomposes the rock and leads to the formation of inclined relict joints (Corley et al. 1999; Aydin et al. 2000). Preferential seepage along these joints may hence occur. As the degree of weathering decreases along depth, the CDT at deeper depth is suspected to have smaller average pore sizes but greater uniformity of individual pore size (Ng et al. 2010). In other words, the permeability of the CDT2 is expected to behave more isotropically. Considering these field characteristics, the k-ratio and for each material are deduced and are shown in Table 2 (see Case 6). The simulated results using the parameters adopted in Case 6 can adequately mimic the overall in situ seepage characteristics in the hillslope (see Figure 6). The PWP profiles computed on Day 0 and 4 are fairly close to the measured ones in terms of the magnitude of PWPs as well as the GWT level (see Figure 6 (a) and (b)). However, the positive PWPs observed at 1.5 m depth, which is the location of colluvium/CDT interface, on Day 0 and 4

-20 0 -10 0 10 20
-20 0
-10
10
20
(a)
1
Antecedent rainfall
colluvium CDT
(b)
1
1st ponding
colluvium CDT
3
Day 0_field Day 0_Case 3 Day 0_Case 4 Day 0_Case 5 Day 0_Case 6 Hydrostatic line
Depth (m)
4
-60 0
4
-20 0 20
-40
-20 0
-10
10
20
(c)
colluvium
1st drainage
1
-10 0
(d)
2nd drainage
colluvium CDT
CDT
-20
2
1
3
Figure 6. Comparisons of measured and computed PWP profiles due to on (a) Day 0 (b) Day 4 (c) Day 28 and (d) Day 48 at section A-A.
1259
cannot be captured. This is probably due to the oversimplification of the in situ soil profile in this numerical analysis. Except the PWPs at colluvium on Day 28, reasonably close agreement between measured and computed PWP is exhibited upon the two drainage phases (see Figure 6 (c) and (d)). Considering the close agreement between measured and computed PWP profiles in Case 6, it may be reasonable to postulate that branches of discontinuities dipping at 30 in average are existed within the CDT stratum at 1 m to 3 m depth in this saprolitic hillslope. Lateral seepage along these preferential flow paths thus reduce the downward seepage and hence limits the GWT rise. 5 SUMMARY AND CONSLUSIONS
Limited and the Hong Kong University of Science and Technology, respectively. REFERENCES
Aydin, A., Duzgoren-Aydin, N.S. and Malpas, J. (2000). A review of igneous and metamorphic saprolities, Journal of Nepal Geological Society, 22: 1116. Bear, J. (1972). Dynamics of fluids in porous media. Elsevier, New York. Buckingham, E. (1907). Studies on the movement of soil moisture. U.S. Dept. Agr., Bur. Soils, 38. Corley, H.P., Martin, R.L. and Macklin, R.L. (1999). Characterisation of saprolite heterogeneities using innovative techniques. Geotechnical Special Publication No. 92, ASCE, pp. 6476. Dmitriev, M.N., Dmitriev, N.M. and Kadet, V.V. (2003). Generalized Darcys law and the structure of the phase and relative phase permeabilities for two-phase flows through anisotropic porous media. Fluid Dynamics. 38(2): 284292. Fredlund, D.G. and Xing, A. (1994). Equations for the soil-water characteristic curve. Can. Geotech. J., Ottawa, 31: 521532. Fredlund, D.G., Xing, A. and Haung, S.Y. (1994). Predicting the permeability function for unsaturated soils using the soil-water characteristics curve. Can. Geotech. J., 31: 533546. Iverson, R.M. (1990). Groundwater flow field in infinite slopes. Geotechnique, 40(1): 139143. Leung, A.K. and Ng, C.W.W. (2009). Experimental study of the soil-water characteristics of a decomposed soil using the Relative Humidity (RH) control technique. Proc. 4th Asia-Pacific Conf. on Unsaturated Soils, 2325 Nov. Newcastle, Australia. pp. 199204. Ng, C.W.W. and Pang, Y.W. (2000). Influence of stress state on soil-water characteristics and slope stability. Journal of Geotechnical and Geoenvironmental Engineering. 126(2): pp. 157166. Ng, C.W.W. and Shi, Q. (1998). A numerical investigation of the stability of unsaturated soil slopes subjected to transient seepage. Computers and Geotechnics, 22(1): 128. Ng, C.W.W., Wang, B. and Tung, Y.K. (2001). Threedimensional numerical investigations of groundwater responses in an unsaturated slope subjected to various rainfall patterns. Canadian Geotechnical Journal, 38(5): 10491062. Ng, C.W.W., H.N. Wong, Y.M. Tse, J. Pappin, Sun. H.W., S. Millis, and Leung, A.K. (2010). Field study of stress-dependent soil-water characteristic curves and Hydraulic conductivity in a saprolitic slope. Geotechnique (provisionally accepted). Oda, M. (1985). Permeability tensor for discontinuous rock masses. Geotechnique, 35(4): 483495. Reid, M.E. (1997). Slope instability caused by small variations in hydraulic conductivity. Journal of Geotechnical and Geoenvironmental Engineering (ACSE), 123(8): 717725. SEEP/W (Version 4) for finite element seepage analysis. (1998). Geo-slope International, Canada.
This paper examines the effect of principal permeability ratio (k ratio kx/ky) and principal flow orientation () on the PWP distribution in a saprolitic hillslope. A series of parametric study is carried out in an attempt to deduce the hydraulic properties for field measurement. The computed PWPs at colluvium are consistently found to be insensitive to both k-ratio and for all cases. It is probably because the permeability of colluvium is about 2 orders larger than the average infiltration rate, leading to fast water percolation. In contrast, the level of GWT is highly sensitive and can be significantly affected by these two parameters. The GWT level decreases by about 40% when the k-ratio increases from 0 to 10. Conversely, the increase of from 0 to 45 results in about 40% rise of GWT. Based on the current limited numerical analysis, the effect of on transient seepage in a hillslope should not be overlooked. Considering the in situ geological features, a set of k-ratio and is deduced for each material, aiming to mimic the overall seepage characteristic in the hillslope. The computed PWP profiles at all selected days are satisfactorily close to the measured ones. It is thereby postulated that water infiltrated to the CDT stratum at 1 m to 3 m depth would preferentially flow along branches of discontinuities dipping at 30 in average. The ratio of the principal permeability along these preferential flow paths may be as high as 10. Besides, the permeability of the CDT at deeper depth is believed to behave isotropically, as a result of low degree of weathering. ACKNOWLEDGEMENTS The authors would like to thank the research grants OAP06/07.EG01 and HKUST9/CRF-SF/09 provided by Ove Arup and Partners Hong Kong
1260
The significance of curved shear strength envelope in modelling rainfall induced shallow failure of slopes
M.J. Md Noor & B.A. Hadi
ABSTRACT: Shallow type of slope failure triggered by rainfall infiltration is common in tropical countries. The strong influence of shear strength in governing slope failure has been well established. However there are still difficulties to back analyse the shallow slope failure when using conventional slope stability methods. The problem lies on the method of analysis that did not replicate the actual mechanism of failure associated with surface water infiltration and not applying the realistic soil shear strength behaviour. The soil shear strength behaviour is primarily being governed by two independent stress state variables namely net stress, ( ua) and suction, (ua uw). The shear strength behaviour relative to these stress state variables is essentially curvi-linear and non-linear respectively. Apparently the characteristics of the shear strength behaviour at low stress levels and low suctions have a strong influence on the occurrence of the shallow rainfall-induced slope failure. This type of shear strength behaviour can be defined using the curved shear strength envelope. The main advantage of the model is that it exhibits steep drop in shear strength when the soil condition approaches saturation and when net stress approaches zero. The former is much related to the effect of infiltration in reducing the shear strength and the latter is associated with the steep reduction in strength as the depth approaches the slope surface. These aspects of strength are very much associated with the shallow infiltration triggered slope failure. This realistic curved shear strength envelope is applied in a slope stability equation. The equation is formulated in such a way that it can replicate the actual mechanism of rainfall infiltration in reducing the soil strength and increasing the soil weight. Analysis on cut slopes with inclination of 1:1 will be presented to demonstrate the significant of applying the curved shear strength envelope in slope failure investigation involving surface water infiltration. 1 INTRODUCTION intercept was realised for both fine-grained and coarse-grained soil as early as 1966 (Bishop). This is in fact very relevant for the modelling of shallow rainfall induced failure since the stress involved falls within the lower range (100 kPa). Another characteristic of shear strength behaviour relevant to shallow rainfall induced failure is the non-linear steep reduction in shear strength nearing saturation. However these aspects of shear strength behaviour have never been capitalised in slope stability analysis. This may be because of the difficulty of finding the best mathematical representation of the behaviour. Not applying the true shear strength of behaviour is the centre of the problem in understanding the occurrence of the shallow mode of landslide. The rainfall induced landslide must be modelled as the advancement of the wetting front into the slope (Md. Noor et al., 2009) rather than elevating of the groundwater table. In highland area the groundwater is too deep to have influence on any landslide (Brand, 1989). These aspects of shear strength and slope stability modelling have been incorporated in a slope stability computer program in order to study the effect of infiltration on slope stability.
The application of linear shear strength behaviour (Terzaghi, 1936 and Fredlund et al.1978) in slope stability analysis has the problem in modelling shallow (3.0 m deep) mode of rainfall induced failure. This is because the shear strength models do not replicate the actual behaviour at low stress levels and at suctions nearing saturation. In addition, slope stability analysis ignoring the presence of suction in the zone above groundwater table would indicate failure for a standing slope (Othman, 1989). These make the understanding of landslide behaviour very complex. This article is emphasizing on the advantage of applying the real curvi-linear shear strength behaviour relative to effective or net stress (Bishop, 1966; Md. Noor and Anderson, 2006 and Md. Noor et al., 2009) and the non-linear shear strength behaviour relative to suction (Escario and Saez, 1986; Escario and Juca, 1989; Gan and Fredlund, 1996 and Toll et al., 2000) in replicating the shallow mode of infiltration induced landslide. The non-linear shear strength behaviour relative to effective stress at low stress levels with zero cohesion
1261
SLOPE DESCRIPTION AND FAILURE MODE
It has been a standard practice in Malaysia that cut slopes are made at 1:1 inclination while fill slopes are constructed at inclination of 1 vertical: 2 horizontal. A typical cut slope with 1:1 gradient in Malaysia is as shown in Figure 1. The slope is covered with grass and the central portion of the slope has already undergone a localise slip. This article will focus on the stability of such cut slopes. Most of the cut slopes are in highways that cut through hilly terrain that made up of either granitic residual soil or sedimentary formations. These soils are usually very dense or stiff when they are freshly exposed. There is no worry on the short term stability of these slopes. However, once exposed to the atmosphere after the construction they deteriorate or weather very fast and this jeopardise their long term stability. These cut slopes usually undergo shallow slides or normally known as localize failure with depth of failure usually less than 4.0 m (Brand, 1989). This type of shallow slides is very difficult to quantify theoretically and thence they cannot be predicted. As a result this poses danger to road user especially during monsoon season when geotechnical engineers cannot quantify the real state of stability of the slopes with respect to infiltration. However the availability of a new infiltration type slope stability method (Md. Noor, 2007) which deployed the realistic curved shear strength envelope and consider the propagation of the wetting front into the slope can simulate exactly the shallow mode of the infiltration induced failure. With this the real state of stability with respect to infiltration can be determined and by this the safety of the road users can be improved. The problem faced with shallow slope failure triggered by rainfall infiltration in granitic residual soil grade VI in Rawang, Malaysia has been chosen for this study.
SLOPE PROFILE AND SHEAR STRENGTH BEHAVIOUR OF GRANITIC RESIDUAL SOIL GRADE VI
The cut slope is 30 m high and the profile is divided into two layers as shown in Figure 2. Three undisturbed samples were taken from the top layer at depth 1.0 to 2.0 m in granitic residual soil grade VI from Rawang, Malaysia. And consolidated drained triaxial tests were conducted on saturated specimens. The resulted Mohr circles and the shear strength envelope of the soil are interpreted according to curvi-linear shear strength model of Md. Noor and Anderson (2006) and linear shear strength envelope of Fredlund et al., (1978) as shown in Figure 3. Apparently according to the curvi-linear shear strength model the strength of the top and bottom soil undergo non-linear steep drop when effective stress drops from transition effective stress of 140 kPa and 120 kPa respectively. Furthermore, notice that generally the linear shear strength envelope slightly over-estimate the strength on both low and high stress levels for the top layer compared to the curvi-linear shear strength envelope in this
Figure 2. Cut slope profile consisting of top soil layer overlying a denser layer.
Figure 1. Typical cut slope with 1:1 gradient in Malaysia.
Figure 3. Linear and curvi-linear shear strength envelopes of the top and bottom soil layers with respect to effective stress.
1262
specific case. However for the bottom soils the linear shear strength model only over-estimate the strength at the low stress levels. The shear strength envelopes of the bottom soil layer are assumed to be slightly greater than the top overlying soil layer and this assumption is based on a higher SPT N value for the top layer to be less than 10 while for the bottom soil which is greater
than 10 as determined from borehole tests. The linear and non-linear variations of apparent shear strength with respect to suction for both soils are assumed as shown in Figure 4. Notice that for the non-linear behaviour the increase in shear strength is not indefinite but curved towards a maximum value of apparent shear strength. This is the kind of behaviour for residual soils as reported by Gan and Fredlund, (1996) and Toll et al., (2000). At the maximum value the corresponding suction is taken as the residual suction. The bottom soil layer is taken to have higher apparent shear strength and the shear strength parameters according to both types of shear strength models and the soil properties applied in the stability analysis are shown in Table 1. The field condition refers to the infiltration unaffected zone below the wetting front while the wetted condition refers to the zone above the wetting front. 4 STABILITY ANALYSIS TO DETECT SHALLOW LOCALISE FAILURE
Figure 4. Linear and non-linear apparent shear strength behaviour of the top and bottom soil layers with respect to suction.
The stability analysis was carried out using slope stability equation of Md. Noor (2007) as shown in Equation 1
Fos
Table 1. Shear strength parameters of the soil layers according to curved shear strength and plane shear strength envelopes and soil properties. Soil layer Top Bottom
{
} R
dry Vi field 9.81 V field wet 9.81 Vwet xi
(1) where i R try Vi field Vfield wet Vwet xi is the shear strength at the base of slice is the length of the slice base is the radius of the slip circle dry unit weight of soil is the volume of each slice is the field volumetric moisture content is volume of slice having field condition is the wetted volumetric moisture content is volume of slice that has been wetted is the distance of each slice from the centre of rotation
Curved shear strength envelope (Md. Noor and Anderson, 2006) Minimum friction 30 38 angle at failure min f (deg) Transition shear strength, 105 113 t (kN/m2) Transition effective stress, 140 120 ( uw)t (kPa) Residual suction, 120 150 (ua uw)r (kPa) Maximum apparent shear 45 65 strength, csmax (kN/m2) Ultimate suction (kPa) 400 400 16 17 Dry unit wt. try (kN/m3) Saturated or wet 0.2 0.15 volumetric m.c. (%) Field volumetric m.c. (%) 0.1 0.1 Wet suction (kPa) 0 0 Field suction (kPa) 60 80 Plane-envelope shear strength (Fredlund et al., 1978) Internal friction angle, b Cohesion, c (kN/m2) 33 33 12 38 38 20
The numerator is the resisting moment. It depends on the shear strength at the base of slice. The analysis considered two types of shear strength, i which are the curved and the plane envelopes as in Equations 2 and 3 respectively.
uw t uw uw 1 t N uw t uw t ua uw ua uw r ua uw max 1 cs ua uw r ua uw r
(2a)
1263
for; net stress transition net stress and suction residual suction, where N; N 1 tan min f 1 uw t t
uw tan min t uw t tan min
ua uw r ua uw max ua uw 1 cs u u a w r ua uw r
(2b)
for; net stress transition net stress and suction residual suction.
c ua tan ua uw tan b
(3)
Note that since the wetted and the field suctions are less than their respective residual suctions 120 kPa and 150 kPa of the soils then the analysis according to Equations 2(a) and (b) are only considering the changes in shear strength within the residual suctions. 5 EFFECT OF INFILTRATION ON STABILITY
The main objective of the slope stability analysis conducted in this section is to model the shallow failure (4 m) as commonly occurred in the tropical countries triggered by rainfall infiltration. The results of considering the depths of infiltration of 2, 3 and 4 m are shown in Figure 5(a), (b) and (c) respectively. The analysis is conducted through the application of a slope stability computer program which utilizes slope stability Equation 1 as described in the previous section. The analysis apparently shows that the 1:1 cut slope with the described shear strengths is not stable with respect to infiltration base on the slope stability analysis curved shear strength envelope of Md. Noor and Anderson (2006). This explains why the slope of this kind fails under prolong rainfall period which permits the wetting to propagate deeper. Nevertheless some slopes may not fail at 1.0 m infiltration. This is due to the extra strength of the top soil provided by the reinforcement by the grass roots. This is the main advantage of applying the real curved shear strength envelope where the actual state of stability can be determined and this allows for early preventive action and this can improve the public safety. However the application of the linear-plane type shear strength envelope of Fredlund et al., (1978) does not indicate failure even up to 4.0 m depth of infiltration. Nevertheless, it shows the reduction in the stability factor with the increase in depth of infiltration.
Figure 5. Stability analysis for depths of infiltration of 1, 2, 3 and 4 m.
Figure 6. Conventional slope stability analysis for rainfall induced failure by elevating groundwater table.
1264
If the shallow rainfall infiltration induced landslide is to be modelled by elevating the groundwater table, the result is shown in Figure 6. Evidently the groundwater table has to be very close to the slope surface in order to achieve failure condition. However the analysis using plane type shear strength model of Fredlund et al., (1978) still does not indicate failure. Nevertheless this is not the actual condition of how rainfall triggered landslide (Brand, 1989). 6 CONCLUSIONS
The main conclusion that can be deduced from this study is that in determining the state of stability with respect to rainfall infiltration, inevitable the real curved shear strength envelope behaviour needs to be applied. Otherwise the shallow type of rainfall induced failure that may occur during prolong heavy rainfall cannot be determined and realised. The method of elevating the groundwater table is not really replicating the field condition. REFERENCES
Bishop, A.W. (1966). The strength of soils as engineering materials. Geotechnique, 16(2), 91130. Brand, E.W. 1989. Occurance and Significant of Landslides in Southeast Asia. International Geological Congress: Symposium on Landslides, Washington D.C. Escario, V. & Juca, J. (1989). Strength and deformation of partly saturated soils. 12th. International Conference on Soil Mechanics and Foundation Engineering, Rio de Janeiro, 3, 4346.
Escario, V. & Saez, J. (1986). The shear strength of partly saturated soils. Geotechnique, 36(3), 453456. Fredlund, D.G., Morgenstern, N.R. & Widger R. (1978). The shear strength of unsaturated soils. Canadian Geotechnical Journal, 15(3), pp. 313322. Gan, J.K.M. & Fredlund, D.G. (1996). Shear strength characteristics of two saprolitic soils. Canadian Geotechnical Journal, 33, 595609. Md. Noor, M.J. & Anderson, W.F. (2006). A comprehensive shear strength model for saturated and unsaturated soils Proc. 4th Int. Conf. on Unsaturated Soils, ASCE Geotechnical Special Publication No. 147, Carefree, Arizona, Vol. 2, pp 19922003 ISBN 0-7844-0802-5. Md. Noor, M.J. (2007). Slope Stability Method Incorporating the Curved Surface Envelope Shear Strength Model. Proc. 16th South East Asian Geotechnical Conference, Kuala Lumpur, Malaysia. Md. Noor, Mohamed Jais M.J., I.B. & Hafez M.A. (2009). Modeling of shallow slope failure using curvedsurface envelope extended Mohr-Coulomb model. 4th Asia Pacific Conference on Unsaturated Soils, Newcastle, Australia, 23th25th November 2009. Othman, M.A. (1989). Highway cut slope instability problems in West Malaysia. Ph.D. Thesis, Department of Geography, University of Bristol, United Kingdom. Terzaghi, K. (1936). The shear resistance of saturated soils. Proceedings for the 1st. International Conference on Soil Mechanics and Foundation Engineering (Cambridge, MA), 1, 5456. Toll, D. G., Ong, B. H. & Raharjo, H. (2000). Triaxial testing of unsaturated samples of undisturbed residual soil from Singapore. Proceedings of the Conference on Unsaturated Soils for Asia, Singapore, Balkema, 581586.
1265
The influence on the slope stability of the initial degree of saturation of a fill ground
T. Mori, R. Uzuoka & M. Kazama
Department of Civil Engineering, Graduate School of Engineering, Tohoku University, Japan
ABSTRACT: This paper presents a case study on a numerical prediction of seepage and seismic behaviors of an unsaturated fill slope. In this study, the numerical prediction is performed by using a dynamic threephase (soil, water and air) coupled analysis based on porous media theory and constitutive models. In-situ observation at the target fill slope of the case study has been performed since 2006. The ground water level, the soil moisture and the water quantity gushed up from the toe of the fill slope are monitored continuously. These observation data were used for verification of the analysis. First, distributions of water saturation in the fill slope are simulated by performing seepage analyses to reproduce the in-situ ground water level in the fill slope. Next, seismic responses of the fill slope during a future scenario earthquake are predicted under different moisture conditions of unsaturated fill. 1 INTRODUCTION scenario earthquake are predicted under different initial moisture conditions of unsaturated fill. 2 2.1 NUMERICAL METHOD Balance and constitutive equations
Fill slopes located on old valleys have been damaged due to the high water content in the fill during earthquakes. Moreover the capillary zone in a fill slope composed of a volcanic soil with high water retention is usually thick; therefore it possibly liquefies during an earthquake (Uzuoka et al. 2005). Recently liquefaction mechanism of unsaturated soil has been discussed (e.g. Okamura & Soga 2006, Unno et al. 2008) and it is suggested that the behaviors of pore air and suction play an important role during liquefaction of unsaturated soil (Kazama et al. 2006). This study presents a case study on a numerical prediction of seepage and seismic behaviors of an unsaturated fill slope. In this study, the numerical prediction is performed by using a dynamic threephase (soil, water and air) coupled analysis based on porous media theory and constitutive models. A simplified Soil Water Characteristic Curve (SWCC) model with logistic function is proposed to reproduce the wetting process during cyclic loading. In-situ observation at the target fill slope of the case study has been also performed since 2006. In-situ tests and laboratory tests with disturbed soil samples are conducted before the predictions. The material parameters of the constitutive models are determined through the calibration of the laboratory tests. Firstly, distributions of water saturation in the fill slope are simulated by performing seepage analyses to reproduce the in-situ ground water level in the fill slope. Next, seismic responses of the fill slope during a future
The basic equations are derived based on porous media theory (e.g. de Boer 2000, Schrefler 2002). The motion of soil skeleton is expressed with Lagrange description, and the motion of pore water and pore air is expressed with Euler description. We derive simplified governing equations which include the momentum balance equations of the overall three-phase material and the mass and momentum balance equations (continuity equations) of the pore water and air with the following assumptions. 1) An isothermal condition is assumed, 2) the soil particle is incompressible, 3) the mass exchange among phases is neglected, and 4) the material time derivative of relative velocities and advection terms of pore fluids to the soil skeleton are neglected. The momentum balance equations of the overall three-phase material is derived as
a s div ( s w pw s a pa )I b
(1)
where is the overall density of three-phase material, as is the acceleration vector of soil skeleton, is the skeleton stress tensor, sw is the degree of saturation, pw is the pore water pressure, pa is the pore air pressure, b is the body force vector. The stress and the pressures are defined as positive in tension and compression respectively.
1267
The mass and momentum balance equations of the pore water is derived as ns w wR D s pw D s pa n wRc n wRc w Dt K Dt s w wR div v s ws k div grad pw wR b wRa s 0 (2) g
g M m ln p / pa 0
(6)
where Mm is the material parameter which defines the critical state ratio, pa is p when 0. Finally the elastic module are assumed as K e K * p G e G * p
e e
(7)
where n is the porosity, wR is the real densities of pore water, Kw is the bulk module of pore water, c is the specific water capacity, Ds /Dt is the material time derivative with respect to soil skeleton, vs is the velocity of soil skeleton, kws is the coefficient of water permeability, g is the gravity acceleration. The mass and momentum balance equations of the air is derived as
s a s w ns a aR D p aR D p R n c Dt n c Dt s a aR div v s k as div grad pa aR b aRa s g
where K is the elastic bulk modulus, G is the elastic shear modulus, K * and G* are the dimensionless elastic module respectively. 3 3.1 THE FILL SLOPE AND MATERIAL PARAMETER The fill slope for a case study
0.
(3)
where aR is the real densities of pore air, is the absolute temperature, R is the specific gas constant of air, k as is the coefficient of air permeability. This simplified formulation is called u-pw-pa formulation. Although the governing equations are derived in the regime of finite strain, we assume infinitesimal strain in the following study for simplicity. 2.2 Constitutive equation for skeleton stress
A simplified constitutive model for a saturated sandy soil is used for an unsaturated soil with using a skeleton stress (e.g. Gallipoli et al. 2003) in place of an effective stress of a saturated soil. Assuming that plastic deformation occurs only when the deviatoric stress ratio changes, the yield function is assumed as f k s / p k 0 (4)
The fill slope is located in a housing area developed on hill ground of northeastern Sendai City in the 1970s. Figure 1 shows a comparison of the present topographical map around the fill slope and a topographical map made in 1966 before the development. An estimated cross sectional view is presented in Fig. 3. This fill slope was buried with crushed material of peripheral natural ground (tuff). The fill slope height is about 25 meters and its maximum thickness is about 17 meters. The N-value in the original ground area is greater than 50. The N-value in the fill part is roughly less than five. Four seismographs were installed in the fill slope and around ground. The positions of measuring instruments are portrayed in Figure 2 and Figure 3. The seismographs, designated as E1, E2, E3, and E4 were respectively installed in the upper natural ground, in the top of the fill (fill thickness is 17 m), in the middle of the fill (fill thickness is 8 m), and in the lower natural ground. In addition, a pore water pressure gauge (G.L. 9.25 m), two water content gauges (G.L. 2.0 m and 4.0 m) and a rain gauge were installed around the location of seismograph E3 in the middle of the fill. 3.2 Material parameters
where p ( tr()) is the mean skeleton stress, s is the deviatoric stress tensor, k is the material parameter which defines the elastic region. is the kinematic hardening parameter (back stress) and its nonlinear evolution rule (Armstrong & Frederick 1966) is assumed as a be sp p
The fill material is silty sand. The physical and mechanical properties are investigated with in-situ
sp e p
(5)
where a, b are the material parameters, e p is the plastic deviatoric strain rate tensor. With non-associated flow rule, the plastic potential function is assumed as
Figure 1.
Topographies in 1966 and 1993.
1268
Figure 2.
Plane figure.
Figure 4.
with simple scaling method in Figure 4 (b). The modified logistic SWCC function is derived as
w se w w c 1 se 1 se c 0 0 lg 1 1 exp( a p b ) lg a lg w w 1 se 1 se 1 1 c c p1 p0 c c c c pa c c ( p p0 ) p0 c p0 p0 p2 (8)
Figure 3.
Cross section.
and laboratory tests. The material parameters of the constitutive models for skeleton stress and SWCC are determined through the calibration of the undrained cyclic triaxial tests with unsaturated soil. The detailed description of the testing method is referred to Unno et al. (2008). The specimen was made of silty sand obtained from the in-situ fill. The initial dry density of the specimen was about 1.2 g/cm3. The effective degree of water saturation was from about 23% to 60% by controlling air pressure during the isotropic consolidation process. The pore water pressure was almost zero after the consolidation and the pore air pressure increased with the decrease in water saturation. The net stress was about 45 kPa for all specimens and the mean skeleton stress varied with the initial suction dependent on initial water saturation. The cyclic shear was applied to the specimen under undrained air and water conditions. The input axial strain was the sinusoidal wave with multi-step amplitudes whose single amplitudes were 0.2, 0.4, 0.8, 1.2, 1.6, and 2.0 with every ten cycles. The frequency of the sinusoidal wave was 0.005 Hz. This loading rate is slow enough to achieve an equilibrium condition between air and water pressure. Figure 4 (a) shows SWCC during the undrained cyclic triaxial tests. The measured SWCC are not on the main drying and wetting curves obtained from water retention tests; therefore the scanning curves should be used to reproduce the SWCC during undrained cyclic shear. The logistic SWCC function is modified to reproduce the scanning curves
where pc0 and swe0 are the initial suction and effective water saturation respectively, pc1 ( pc0) and swe1 are the suction and effective water saturation on main wetting curve respectively and pc2 is the suction on the scanning curve at swe 1. The modified logistic SWCC roughly reproduce the measured SWCC during undrained shear in Figure 4 (a). Further investigation on SWCC during undrained shear is necessary for more precise reproduction. In the triaxial test simulations, the finite element formulation presented in the previous section is not used. Assuming that the variables in the specimen are homogeneous, only the local equilibrium is considered. Table 1 shows the calibrated material parameters of the constitutive model. Figure 5 shows the time histories of pore water pressure, pore air pressure, suction, mean skeleton stress (positive in compression) and void ratio from tests and simulations in the case with the initial effective water saturation of 46%. In the test results (denoted Test in the figures), the pore water and air pressure increase, while the suction and mean skeleton stress decrease during cyclic undrained shear. In this case, the suction and the mean skeleton stress do not attain zero, which means that the specimen do not liquefy completely. In the simulated results (denoted Model in the figures), the model well reproduces the overall tendency of the test results. Therefore, the simplified constitutive equation of soil skeleton can be applied to predict pore water and air responses of unsaturated soil in the framework of three-phase porous media theory. The modifications of the constitutive equations are necessary for more precise reproduction.
1269
Table 1. Material parameters of silty sand from the fill. Elasto-plastic model parameters Dimensionless shear modulus, G * Dimensionless bulk modulus, K * Nonlinear hardening parameter, a Nonlinear hardening parameter, b Critical state stress ratio, Mm Yield function parameter, k SWCC parameters Maximum degree of saturation, sws Minimum degree of saturation, swr Main drying curve, alg, blg, clg Main wetting curve, alg, blg, clg Scaling parameter, pc2 (kPa) Physical parameters of water and air Bulk modulus of water, K w (kPa) Real density of air, aR (t/m3) Gas parameter, 1/(R ) (s2/m2) 140 600 5512 1.8 1.8 0.0245 0.99 0.35 1.5, 3.0, 0.05 0.25, 2.5, 0.2 15 1.0 106 1.23 103 1.25 105
Figure 6. Table 2.
Cross section of the fill. Material parameters for seismic analysis. Rock Fill 0.57 2.59 1.00 1.0 107 1.0 108 3.0/0.05 127351 31838 0.001 Clay 0.64 2.72 1.00 1.0 1010 1.0 1011 99915 11102 0.001 Air 1.00 0.00 1.00 10.0 10.0 0.01 0.001 0.001
n sR (t/m3) wR (t/m3) kws (m/s) kas (m/s) k /k (kPa) (kPa)
0.37 2.72 1.00 1.0 1012 1.0 1013 293876 293877 0.001
Figure 5. Test and simulation with initial effective water saturation of 46%.
4 4.1
SEEPAGE ANALYSIS Numerical data of seepage analysis
Figure 6 shows the cross section for seepage and seismic analysis. The position of cross section is shown in Figure 1. Assuming a plane strain condition, the cross section is used for finite element modeling. In the seepage analyses, all materials are linear elastic materials with Lames coefficients and in Tables 2. The porosity and density of the fill was obtained from in-situ soil sample. The coefficient of water permeability of the fill was obtained from permeation velocity of rainfall based on in-situ observation and laboratory tests. Other parameters (rock and clay) were decided to refer to the report by Sendai City (Sendai City 2002).
Two typical cases are considered in this study. Case 1 is the case with low ground water table after light precipitation and Case 2 is the case with high groundwater table after heavy precipitation. Rainfall is applied on the top boundary in the seepage analyses. Because of difficulties to reproduce the actual rainfall history since distant past, the precipitation in Case 1 is assumed as 175 mm/year. This amount of infiltration roughly corresponds to the amount of measured flux at the toe of the fill when it has not rained for about a week before the measurement. While the rainfall strength in Case 2 is 236 mm/year which is determined to reproduce the actual measured ground water level when it has heavy rain. Smooth infiltration of rain on the surface of the fill is simulated with aerial elements which have a special SWCC and large permeability of water in vertical direction (Uzuoka et al. 2008). Use of the aerial elements makes no numerical treatment (e.g. switching between natural and basic boundaries) on the surface. The soil displacement at the bottom boundary is fixed in all directions and the lateral boundaries are vertical rollers. The bottom and lateral boundaries are impermeable and a part of the surface on the clay layer nearby the pond is permeable with zero water pressure. The seepage analyses are quasistatic analyses under the gravity force without the inertia terms. A large calculation time is adopted to obtain the steady state conditions of water saturation. The coefficient in Newmark integration method is 1.0 and the time increments are determined to ensure the numerical convergence. 4.2 Numerical results of seepage analysis
Figure 7 shows the distributions of water saturation after the precipitation in both cases after the
1270
Figure 8. Comparison of water saturation (middle of the fill slope).
reproduced the ground water level to assume that the fill ground is homogeneous. Further studies on soil water characteristics in a heterogeneous ground are necessary to reproduce the actual distribution of water saturation in the unsaturated soil. 5
Figure 7. Distribution of water saturation before an earth quake (fill slope area).
PREDICTIONS OF SEISMIC RESPONSE WITH DIFFERENT DISTRIBUTION OF WATER SATURATION Numerical conditions
5.1
steady state is achieved. Figure 7 is the enlarged figure around the part of fill slope. The most upper figure (initial state) in Figures 7 shows the initial distributions of water saturation without rainfalls. The distributions show the balanced state of soil moisture which is obtained from applying free water surface on the pond of the finite element model. Because soil moisture in the fill depends on the suck from free water surface, most part of the fill is dry. Constant rain fall is applied to the initial state and then the water saturation gradually rises up. According to applying rainfall, the water quantity of input and output becomes to be balance and the distribution of water saturation becomes a steady state. We define this state as the initial distribution of water saturation in the fill before an earthquake in Fig. 7 (Case 1 and Case 2). Figure 8 shows the comparison between the actual observed and computed results of water saturation in the fill. The comparison of water saturation is performed at the middle of the fill slope (refer to Fig. 7). In both cases 1 and 2, the ground water level is well reproduced. However, the computed water saturation in the unsaturated part above the ground water level is smaller than the actual measured value. In the numerical analysis, the fill ground is assumed to be homogeneous and the moisture distribution is expressed by the continuous soil water characteristic curve which starts from ground water level. Meanwhile the actual fill is not homogeneous and contains air spaces and soil blocks everywhere. Therefore the moisture distribution of the actual fill doesnt follow the continuous soil water characteristic curve. In this study, we
The numerical conditions are the same conditions mentioned in Section 4 except for the initial conditions. The two initial distributions of water saturation in the fill with Case 1 and Case 2 are considered in the predictions (refer to fig. 7). The difference of ground water level between Case 1 and Case 2 is about 1.5 meters at the middle of the slope. The calculated scenario earthquake wave is used for the input wave. This calculated wave is obtained by a numerical simulation of future earthquake (Sendai city. 2002). The time history of input acceleration is shown in figure 9. 5.2 Numerical results
Figure 9 shows the acceleration responses at the top of the slope with different initial distribution of water saturation. Until about 9 seconds, the difference in the acceleration between the two cases is little; therefore the influence of initial distribution of water saturation is not significant. However, the differences of acceleration responses appear after about 9 seconds. The maximum acceleration responses in the Case 1 and Case 2 are 453.0 cm/s2 and 325.0 cm/ s2 respectively. The maximum acceleration response in the Case 1 is larger than that in the Case 2. Figure 10 shows the distributions of Skeleton Stress Reduction Ratio (SSRR hereinafter) in Case 1 and Case 2. The SSRR is defined as 1 p/p0 where p0 is the initial value of p. In Case 1, the SSRR increases from central fill and the area with large SSRR expand to the around area. The distribution of SSRR in the unsaturated parts (above ground water level) becomes dappled.
1271
CONCLUSIONS
Following findings were obtained. 1. The distributions of water saturation were simulated with the seepage analyses with artificial rainfalls. The observed groundwater level was reproduced while the distribution of water saturation above the ground water level was not reproduced. Further studies on soil water characteristics in a heterogeneous ground are necessary to reproduce the actual distribution of water saturation in the unsaturated soil. 2. The seismic response analyses of the unsaturated fill slope which had different distributions of water saturation were performed. The numerical results clearly show that skeleton stress and the residual displacement of the fill slope are affected by the initial distribution of water saturation in the fill ground. REFERENCES
Armstrong, P.J. & Frederick, C.O. 1966. A mathematical representation of the multiaxial Bauschinger effect, C.E.G.B. Report RD/B/N731, Berkeley Nuclear Laboratories, Berkeley, UK. de Boer, R. 2000. Contemporary progress in porous media theory, Applied Mechanics Reviews, 53(12): 323369. Gallipoli, D. et al. 2003. An elasto-plastic model for unsaturated soil incorporating the effects of suction and degree of saturation on mechanical behaviour, Geotechnique, 53(1): 123135. Kazama, M. et al. 2006. Liquefaction mechanism of unsaturated volcanic sandy soils, Journal of Geotechnical Engineering, Japan Society of Civil Engineers, 62(2): 546561 (in Japanese). Okamura, M. & Soga, Y. 2006. Effects of pore fluid compressibility on liquefaction resistance of partially saturated sand, Soils and Foundations, 46(5): 93104. Schrefler, B.A. 2002. Mechanics and thermodynamics of saturated/unsaturated porous materials and quantitative solutions, Applied Mechanics Reviews, 55(4): 351388. Sendai City. 2002. Report on earthquake damage investigation (in Japanese). Unno, T. et al. 2008. Liquefaction of unsaturated sand considering the pore air pressure and volume compressibility of the soil particle skeleton, Soils and Foundations, 48(1): 8799. Uzuoka, R. et al. 2005. Landslides during the earthquake on May 26 and July 26, 2003 in Miyagi, Japan, Soils and Foundations, 45(4): 149163. Uzuoka, R. et al. 2007. Finite element analyses for dynamic finite deformation in unsaturated soil, Proc. the Conf. on Computational Engineering and Science, 13: 523526 (in Japanese).
Figure 9.
Time history of acceleration response.
Figure 10. Distribution of skeleton stress reduction ratio.
In Case 2, the SSRR above the ground water level increases until about 8 seconds when the SSRR is less than 1.0 at the bottom of the fill. However, the SSRR above the ground water level changes little after about 10 seconds when the SSRR reaches 1.0 at the bottom of the fill. These phenomena are due to the softening of saturated soil and the shear stress does not propagate to the unsaturated part above the ground water level. The SSRR in the central fill is larger than that around the toe of the fill. The residual horizontal displacement at the top of the slope in Case 1 and Case 2 are about 0.3 meters and about 0.9 meters respectively.
1272
Test site experience on mechanisms triggering mudflows in unsaturated pyroclastic soils in southern Italy
M. Pirone, R. Papa & M.V. Nicotera
Dipartimento di Ingegneria Geotecnica, Universit Federico II, Napoli, Italy
ABSTRACT: The paper describes the results of an ongoing experimental research project on mudflows. Based on geological and geomorphologic considerations, the test site of Monteforte Irpino was selected to collect experimental data through laboratory testing on undisturbed samples recovered in situ and by field monitoring of climatic conditions, matric suction and water content. The site instrumentation consisted of 94 tensiometers, 40 TDR probes, six Casagrande piezometers and a weather station. To date, the collected suction data span about three years of observations, those of water content about a year and a half. They allow clear identification of seasonal variations in matric suction and water content and the hysteretic behaviour of the pyroclastic soils. Site monitoring results and analysis of water flow in the subsoil are presented, and some considerations are made on the hydraulic behaviour of the pumices interposed between the pyroclastic layers. 1 INTRODUCTION some important additional understanding about the water flow in this layer is also achieved. 2 TEST SITE & INSTRUMENTATION DESIGN
In the last years a series of catastrophic mud-flows, involving pyroclastic unsaturated soils, has caused severe damage and a number of fatalities in the region of Campania (Italy). Mudflows affect the pyroclastic cover resting on the limestone massif in the region. Rainwater infiltration is the likely mechanism that leads or predisposes to slope failures by reducing matric suction in unsaturated soils, thereby reducing the shear strength. Although local factors could be responsible for landslides, it is the general condition of potential instability, produced by rain infiltration from ground surface, which predisposes the subsoil to failure. Based on geological and geomorphological considerations, a test site was selected to collect experimental data from laboratory tests on undisturbed samples, monitoring climatic conditions (affecting infiltration), matric suction and water content in the subsoil. In this paper suction and volumetric water content measurements are shown. In particular, seasonal variations in matric suction and water content in the different layers of pyroclastic cover are detected by averaging the measurements of all instruments situated at the same depth. Matric suction measurements from tensiometers arranged along the longitudinal section B-B of the instrumented area are used to obtain the groundwater head field in the cover. Vertical water flows through the top and the intermediate part of the soil profile are estimated. Water mass balance in pumice layer 3 is performed and
The test site is situated on the west side of a limestone relief called Monte Faggeto about 40 km northwest of the volcano Somma-Vesuvius. The pyroclastic unsaturated soil cover (56 meters thick) on the limestone massif consists of the products of a series of eruptions of Somma-Vesuvius rests (Di Crescenzo et al., 2007). The morphological and stratigraphic features of the test site and the instrumentation design have been described extensively elsewhere (Papa, 2007; Nicotera et al., 2008). Hence the stratigraphy and instrumentation design is only briefly reported herein. A series of soil layers essentially parallel to the ground surface are present in situ. However, a simplified profile was derived (Figure 1), based on experimental investigation of the geotechnical properties of the pyroclastic cover (Papa, 2007). The field instrumentation was designed to measure matric suction and water content in the pyroclastic cover and the climatic conditions as well. An area of about 230 m2 was chosen on the slope. Twenty instrumented vertical sections were set up along three longitudinal alignments (Papa, 2007). The tensiometers and sensor probes were installed in the top part of the soil profile (superficial layers 1 & 2), the intermediate part (layer 4) and
1273
TENSIOMETER TDR PIEZOMETER
sections: AA ; C-C B-B
depth(m) 1-2: topsoil & ashysoil (0.000.80m) 3: pumiceous deposits of Avellino eruption 3.7 ky b.p.
(0.801.20m)
0.25 0.45
4: palaeosoil
(weathered volcanic ashes)(1.202.80m)
1.50 1.60 1.70
5: pumiceous deposits of Ottaviano eruption8.0 ky b.p.

(2.803.50m)
6: palaeosoil
(weathered volcanic ashes)(3.504.40m)
4.00
7: volcanic sand (4.404.90m) 8: highly weathered fine-grained ashy soil (4.905.50m) Fractured limestone
5.00 5.50
Figure 1. Instrumentation design of the test site: vertical distribution of sensors.
the bottom part (deep layers 6,7 and 8) (Nicotera et al., 2008). However, tensiometers and TDR probes were judged ineffective for measuring matric suction and water content inside the pumiceous strata (layers 3 and 5). Four tensiometers were installed inside the superficial and intermediate soil layers (see Figure 1) for all the instrumented vertical sections; two more tensiometers in the deeper soil layers and six TDR probes (two probes for each layer apart from the pumiceous ones) were installed only in the vertical sections arranged along the central longitudinal alignment. Furthermore, in each of these vertical sections a piezometer at the bottom of the pyroclastic cover was set up to measure any positive pore water pressure at the limestone upper surface (Figure 1). Finally, a weather station was installed to monitor the climate conditions affecting the pore water pressure field in the soil cover (i.e. rainfall, solar net radiation, soil temperature, air temperature, air pressure, air humidity, wind speed, etc.).
MONITORING RESULTS
Measurements for suction monitoring started on 13th October 2005 and were collected weekly by means of portable quick draw tensiometers at a number of points inside the instrumented area at depths of 0.30 cm and 0.60 cm (Fig. 2, QD1, QD2). These measurements proceeded until summer 2006. Tensiometers in layers 1 & 2 and in layer 4 were installed in mid-autumn 2006 and monitoring continued by means of these instruments. The tensiometers in layers 6 and 8 were installed later and the data were only available from 18th July 2008. In Figure 2 some of the monitoring results are presented. Figure 2a reports daily rainfall registered
by the Monteforte rain-gauge station (502 m a.s.l.), and the records from another rain-gauge installed at the site (605 m a.s.l.). The first measurements are available over the whole monitoring period, while the second are available only from spring 2008 to summer 2008 and from summer 2009. Figure 2b represents matric suction measurements: as regards preliminary monitoring carried out using the portable tensiometers, the data collected are reported as mean measurements taken at two different depths (0.3 m and 0.6 m). Analogously, the mean measurements obtained with the tensiometers are presented as mean measurements carried out at similar depths. The data collected during preliminary monitoring with the portable tensiometers are in good agreement with subsequent measurements. Data reported in Figure 2b clearly show the seasonal oscillations in the measurements at different depths. Matric suction appears almost constant in winter while it increases in spring and summer; maximum suction values are measured by shallower tensiometers (TL1 and TL2) in summer, sometimes exceeding the upper limit of the tensiometers operation range (about 70 kPa). Finally, matric suction progressively decreases as the wet season begins. In the superficial layers (see Fig. 2.2, TL1, TL2) the suction measurements are heavily influenced by a single rainy event. However, seasonal variations generally play a major rule. Conversely, the suction measurements in the intermediate and deeper layers are unaffected by individual rainfall events. Furthermore, it must be observed that the seasonal cycle in the intermediate and deep part of the cover is quite delayed and the amplitude of seasonal obscillation, especially in soils 6 and 8, is smaller than that in the superficial layers. Volumetric water content monitoring started on 30th April 2008, the measurements being collected with TDR probes. The measurements concern the verticals of the central longitudinal section and the probes are installed at the same depth as the tensiometers. In Figure 3b and 3c the mean suction and mean volumetric water content measurements carried out at the same depth in soils 1, 2 and 4 are plotted against time. Whereas in the superficial layers (Fig. 3c, TDR1, TDR2) the water content measurements are influenced by single rainy days, seasonal variation plays a major role with suction measurements collected at the same depth. In the intermediate layers (Fig. 3c, TDR4 sup, TDR4 inf) measurements are not influenced by a single rainy day just as with suction measurements collected at the same depth. Hence, good agreement can be observed (Figure 3b and 3c) between the suction and volumetric water content measurements: both are characterized by a seasonal trend and show an amplitude of seasonal oscillation function of the distance from the soil surface.
1274
daily rainfall (mm/day)
120 100 80 60 40 20 0 70
Quick draw tensiometers TL1 z=0.20-0.25 m TL2 z=0.40-0.45 m TL4sup z=1.30-2.20 m TL4inf z=1.35-2.30 m TL6 z=3.10-3.50 m TL8 z= 3.50-3.90 m QD1 z=0.30 m QD2 z=0.60 m
Monteforte Rain Gauge Site Rain Gauge
a)
Jet Fill and SDEC France Tensiometers
matric suction(kPa)
b)
60 50 40 30 20 10 0
O N
M M
A M MJ J
JA A
O N
A M J J
JA A
O N
A MJ J
JA A
O N
F M
J A
2005
2006
2007
2008
2009
Figure 2.
Main monitoring results: a) Site raingauge and Monteforte raingauge readings; b) mean matric suction.
120 100 80 60 40 20 Monteforte Rain Gauge Site Rain Gauge
a)
daily rainfall (mm/day) matric suction(kPa)
80 70 60 50 40 30 20 10 0 0.60
TL1 TL2 TL4inf z=0.20-0.25 m z=0.40-0.45 m z=1.35-2.30 m
TL4sup z=1.30-2.20 m
b)
Vertical water flow(mm/day) Volumetric water content
0.50 0.40 0.30 0.20 0.10

8.0 6.0 4.0 2.0 0.0 -2.0 -4.0 -6.0
q2 soil 1-2 q2 soil 4 q1 soil 1-2 q1 soil 4
TDR1 TDR2
z=0.20m z=0.40m
TDR4sup z=1.35-2.20m TDR4inf z=1.45-2.30m
c)
d)
Cumulated water flow(mm)
2900 2400 1900 1400 900 400 -100

A M JJ J J A S O N D JA J F M A M J J A S
Cumulated Rainfall Cumulated flow soil 1-2 Cumulated flow soil 4
e)
2008
2009
Figure 3. Main monitoring results: a) Site raingauge readings and Monteforte raingauge readings; b) mean matric suction; c) mean water content; d) vertical water flow derived from measurements; e) total vertical water flow and cumulative rainfall.
1275
0 .8 0.7 0.6
volumetric water content
2
Suction measures in this area have been corrected as explained in the text
0.8 0.7 volumetric water content 0.6 0.5 0.4 0.3 0.2
lab drying retention curve soil 4
4
0.5 0.4 0.3 0.2
0.1
a)
0.1
suction-water content suction-water content
measurements z=0.25m measurements z=0.45m
0.1
suction-water content measurements z=1.35-2.20m suction-water content measurements z=1.45-2.30m
b)
0
0.1 1
10 100 matric suction (kPa)
1000
matric suction (kPa)
10
100
1000
Figure 4. Main monitoring results: matric suction and volumetric water content measurements and the lab drying retention curves (Papa, 2007) a) for soils 12; b) for soil 4.
In Figure 4a and 4b the water content and suction measurements collected in all the verticals instrumented along the central section B-B and the dry retention curves obtained in the lab (Papa, 2007) are plotted for the superficial layers (12) and the intermediate layers (4) (Nicotera et al., 2008). The paths described in the water retention plane by the data points collected in situ from 30th April 2008 were always below the drying retention curves but inside the expected hysteresis domain (Fig. 4a, b). It is worth to noticing that during the drier seasons some of the water retention paths bended towards the suction axis indicating that the in situ suction value exceeded the upper boundary of the tensiometer measurement range (i.e. 7080 kPa); however in these cases the suction values have been corrected by simply assuming as suction value the one corresponding to the measured water content on the main drying water retention curve (see Fig. 4a). 4 ANALYSIS OF THE GROUNDWATER FLOW
The total fluxes of water filtrating vertically into the upper and intermediate parts of the soil cover were estimated on the basis of both monitoring and lab data. Concurrent readings of the tensiometers installed along the same vertical section were used to estimate the hydraulic gradients in each vertex of the instrumented grid at two different depths (i.e. between tensiometers TL1 and TL2 and between tensiometers TL3 and TL4). The parameters of the Mualem-van Genuchten model for water retention and hydraulic conductivity were derived from lab tests on undisturbed samples recovered in each soil layer (Papa, 2007). These parameters were determined during drying tests; however, according to Topp and Miller (1966) the relationship between volumetric water
content and water permeability was assumed to be non-hysteretic. Therefore a preliminary estimate of the hydraulic conductivity corresponding to each of the above gradients was derived by generating the main-wetting retention curve from the main experimental drying curve (Papa, 2007) by simply doubling the value of parameter , as suggested in the literature (Kool and Parker, 1987, Nielsen and Luckner, 1992). Hence the value of the permeability was obtained from the value of the volumetric water content corresponding to the measured matric suction value along the estimated main wetting curve. Finally, the flows (see Fig. 3d, q1) calculated by applying Darcys law were integrated over the whole instrumented area. Nevertheless, when the in situ measurements of water content became available a second more reliable estimate of the hydraulic conductivity was obtained by directly calculating permeability as a function of the volumetric water content on the basis of the lab test results. The flows (see q2 Fig. 3d) calculated by applying Darcys law were integrated over the central longitudinal section of the site. The results of these analyses are reviewed in Figure 3d, 3e. In Figure 3d the mean calculated fluxes (positive when directed downward) are represented while in Figure 3e the cumulative flows Q2 and the cumulative rainfall recorded at the Monteforte rain-gauge stations are plotted. The vertical water flows q2 (continuous line in Fig. 3d) are compared with the flows q1 (dotted line in Fig. 3d). It should be noted that the infiltration and evaporation peaks of q2 are smaller than those of q1. Indeed, the hydraulic conductivity used to calculate q2 is a function of the volumetric water content measured in situ and is smaller than that obtained using the permeability curve in drying conditions with double the values of . However, the hydraulic conductivity calculated as a function of water content measurement is
1276
more reliable because it takes into account the actual paths described by the suction and water content data in the water retention plane. Vertical water flow q2 (Fig. 3d) in the superficial layers is positive (infiltration) throughout the year apart from summer when the outward flow reaches values around 1 mm/day, and its intensity is strongly correlated to rainfall. Vertical water flow q2, in the intermediate layers is negative (evaporation) (Fig. 3d) from spring to summer and assumes an absolute value of around 0.51.0 mm/day. By evaluating the total flows (Fig. 3e), most of the rainfall does not infiltrate into the superficial layers but is lost in run-off or intercepted by vegetation. Cumulative water flow over one year through the superficial soil is not nil but the inward flows (due to infiltration) exceed the outward flows (due to evaporation). Just about 15% of the cumulated vertical flow in the superficial layers reaches the intermediate part of the soil cover. In these layers the cumulative flow over one year is almost nil, hence the inward flows (produced by infiltration) balances the outward flows (generated by evaporation phenomena at the soil surface). The mean value of the modulus and mean direction of the water flow vectors inside each layer were determined (Fig. 5a, b). The groundwater head field along the central longitudinal alignment of the instrumented area was estimated by a spatially interpolation of the available matric suction measurements; hence the groundwater head gradients were derived from the estimated field. The hydraulic permeability in each point of the analysed section was derived by estimating the soil water content along a scanning curve contained in the water retention hysteresis domain. The modulus of the groundwater flow vectors in each point of the interpolation grid were finally calculated by
Direction of the water flows() Water flow modulus(mm/day)
8 7 6 5 4 3 2 1 0
applying Darcys law. The mean values of modulus were calculated by integrating the water flows over each of layers 1, 2 and 4 (Fig. 5a). In Figure 5b the angle between the vertical direction and the mean direction of the groundwater flow vectors is diagrammed as a function of time t. In winter the direction of the water flow is vertical in all the layers because of a high infiltration flow, while in spring the directions rotate and become parallel to the slope. In dry periods the mean direction described by the superficial gradients is around 150 (evaporation flow perpendicular to the slope), while in the intermediate and deep layer it is parallel to the slope. Evaporation in the intermediate and deep layers occurs mostly when the directions of the superficial flows are downward (September, October and November 2008). Moreover, water flow is very small during the summer in all the layers even if the gradients are very high because of the lower hydraulic conductivity in the dry period. In the remaining part of the year, the water flow modulus is five time higher than in the summer in the superficial soils, four times in the intermediate soils. Total water content inside each soil layer was estimated by means of TDR measurements. In Figure 6a and 6c the cumulated vertical flows in the superficial and intermediate layers are compared with the variation in the water volume calculated. The shape of the water volume curves in the superficial and intermediate soils is similar. The water balance over one year for both layers appears to be in equilibrium. By contrast, while the cumulated flow in the intermediate soils is nil, the cumulated inward flow through the superficial soils in the wet period is not balanced by the outward flow in the dry period. Hence a large cumulated flow infiltrates through the pumiceous layer 3.
soil 1 soil 2 soil 4
a)
180 150 120 90 60 30 0

A A M M J J J J
normal to the slope soil 1 soil 2 soil 4 parallel to the slope
b)
-30
2008 ASON S A 2008 O N
D J
FM F
2009
J A
Figure 5.
Water flow estimated on the basis of in situ suction and water measurements: a) modulus; b) direction.
1277
250 200 Q, DW (mm) 150 100 50 0 -50 -100 250

Q + DW (mm)
200 150 100 50 0 250 200 150 100 50 0
Qcumulated vertical water flow soil1-2 DW variation of water volume soil 1-2
a)
Q+DW, cumulated water flow parallel to the slope and variation of water volume soil 3
b)
Q, cumulated vertical water flow soil 4 DW, variation of water volume soil 4
c)
Q, DW (mm)
-50 -100
J JJ
2008
J JA
2009
Figure 6. Total vertical flows and water volume in a) soils 12; b) Total parallel flow and variation in water volume for soil 3 and c) soil 4.
In order to analyze the hydraulic behavior of this pumiceous layer interposed between the superficial and intermediate layers, the difference between the normal components to the slope of the cumulated vertical flows in the superficial and the intermediate layers was calculated (Fig. 6b). This difference is equal to the sum of two terms: the cumulative water flow which develops throughout the pumices in the direction of the layer (i.e. parallel to the slope); the variation in water volume contained inside layer 3. It is worth noting that the above sum rises during the observation period (see Fig. 6b) while the water balance in the layer is likely to be in equilibrium as previously observed for layers 1, 2 and 4. Therefore it can be concluded that most of the water leaves this soil as flow parallel to the slope. 5 CONCLUDING REMARKS
slope and a contribution to the variation in water content. If the water balance over one year were also assumed stationary in the pumices, most of the water from the superficial layers would disperse through the pumice in a parallel direction. REFERENCES
Di Crescenzo, G., Rotella, M. & Santo. A, 2007. Il contributo della geologia per lo studio dei meccanismi di innesco di colate rapide di fango al campo sperimentale di Monteforte Irpino (AV). In C. Nunziata (ed.) Piattaforme Evolute di Telecomunicazioni e di Information Technology per lOfferta di Servizi al settore Ambiente Petit-Osa: 263272. Rome: Aracne. Evangelista, A., Nicotera, M.V., Papa, R. & Urciuoli G. 2008. Field investigation on triggering mechanisms of fast landslides in unsaturated pyroclastic soils. UN-SAT, Durham UK. 24 July 2008. (pp. 909915). LONDON: Taylor & Francis Group plc (UK). Kool, J.B. & Parker J.C. 1987. Development of closed-form expressions for hysteretic soil hydraulic properties, Water Resour. Res., 23(1): 105114 . Papa, R., Evangelista, A., Nicotera, M.V. & Urciuoli G. 2008. Mechanical properties of unsaturated pyroclastic soils affected by fast landslide phenomena. UN-SAT Durham, UK. 24 July 2008. (pp. 909915). LONDON: Taylor & Francis Group plc (UK). Papa, R. 2007. Indagine sperimentale di una copertura piroclastica di un versante della Campania. PhD thesis,. Universit Federico II DI Napoli Topp, G.C. & Miller, E.E. 1966. Hysteretic moisture characteristics and hydraulic conductivities for glass-bead media. Proc. Soil. Sci. Amer. Soc., 30:156162.
Analysis of the suction and water content measurements collected in situ shows some features of the hydraulic regime in the subsoil. The monitoring results clearly show the seasonal variation in matric suction and volumetric water content in situ. While the matric suction and volumetric water content in the top part of the cover seem affected by single rainfall events, the corresponding variations are relatively small if compared to seasonal ones; conversely, in the intermediate and deeper parts of the cover they follow a trend unaffected by individual rainfall events. In the pumice layer the total water is divided into two terms: a flow parallel to the
1278
Performance of an instrumented slope under a capillary barrier system

H. Rahardjo, C.J. Hua, E.C. Leong & V.A. Santoso
School of Civil and Environmental Engineering, Nanyang Technological University, Singapore
ABSTRACT: A capillary barrier is a two-layer system of distinct hydraulic properties that is used as a cover system. The distinct hydraulic properties prevent water infiltration into the soil below the capillary barrier system by utilizing unsaturated soil mechanics principles. This paper illustrates the application of the capillary barrier system as a slope cover to prevent rainfall-induced slope failures. The capillary barrier was constructed on a slope which experienced a shallow slip surface. In this study, the capillary barrier system was designed as a cover system for slopes with a steep slope angle under heavy rainfall conditions of the tropics. The capillary barrier system was constructed using fine sand as the fine-grained layer and granite chips as the coarse-grained layer. Both layers were contained in geocells. The slope was instrumented with tensiometers and piezometers. The tensiometers were installed at different depths from about 0.5 m to 2.0 m below the slope surface. In addition, the adjacent original slope without the capillary barrier system was also instrumented using tensiometers in order to investigate the performance and effectiveness of the capillary barrier system in reducing rainwater infiltration and maintaining negative pore-water pressure in the slope. The measurement results showed that the capillary barrier system was effective in maintaining the negative pore-water pressures during rainfalls. Results of field measurements and numerical analyses are presented in the paper. The measurement and numerical results were in agreement, demonstrating the use of unsaturated soil mechanics principles in capillary barrier system. Keywords: capillary barrier, pore-water pressure, matric suction, residual soil, slope instrumentation 1 INTRODUCTION grained layer limits the downward movement of water through capillary barrier effect. The infiltrated water is then stored in the fine-grained layer by capillary forces. This infiltrated water is ultimately removed by evaporation and transpiration, lateral drainage through the slope or percolation into the underlying layer. When percolation (breakthrough) takes place, the capillary barrier no longer impedes water from infiltrating into the slope. Previous research works have indicated the effectiveness of the capillary barrier system as a soil cover in reducing rainfall infiltration (Tami et al. 2004; Khire et al. 2000; Morris & Stormont 1997a,b; Stormont 1996). Rahardjo et al. (2007) conducted a 1-D laboratory test to investigate the infiltration characteristics through a capillary barrier system and the storage of the fine-grained layer. The performance of capillary barrier models constructed using different materials (i.e. geosynthetic material and gravelly sand) as coarse-grained layer was also studied. Krisdani et al. (2006) also constructed a capillary barrier system soil cover using fine sand as the fine-grained layer and geosynthetic material as the coarse-grained layer on a cut slope which experienced a shallow slip. In this study, a Capillary Barrier System (CBS) using fine sand as the fine-grained layer and gran-
Rainfall-induced landslide is one of the most common natural disasters that occur in many residual soil slopes in tropical areas. Residual soils cover about two-thirds of the land in Singapore (Pitts 1984). The mechanism of failure for a rainfall-induced landslide can be described as follows: infiltrating water from rainfall events goes into the slope, resulting in a decrease of matric suction due to an increase in the pore-water pressures. The reduction of matric suction in unsaturated residual soils is equivalent to a decrease in shear strength of the soil along the potential slip surface (Brand 1982). Since rainwater infiltration into soil slopes is the major cause of rainfall-induced landslides, it is of value and interest to study preventive measures that can prevent or minimize rainwater infiltration into soil slopes. A capillary barrier is an earthen cover system using a fine-grained layer of soil overlying a coarse-grained layer of soil (e.g. Ross 1990a,b; Stormont 1996). The principle of the capillary barrier system is based on the contrast in unsaturated hydraulic properties (soil-water characteristic curves and permeability functions) of each material. Under unsaturated conditions, the difference in permeability between the fine-grained layer and the coarse-
1279
ite chip as the coarse-grained layer was constructed on a slope which experienced numerous rainfallinduced slope failures in the past. This slope with capillary barrier system and the adjacent original slope without capillary barrier system were monitored using tensiometers and piezometers to study the effectiveness and performance of the capillary barrier system as a slope cover. 2 CONSTRUCTION SEQUENCES AND FIELD INSTRUMENTATION
Figure 1. Laying of separator layer and fine-grained layer.
A slope which experienced numerous rainfallinduced slope failures was selected for the construction of CBS. Scars on the face of the slope indicated previous slope failures and movements. The total area of the slope to be covered with a capillary barrier system was approximately 140 m2. Geodrain was laid on the soil once the slope has been trimmed to the correct depth. Steel wires were then used to secure the geodrain to the soil to prevent slippage. The purpose of the geodrain was to provide drainage if a breakthrough were to occur. Geocells were laid over the geodrain and granite chips were used to fill up the entire geocells to form the underlying coarse-grained layer as shown in Figure 1. Steel J-pins of length 550 mm were used to secure the underlying geocells onto the ground. Manual tamping was carried out to compact the coarse-grained layer to the desired density. In-situ density tests were conducted on several locations of the coarse-grained layer to ensure that the layer has been compacted to the desired density. A layer of geofabric was laid on top of the coarse-grained layer to act as a separator between the coarse-grained and the fine-grained layers. A second layer of geocells was then laid above the geofabric. Steel J-pins of length 750 mm were used to secure the overlying geocells onto the ground (Fig. 1). Fine sands were used to fill up the geocells to form the fine-grained layer. Manual tamping was carried out to compact the fine-grained layer. Density tests were also conducted to ensure that the layer has been compacted to the desired density. Twenty four tensiometers were installed on the capillary barrier system, with 8 tensiometers being located on the capillary barrier system with topsoil and cow grass, 8 tensiometers on the capillary barrier system without topsoil and cow grass (Fig. 2), and 8 tensiometers on the original slope. From the 8 numbers of tensiometer for each area; 4 numbers of tensiometers were installed near the crest of the capillary barrier system representing Row A, and the other 4 tensiometers installed near the middle of the capillary barrier system representing Row B. Under the capillary barrier system
Figure 2. Capillary barrier system without topsoil and cow grass.
with topsoil and cow grass; tensiometers installed at the crest (Row A) were named A1, A2, A3 and A4 with a spacing of 0.5 m and insertion depths of 0.69 m, 1.3 m, 1.59 m and 2.18 m, respectively. Tensiometers installed at the mid slope (Row B) were named B1, B2, B3 and B4 with a spacing of 0.5 m and insertion depths of 0.69 m, 1.27 m, 1.57 m and 2.12 m, respectively. Similarly, under the capillary barrier system without topsoil and cow grass, tensiometers installed at the crest (Row A) were named A5, A6, A7 and A8 at 0.5 m spacing and insertion depths of 0.54 m, 1.15 m, 1.48 m and 2.05 m, respectively. Tensiometers installed at the mid slope (Row B) were named B5, B6, B7 and B8 at 0.5 m spacing and insertion depths of 0.59 m, 1.2 m, 1.5 m and 2.1 m, respectively. Under the original slope, tensiometers installed at the crest (Row A) were named A9, A10, A11 and A12 at 0.5 m spacing and insertion depths of 0.5 m, 1.29 m, 1.6 m and 2.18 m, respectively. Tensiometers installed at the mid slope (Row B) were named B9, B10, B11 and B12 at 0.5 m spacing and insertion depths of 0.65 m, 1.31 m, 1.6 m and 2.18 m, respectively. There were three numbers of piezometers: piezometer 1, piezometer 2 and piezometer 3 located at the crest, middle and toe of the slope, respectively. A dip-meter was used to measure the level of groundwater table.
1280
Figure 3. Schematic diagram of capillary barrier system without topsoil and cow grass.
Figure 4. Pore-water pressures versus rainfall (slope with capillary barrier system).
FIELD PERFORMANCE RESULTS
Manual monitoring of the tensiometers and piezometers was done 5 times a week (MondayFriday) at the same time for the first month. Subsequently, manual monitoring was carried out 3 times a week (Monday, Wednesday and Friday) at the same time. Rainfall data was obtained from the nearest rainfall station which was about 0.9 km away from the site. Manual monitoring was conducted for a period of 10 months, starting from February to November 2008. Pore-water pressures measured by the tensiometers were plotted against the rainfall data. In general, pore-water pressures under the slope with capillary barrier system were able to maintain negative pore-water pressures or matric suction under rainfall conditions as illustrated in Figure 4. Although there was a rise in pore-water pressure due to percolation, the pore-water pressure was still able to maintain negative values. This could be attributed to the lateral drainage in the fine-grained layer that reduced the amount of rainfall infiltration into the soil below the capillary barrier system significantly. As a result, the presence of negative pore-water pressure contributed to the shear strength of the soil, resulting in the slopes to be less susceptible to failure. On the other hand, the pore-water pressure under the original slope was easily affected by the rainfall infiltration. Figure 5 illustrates that the pore-water pressures under the original slope followed the rise and fall of rainwater infiltration. Pore-water pressures under the original slope were unable to maintain negative values when rainfall occurred. Based on the manual monitoring, the water level in piezometer 3 was always at the toe of the slope as plotted in Figure 6. The graph of piezometer readings versus rainfall data (Fig. 6) illustrates that the pattern of the piezometer readings followed exactly the pattern of the rainfall data.
Figure 5. slope).
Pore-water pressures versus rainfall (original
Figure 6. Groundwater level versus rainfall.
A rise in the intensity of rainfall was accompanied by a rise in the groundwater level as illustrated in Figure 6. It was suspected that during rainfall events, there was an internal water flow from the uphill part of the slope, contributing significantly to the rise of groundwater level. As a result, the pore-water pressure variations in both slopes with and without capillary barrier system were controlled by the fluctuations
1281
Granite and is located at Ang Mo Kio St. 21. The slope has a slope height of 8 m, slope angle of 37 at the toe of the slope and 33 at the repaired area (Fig. 3). Test results indicated that the water content, specific gravity, liquid limit, and plastic limit range of the residual soil are 3540%, 2.642.68, 5366%, and 3638%, respectively. The slope consisted of clayey silt with a unit weight of 20 kN/m3, an effective cohesion of 8 kPa, an effective friction angle of 33, and a b angle of 25. 4.1.2 Soil properties Figures 9 and 10 show the Soil-Water Characteristic Curves (SWCCs) and permeability functions, respectively, for fine-grained, coarse-grained, and residual soils at Ang Mo Kio St. 21. The SWCC of the soils in the slope was best-fitted using Fredlund & Xing equation (1994):
Figure 7. Pore-water pressures versus groundwater level (slope with capillary barrier system).
w C ( )
s
n ua uw ln e a
(1)
Figure 8. Pore-water pressures versus groundwater level (original slope).
in groundwater level. In addition, the pore-water pressure variation in the original slope was also affected by rainfall infiltration from the slope surface. However, this was not the case for the slope with capillary barrier since the capillary barrier system minimized rainfall infiltration into the underlying soils. Therefore, during rainfalls, the pore-water pressures in the slope with capillary barrier system (Fig. 7) were generally lower than the pore-water pressures in the original slope (Fig. 8), demonstrating the effectiveness of capillary barrier system in minimizing rainfall infiltration into the slope. 4 NUMERICAL STUDIES
Figure 9. Soil-water characteristic curves for finegrained, coarse-grained and residual soils at slope with capillary barrier system.
Two-dimensional seepage analyses were performed using finite element software, SEEP/W (Geoslope International Pte. Ltd. 2004). 4.1 Modelling of slope with capillary barrier system
Figure 10. Permeability functions for fine-grained, coarse-grained and residual soils at slope with capillary barrier system.
4.1.1 Slope geometry The slope constructed with the capillary barrier system lies on residual soils of Bukit Timah
1282
where w is volumetric water content, s is saturated volumetric water content, C() is correction factor, (ua uw) is matric suction (kPa), e is natural number (2.71828). Correction factor is taken as 1 as suggested by Leong & Rahardjo (1997), while fitting parameters fitted are a 147.88 kPa, n 1.44, and m 1.63. The SWCCs of the fine sand and granite chips in Ang Mo Kio St. 21 slope were estimated using Fredlund & Xing equation (1994). The fitting parameters for the fine sand are a 1.74 kPa, n 3.30, m 3.6, and those for the granite chip are a 0.12 kPa, n 1.66, m 1.22. The measured saturated permeability of the clayey silt soil, fine sand, and granite chip of Ang Mo Kio St. 21 slope were 6 106, 2.7 104, and 7.6 102 m/s, respectively. 4.1.3 Boundary conditions The boundary conditions applied to the finite element model are illustrated in Figure 3. The boundary of the slope model was set at three times the height of the slope. The non-ponding condition was selected to avoid excessive accumulation of rainwater on slope surface. On the ground surface, surface runoff would occur because the increase in pore-water pressures was prevented and the maximum computed pore-water pressure was limited to zero. Nodal flux, Q, equal to zero was applied along the bottom of the slope model and along the sides of the slope model above the groundwater table to simulate a no flow zone. The constant total head, hw, corresponding to each side was applied as the boundary along the sides of the slope model below the groundwater table. The actual rainfall intensity and its duration were applied to the surface of the slope model as flux boundary, q. Transient seepage analysis was performed on the slope for a natural rainfall from 26 May to 13 June 2008 and the results were then compared with the results from the manual monitoring of pore-water pressures in the slope. The total amount of rainfall from 26 May to 13 June 2008 was 157.6 mm and the maximum rainfall intensity was 33.6 mm/h. The rainfall pattern from 26 May to 13 June 2008 is shown in Figure 11. 4.1.4 Seepage analysis results Comparison of pore-water pressure profiles obtained from the numerical analyses and field measurements are presented in Figures 12 and 13 for the crest and the middle of the slope, respectively. At the start of the analyses, the groundwater table (measured by piezometers) was significantly high due to the previous rainfall events and a suspected source of water flow from the uphill part of the slope. For this reason, the initial groundwater table before applying
Figure 11. Rainfall data on 26 May 2008.
the actual rainfalls for the period of analysis (i.e. 26 May to 13 June 2008), was not in a steady state condition. This explains the reason for the more negative pore-water pressures at elapsed times greater than zero (i.e. up to 4 days) as depicted in both numerical results and field measurements. The problem of high water table during rainfalls was later resolved by installing one row of horizontal drains near the toe of the slope on 24 November 2008. In general, the numerical analysis shows a reasonably good agreement in the trend of the pore-water pressure profile with those obtained from field measurements. The discrepancies can be attributed to the fact that rainfall data were obtained from the nearest rainfall station and might not reflect the actual rainfalls on the slope. The problem was overcome later by installing a rain gauge on the slope in May 2009 to measure the actual rainfalls on the slope. In addition, the SWCCs used in the analyses were estimated from the grain size distribution. 5 CONCLUSIONS
Based on the pore-water pressure data measured by the tensiometers, the capillary barrier system constructed on the slope was able to reduce the rainwater infiltration, therefore maintaining the negative porewater pressures in the unsaturated zone. As a result, the shear strength in the unsaturated zone of the slope with capillary barrier system can be maintained, causing the slope to be less susceptible to future instabilities. However, further measurements are required to have a better understanding of the performance of capillary barrier system as a slope cover including the effect of fluctuation in groundwater level on the effectiveness of the capillary barrier. The measurement
1283
Pore-Water Pressure (kPa) -50 122 Elevation, z (m) 121 120 119 118 117
t = 0 (Numerical) t = 4days (Numerical) t = 9days (Numerical) t = 11days (Numerical) t = 16days (Numerical) t = 18days (Numerical) t = 0 (Field Data) t = 4days (Field Data) t = 9days (Field Data) t = 11days (Field Data) t = 16days (Field Data) t = 18days (Field Data)
ACKNOWLEDGEMENT
10 20
-40
-30
-20
-10
The work described in this paper is supported by the Housing & Development Board and Nanyang Technological University, Singapore. REFERENCES
Brand, E.W. 1982. Analysis & Design in Residual Soils, in Proc. ASCE Specialty Conf. Eng. Ad Construction in Tropical and Residual Soils (Honolulu, HI): 89143. Fredlund, D.G. & Xing, A. 1994. Equations for the SoilWater Characteristic Curve, Canadian Geotechnical Journal 31: 533546. Geoslope International Pte. Ltd. 2004. SEEP/W Users Guide for Slope Stability Analysis. Geoslope International Ltd., Calgary, Alberta, Canada. Khire, M., Benson, C. & Bosscher, P. 2000. Capillary Barriers: Design Variables and Water Balance, J. of Geotech and Geoenvir. Engrg., ASCE 126(8): 695708. Krisdani, H. 2006. The Use of Residual Soil and Geosynthetic Material in Capillary Barrier System. PhD thesis, School of Civil and Environmental Engineering, Nanyang Technological University, Singapore. Leong, E.C. & Rahardjo, H. 1997. A Review on SoilWater Characteristic Curve Equations, ASCE Journal of Geotechnical and Geoenvironmental Engineering, December, 123(12): 11061117. Morris, C.E. & Stormont, J.C. 1997a. Capillary Barriers and Subtitle D Covers: Estimating Equivalency, J. of Envir Engrg., ASCE 123(1): 310. Morris, C.E. & Stormont, J.C. 1997b. Capillary Barriers and Subtitle D Covers: Estimating Equivalency Closure, J. of Envir Engrg., ASCE 123(1): 310. Pitts, J. 1984. A Review of Geology and Engineering Geology in Singapore, Quarterly Journal of Engineering Geology 17: 93101. Rahardjo, H., Krisdani, H. & Leong, E.C. 2007. Application of Unsaturated Soil Mechanics in Capillary Barrier System, in Proc. 3rd Asian Conf. on Unsaturated Soils: 127137. Ross, B. 1990a. Reply to Comment on the Diversion Capacity of Capillary Barriers by T.S. Steenhuis nad J.Y. Parlange, Water Resources Research 27(8): 2157. Ross, B. 1990b. The Diversion Capacity of Capillary Barriers, Water Resource Research 26(10): 26252629. Stormont, J.C. 1996. The Effectiveness of Two Capillary Barriers on a 10% Slope, Geotechnical and Geological Engrg 14: 243267. Tami, D., Rahardjo, H., Leong, E.C. & Fredlund, D.G. 2004. A Physical Model for Sloping Capillary Barriers, Geotechnical Testing Journal, ASTM 27(2): 173183.
Figure 12. Comparison of pore-water pressure profiles obtained from numerical analyses and pore-water pressure data measured in the field near at the crest of the slope from the beginning of rainfall (t 0) until the end of rainfall (t 18 days) on 26 May 2008.
Pore-Water Pressure (kPa) -50 121 120 119 118 117 116 -40 -30 -20 -10 0 10 20
Elevation, z (m)
t = 0 (Numerical) t = 4days (Numerical) t = 9days (Numerical) t = 11days (Numerical) t = 16days (Numerical) t = 18days (Numerical)
t = 0 (Field Data) t = 4days (Field Data) t = 9days (Field Data) t = 11days (Field Data) t = 16days (Field Data) t = 18days (Field Data)
Figure 13. Comparison of pore-water pressure profiles obtained from numerical analyses and pore-water pressure data measured in the field at middle of the slope from the beginning of rainfall (t 0) until the end of rainfall (t 18 days) on 26 May 2008.
results showed that the capillary barrier system was effective in maintaining the negative pore-water pressures during rainfalls. The numerical results were able to simulate the field measurement reasonably, demonstrating the application of unsaturated soil mechanics principles in modelling capillary barrier system as a slope protection measures.
1284
Matric suction in unsaturated pyroclastic slopes

A. Scotto di Santolo & A. Evangelista
University of Naples Federico II, Italy
ABSTRACT: In situ measurements of matric suction in the pyroclastic covers of four slopes in Naples, which were susceptible to debris flows, were conducted from 2000 to 2005. The water level is several hundred meters below the ground surface. The measured suctions were used to calibrate a numerical hydrological model, which referred to the first four meters of depth that could be affected by flow movements. The volcanic soil of the cover consists of loose deposits of ashy and coarse levels that originated from the recent activities of the Campi Flegrei (Phlegrean Fields). The bedrock is made of Volcanic Tuff (mainly Neapolitan Yellow Tuff). This model makes it possible to reproduce the seasonal cycle of suction in the pyroclastic cover in the study area and could be used to assess an early warning system as part of risk management procedures. 1 LANDSLIDES
In the city of Naples there are hilly areas of variable height (h) ranging from the 100 to 500 m which are partly exposed to the sea and partly located in urbanized areas. These slopes consist of a rocky skeleton of volcanic nature (tuffs of different age) and are covered by pyroclastic products of the activity of the Third Phlegrean period (12,000 years), illustrated in Figure 1. The latter are characterized by alternations of ash and lapilli deposits. They are characterized by high porosity (mean void ratio e 1.28, st. dev. 0.23) and are partially saturated (mean degree of saturation Sr 0.46 st. dev. 0.15). The ticks are very variable, ranging from one meter to some tens of meters on the hills of Camaldoli and Posillipo, and up to several tens of meters in the Conca di Agnano. In such areas landslides are widespread (Beneduce et al., 1988; Pellegrino, 1994; Calcaterra and Guarino, 1999; Scotto di Santolo, 2000a; 2002). The standard type of the phenomena is for 30% complex landslides (translational or rotational slide, or falls that lead to debris flows). The volumes are generally reduced (reaching the maximum rates of 1000 m3) and only in some cases do the covered distances (L) exceed the foot of the hill (Lmax of 300 m). The geometric feature of landslides is compared in Figure 2 with what occurred in the Campania region (Monti Lattari and Monti di Sarno) in recent years. It is observed that the distance covered increases with the difference in height between the crown and the toe of landslides H with a linear relationship for H less than 400 m whereas it increases with a power law for H greater than 500 m.
Figure 1. Location of monitored slopes on the geological map of Campi Flegrei (modified from Isaia et al, 2004).
1000 800 H (m) 600

H = 0,57 L
H = 1,82 L0,76
400 200 0
H = 0,78 L
Napoli (N=107) Penisola Sorrentina (N=17) Sarno-Quindici 1998 (N=26)
1000
2000 L (m)
3000
4000
Figure 2. Relation between travel distance and difference between the height of the crown and the toe of the deposits of debris flows (modified from Scotto di Santolo, 2002).
1285
Moreover, for the same H the mobility is a function of a travel path (Scotto di Santolo and Evangelista, 2008). The reach angle (arctan(H/L)) is constant and equal to 38. Nevertheless the risk is high because of the intense urbanization of the hilly area, Figure 3. The landslides occurred in the rainy period: in Figure 4 the frequency of the landslides in the months of the year are reported for 101 complex landslides which occurred in this deposit in the period between 1986 and 2009 (November). It has been observed that the phenomena mainly take place in winter (the maximum are recorded in January, most of them in January 1997) and the beginning of spring. If compared with average monthly rainfall, as shown in Figure 5, it can be observed that the landslides do not occur in correspondence with the wetter period (October December), but are delayed by several months. By defining the recurrence of landslides as the ratio between the period of time in which these events
took place and their number, it was calculated that this time is less than 10 years in the study area. The delay between rainfall and landslides is a function of climatic conditions and the hydraulic characteristics of the cover which, since it is partially saturated, has a conductivity that varies with water content (Scotto di Santolo and Evangelista, 2009). The triggering of landslides is strongly related to rainfall: the increase of water content in soil is the principal undisputed factor leading to soil failure (e.g. Lumb, 1962; Iverson, 2000; Scotto di Santolo et al., 2000). Infiltration of rainfall leads to a decrease in suction which reduces the shear strength of the shallow soil (e.g. Fredlund et al., 1978; Evangelista & Scotto di Santolo, 2001). The prediction of triggering factors is of extreme importance for hazard mitigation, due to the dynamic features that characterize these phenomena during their evolution. 2 DESCRIPTION OF MONITORING SYSTEM
Figure 3. View of the Astroni Volcano, Naples, showing the high level of urbanization of the hilly zones.
Monthly frequency of landslides 1986-2009
0.8 0.6 0.4 0.2 0.0 J F M A M J J A S O N D
Figure 4.
Average Monthly Rainfall (mm) AMR 1986-2009
Distribution of landslides (19862009).
150 100 50 0 J F M A M J J A S O N D
750 500 250 0
Figure 5.
Average monthly rainfalls (19862009).
Cumulative AMR (mm)
200
1000
Four sites which are representative of the morphologic and stratigraphic conditions of slopes susceptible to debris flows in Naples were monitored with rain gauges, tensiometers, psychrometers, tiltmeters and TDR probes (Evangelista et al., 2003). Two sites are situated on the hill of the Camaldoli, the highest relief in the city (456 m asl), one on the hill of Posillipo and one in the basin of Agnano (Figure 1). All sensors were automatically recorded every hour in routine conditions and the data was sent by modem to a remote computer. The measurements began in February 2000 and were completed in June 2005. It represents one of the first installations (if not the first) for continuum measurements of suction in a pyroclastic cover in southern Italy. Unfortunately the inexperience with the use of innovative equipment (the first in 2000) and initial uncertainties did create some problems. Moreover maintenance always was not carried out on a continuous basis with the result that some instruments were often desaturated for long periods. Therefore some stations sometimes lack readings. The stations that worked best were S1 and the S2 during the period 20012003 and therefore reference is repeatedly made to them. For all of them temporary measuring devices were also available and were used to gain a more complete and wider picture of the phenomenon. In Figure 6 the time histories of average daily readings on the tensiometers are reported
1286
a)
50 40 30 20 10 0 23/09/2001 22/12/2001 22/03/2002 20/06/2002 18/09/2002 17/12/2002 17/03/2003 15/06/2003 13/09/2003 12/12/2003
b)
gauge reading (kPa) -10
z = -4.0 m z = -3.0 m z = -2.0 m z = -1.5 m z = -1.0 m z = -0.5 m
rain (mm/day)
-60
23/09/2001
22/12/2001
22/03/2002
20/06/2002
18/09/2002
17/12/2002
17/03/2003
15/06/2003
13/09/2003
12/12/2003
c)
gauge reading (kPa) -10
-60
23/09/2001
22/12/2001
22/03/2002
20/06/2002
18/09/2002
17/12/2002
17/03/2003
15/06/2003
13/09/2003
12/12/2003
Figure 6. S2 sites.
Results of monitoring: a) rainfall; b) and c) average values at gauge readings on the tensiometers at S1 and
for the stations S1 and S2 for each depth of measurement (0.5; 1; 1.5; 2; 3; 4 m from the p.c.). It was observed that the response at 0.5 m from the surface is closely related to rainfall (also shown in Figure 6a). A delay was recorded according to depth. Therefore while the suction reaches the minimum at 0.5 m in the winter months, at 2.0 m the minimum is reached at the beginning of spring. For greater depths (34 m) less variation was observed so that it was considered to be fairly constant, with the exception of the measurements taken in the spring of 2002. 3 MATRIC SUCTION REGIME
year. The results of the statistical analysis are shown also in the diagrams in Figure 7. The best fit for the statistical model varies according to the depth and the time interval T considered. For each year the coefficient R2 and the standard deviation are estimated by comparing the observed data with the calculated data. The results for all depths are shown in Table 1. As already mentioned, the measurements were often interrupted due to the desaturation of the equipment. This is reflected by the goodness of the fitting parameters of the statistical model. 3.2 Hydrological model
This paragraph provides a review of the available suction data, and a hydraulic model of infiltration that explains the seasonal cycles of the suctions in the pyroclastic soil is presented. 3.1 Review of the suction data The data available in all sites was collected and interpreted in terms of daily averages for each depth (0.5; 1.0; 1.5; 2.0; 3.0 and 4 m). The data, reported in Figure 7, were interpolated from sinusoidal functions of a period equal to a
The governing flow equation for one-dimensional conditions in an unsaturated rigid porous medium is given by the modified form of the Richards equation: K K t z z (1)
where is the volumetric water content, is the pressure head (negative for unsaturated soil), K is the hydraulic conductivity, z is a vertical coordinate positive upward and t denotes time.
1287
500
1000
1500
@ 0.5m
2000
suzione (kPa) suzione (kPa)
0 -20 -40 -60 -80
0 -20 -40 -60 -80

0 -20 -40 -60 -80
@ 1.0m
suzione (kPa)
@ 1.5m
suzione (kPa)
0 -20 -40 -60 -80 0
suzioni (kPa)
@ 2.0m
-20 -40 -60 -80

0 -20 -40 -60 -80 0 500 1000
(day) t Time (giorni)
@ 3.0m
suzione (kPa)
@ 4.0m
1500
2000
Figure 7.
In situ suction measurements (years 20002005) and fitting functions at each depth.
1288
Table 1. Statistical variability of the data and time for which there is the best correlation. Depth from s.l. (m) 0.5 1 1.5 2 3 4 Average in situ suction (kPa) 41 41 41 37 34 27 R2 (1) 0.87 0.87 0.99 0.61 0.88 0.1 T (d) 1460 1825 1460 1825 1460 1825 1460 1825 1460 1825 243 608 Dev. St. (kPa) 8.00 8.10 1.95 8.44 4.30 9.80 T (d) 279 644 937 1302 1004 1469 662 10279 383 718 594 959
0 suzione (m)
0 -1 -2 -3 -4 -5 -6 -7 -8 0 100 200 300 400 Time [days] 500 600 700 800
-2 0 -4 -6 -8
200
400
600
@ 1.0m
0 suzione (m)
Matric suction (m)
-2 0 -4 -6 -8
200
400
600
@ 1.5m
-1.0 -1.5 -2.0 -2.5 -3.0 -3.5 -4.0 0 100 200 300 400 Time [days] 500 600 700 800
0 -2 0 -4 -6 -8 200 400 600
suzione (m)
@ 2.0m
Figure 8.
Sketch of boundary conditions.

suzione (m)
0 -2 0 -4 -6 -8 200 400 600
This equation was solved with the finite-element model HYDRUS 1D code (Simunek et al., 2005) using the well-known constitutive relation proposed by van Genuchten (1980). The stratigraphy and soil parameters were obtained by the authors in previous works (Scotto di Santolo et al., 2005). The model was implemented on the basis of the measurements of suction that were carried out. Time histories of suction in a shallower and deeper layer were used as boundary conditions (0.5 and 4 m; Figure 8). The suctions calculated for the intermediate depths (1,0; 1.5; 2.0; 3.0 m) were then compared with those measured. This approach is useful because it does not take into account the quantification of the water infiltration of the precipitation and does not require simulation of the evaporation and runoff. The procedure derives from the observation that infiltration is conditioned by the shallower layers of the ground, which are extremely variable due to their properties, and by the type and density of vegetation.
@ 3.0m
Time (day)
Figure 9. Comparison between in situ suction and the calculated data at intermediate depth.
t (giorni)
3.3 Comparison between the hydrological model and in situ measurements The calculated suction time histories, expressed in days, are reported in Figure 9 together with the measured ones (1.0 m, 1.5 m, 2.0 m and 3.0 m). We obtained a satisfactory correspondence between measurements and the calculated data, even though the desired coincidence was not possible in cases where the control parameters were highly variable. 4 CONCLUDING REMARKS
In situ measurements of matric suction in four slopes which are susceptible to debris flows in
1289
Naples, Italy, were used to calibrate a numerical hydrological model. The analyzed data provided the mean value of matric suction during the five-year period 20002005 and therefore does not consider situation details which can be verified or which may happen. The measurements and their interpretation enabled us to making some useful considerations about setting up a warning system in order to forecast the trigger conditions of landslides in partially saturated soils. ACKNOWLEDGEMENTS Part of this research was carried out in the context of the PRIN Project (Progetto di Ricerca di Interesse Nazionale 2006082400_002). REFERENCES
Beneduce P., DElia G., Guida M. 1988. Morfodinamica dei versanti dellarea flegrea (Campania): erosione in massa ed erosione lineare. Mem. Soc. Geol. It., 41. Calcaterra D., Guarino P.M. 1999. Dinamica morfologica e fenomeni franosi recenti nellarea collinare napoletana (settore occidentale). Geologia Tecnica ed Ambientale 2/99, Roma. Evangelista, A. & Scotto di Santolo, A. 2001. Mechanical behavior of unsaturated pyroclastic soil. Proc. International Conference on Landslides: Causes, Impacts and Countermeasures, Davos-Switzerland, June 2001, pp. 3544. Evangelista A., Nicotera M.V., Scotto di Santolo A. 2003. Experimental and Theoretical Investigation on matric suction measurements in pyroclastic soils, Proc. Int. Conf. on Fast Slope Movements-Prediction and Prevention for Risk Mitigation, Naples, May 2003. Evangelista A. e Scotto di Santolo A. 2004. Analysis and field monitoring of slope stability in unsaturated pyroclastic soil slopes in Napoli, Italy. Proc. 5th Int. Conference on Case Histories in Geotechnical Engineering, New York 2004. Evangelista, A., Scotto di Santolo, A. & Lombardi, G. 2007. Previsione dellinnesco di fenomeni franosi nelle coltri piroclastiche della citt di Napoli. Atti del XXIII Convegno Nazionale di Geotecnica, Padova 2007, pp. 227234.
Fredlund D.G., Morgenstern, N.R. & Widger R.A. 1978. The shear strength of unsaturated soils. Canadian Geotech. J., 15: 313321. Isaia, R., DAntonio, M., DellErba, F., Di Vito, M., Orsi, F. 2004. The Astroni volcano: the only example of closely spaced eruptions in the same vant area during the recent history of the Campi Flegrei caldera (Italy), Journal of Volcanology and Geothermal Research 133: 171192. Iverson, R.M. 2000. Landslide triggering by rain infiltration. Water Resources Research, 36, 7: 18971910. Lumb, P. 1962. The properties of decomposed granite, Gotechnique J. 12: 226243. Pellegrino A. 1994. I Fenomeni Franosi nellarea Metropolitana Napoletana. Acta Nepolitana, Guida Ed. Napoli, 18, 237256. Scotto di Santolo, A. 2000. Analisi geotecnica dei fenomeni franosi nelle coltri piroclastiche della provincia di Napoli. Tesi di dottorato XII ciclo Consorzio Universit di Napoli e Roma, 1/2000. Scotto di Santolo, A., Nicotera, M.V. & Ramondini, M. 2000. Analysis of instability phenomena affecting a cut slope in unsaturated pyroclastic soils. Proc. 8th Int. Symposium on Landslides, Cardiff, June 2000, pp. 13531360. Scotto di Santolo, A. 2002. Le colate rapide. 118p. Hevelius Ed., Benevento. Scotto di Santolo A., Nicotera M.V., Evangelista A. 2005. Monitoring matric suction profiles in Partially Saturated Pyroclastic Topsoil Slope. Proc. Int. Symposium Advanced Experimental Unsaturated Soil Mechanics, Trento, giugno 2005, pp. 533539. Scotto di Santolo, A. & Evangelista, A. 2009. Reliability of rainfall threshold for pyroclastic debris flows. Proc. Inter. Conference on Soil Mechanics and Geotechnical Engineering (17th ICSMGE), 59 October 2009, Alexandria, Egypt, Vol II, 14911494. Scotto di Santolo, A. & Evangelista, A. 2008. Some observations on the prediction of the dynamic parameters of debris flows in pyroclastic deposits in the Campania region of Italy. Int. Journal of Natural Hazards, 2009, 50, pp. 605622, DOI: 10.1007/s11069-008-9334-3. Simunek J., Huang, K. & van Genuchten, M.Th., 1998. The HYDRUS code for simulating the one-dimensional movement of water, heat, and multiple solutes in variably-saturated media. Version 6.0. R. R. No. 144, U.S. Salinity Laboratory, Riverside, California, 164pp. van Genuchten, M.Th. 1980. A closed form equation for predicting the hydraulic function of unsaturated soils. Journal Soil Science Society of America, 44.
1290
Analysis of influencing factors in strengthening expansive soil slope with the coverage of geo-membrane
J.P. Yuan, J.H. Wu & Z.Z. Yin
Key Laboratory of Ministry of Education for Geomechanics and Embankment Engineering, Hohai University, Nanjing, China Geotechnical Research Institute, Hohai University, Nanjing, China
ABSTRACT: Cracks development caused by wetting-drying cycle is one of the key factors resulting in instability of expansive soil slope. In order to find an effective method of strengthening expansive soil slope, a field model test was carried out. Field test results show that using geo-membrane as coverage on the surface of expansive soil slope can significantly reduce the change range of slope water content, mitigate the cracks development, maintain high soil suction, and then finally strengthen expansive soil slope. Numerical analyses are carried out to discuss the factors influencing the effect of this method including slope configuration, coverage range, permeability of soils, rainfall intensity, rainfall duration, underground water level etc. Keywords: expansive soils, geo-membrane, slope, stability analysis 1 INTRODUCTION 2 MECHANISM OF SLOPE FAILURE AND STRENGTHENING PRINCIPLE WITH THE COVERAGE OF GEO-MEMBRANE
Expansive soil has been found in more than 20 provinces in China with wide distribution, and its damage to the projects could not be neglected (Li 1992). The middle route of South-to-North Water Diversion will pass through the expansive soil areas of more than 300 kilometers. The slope stability of expansive soil is important to the projects for normal operation. So its necessary to study on the slope stability of expansive soil (Yuan 2003). The researches on expansive soil slope at present mainly focus on in stability analyses (Huang et al. 2002, Bao 2004), strengthening methods (Zhang 2006, Ying 2007), numerical calculation (Li et al. 2001, Yin et al. 2002, Wu et al. 2008) etc. there are two kinds of strengthening methods for expansive soil slope: passive and active methods. Passive methods are the methods which restrict the swell-shrinking deformation of expansive soil by supporting structures to avoid oversize deformation, such as retaining wall, geotechnical grille, anti-slide pile, bolt etc. Active methods are the methods which prevent the soil water content from variation by some measures to avoid deformation, such as slope protection with vegetation, coverage with geo-membrane etc. As we know, the swellshrinking characteristic of expansive soil itself is the internal reason of the expansive soil slope failure. So the active strengthening methods definitely have a better effect comparing to the passive strengthening methods.
Studies show that the cracks which caused by drying-wetting cycles are one of the main reasons of expansive soil slope failure (Liu 2008). The permeability of soil without cracks is extremely low. The upper layer shrinks due to water loss by evaporation in dry season, while the soil water content of lower layer does not meanwhile. This results in the uneven shrinkage deformation and cracks will grow when the uneven shrinkage deformation reaches to a certain degree. The cracks further accelerate evaporation and continually grow into deeper layers. The expansive soil strength is reduced both by growing cracks and infiltration effects, which results in slope failure finally. Therefore, to keep the soil water content stable to avoid cracking it is an effective method for slope treatment. Geo-membrane is an impervious material. If expansive soil slope is covered by geo-membrane on surface, both rainfall infiltration and evaporation can be prevent pass through. In this way, the soil water content of slope surface can keep stable and swell-shrinking deformation would not be generated. Cracks would not be generated either. That is to say, expansive soil slope could be strengthened by keeping the soil water content stable with the coverage of geo-membrane. In order to verify this method, field tests with coverage and without coverage respectively had been carried out in
1291
Zhen-Jiang city, Jiang-Su province, China (Yuan et al. 2009). The researches on strengthening effect of the coverage with geo-membrane had been carried out by measuring the soil suction, water content and displacement of slope for about 1.5 years. The results showed that the suction and water content were stable and displacement was much less under the condition with coverage. So the rationality and validity method of coverage with geo-membrane was verified primarily. In this paper, the strengthening effect of this method under different conditions will be analyzed by numerical analysis methods.
method from the reference (Fredlund et al. 1997) by using SWCC. 3.1 Models and schemes The finite element mesh is shown in Figure 3. There are six schemes for analysis which are shown in Table 1. Scheme 1 is the one on which the others are based in order to analyze the influence degree of coverage effect on different slope ratio, overburden thickness, coverage area, soil permeability coefficient, rainfall intensity and duration. All but scheme 4 calculate the seepage field under the condition with coverage and without coverage respectively. 3.2 Initial and boundary conditions
SEEPAGE ANALYSIS ON EXPANSIVE SOIL SLOPE WITH THE COVERAGE OF GEO-MEMBRANE
The finite element software VADOSE/W is adopted to carry out seepage analysis. It can solve seepage problems considering the factors such as rainfall infiltration, groundwater, evaporation conditions etc. 3.1 SWCC & Osmotic function curve Figure 1 shows the adopted SWCC which is measured from some expansive soil. Figure 2 shows the osmotic function curve which is obtained with the
The initial water table was set at Y 3 m, horizontaly; the maximum negative pore water pressure head is 10 m. Water-air exchange exists on top of this model, while the other boundaries are impervious. The initial temperature is 20 m. The calculation process contains 30 steps and one day is one step. The rainfall, wind speed, temperature and humidity shown in Figure 4 and Figure 5 are assumed values to consider the most unfavorable climatic factors. The values shown in Figure 1 to Figure 5 are the basic parameters of scheme 1. The differences between scheme1 and the others are shown in table 1. 3.3 Calculation results & analysis
Figure 1.
Soil water characteristic curve.
Figure 6 shows pore water pressure contours of slope with the coverage of geo-membrane of scheme 1. It shows that the rainfall infiltration and evaporation have little effect on the pore water pressure distribution when with coverage of geo-membrane. In addition, the pore water pressure is invariant in all except the rectangular region shown in top left corner in Figure 6. Although the right part of this rectangular region is located under geo-membrane, atmospheric factors have much effect on the pore water pressure notably. We call it blind area of coverage. Therefore, it should be reasonable to arrange the coverage area in order to prevent the important parts from locating in blind area of coverage. Figure 7 shows the pore water pressure-depth relationship in 11(x 12 m), 22(x 23 m),
Figure 2.
Osmotic function curve.
Figure 3. (unit: m).
Finite element mesh of numerical model
1292
Table 1. No. 1 2 3 4 5 6
Numerical calculation scheme. Slope ratio Overburden thickness Coverage area X 540 m X 540 m X 540 m a X 1540 m b X 1040 m X 540 m a b X 540 m Permeability coefficient k 1e-5 m/s k 1e-5 m/s k 1e-5 m/s k 1e-5 m/s Rainfall intensity and duration Scheme description
1:2 50 cm a 1:1.75 50 cm b 1:2.25 1:2 a 20 cm 1:2 1:2 1:2 b 80 cm 50 cm 50 cm 50 cm
h 50 mm/d, t 10 d basic h 50 mm/d, t 10 d different slope ratio h 50 mm/d, t 10 d different overburden thickness h 50 mm/d, t 10 d different coverage area
k 1e-3 m/s h 50 mm/d, t 10 d different soil permeability k 1e-7 m/s k 1e-5 m/s a h 50 mm/d, t 15 d different rainfall intensity and duration b h 100 mm/d, t 10 d
Where: X is horizontal coordinates; k is saturation permeability coefficient; h is rainfall intensity; t 10 d means that the rainfall-time and dry-time are 10 days respectively.
60 50 rainfall (mm) 40 30 20 10 0 1 6 11 16 21 26 (day) 0 1 wind speed (m/s) rainfall wind speed 1 .5
0 .5
Figure 4. Rainfall-time and wind speed-time relationship.

40 30 20 40 20 0 1 6 11 maximum temperature maximum relative humidity 16 21 26 (day) minimum temperature 10 0 100 relative humidity (%) 80 60
Figure 5. tionship.
Temperature-time and humidity-time rela-
Figure 6. Pore water pressure contours with coverage on the 10th day of scheme 1.
33(x 36 m) sections (Figure 3) at different time under the conditions with and without coverage respectively. The sign 10d-12 m and 10d-12 m-no are the results with coverage and without coverage respectively on the 10th day at 1-1 section; the others follows analogically. The results show that the climatic factors have little effect on the soil pore water pressure at different sections under the condition with coverage. However, the pore water pressure changes notably under the condition without coverage. The values increase when raining and decrease when sunning; the mutative range of pore water pressure has a negative correlation with the depth during the same period. As the limitation of space forbids, following analyses only focus on different schemes with coverage, the results without coverage are just listed out as contrast. Figure 8 shows the pore water pressure-depth relationship in 2-2 section under the condition of different slope ratio at different time. It is shown that the variation gradient of pore water pressure with depth has a positive correlation with the slope ratio. Figure 9 shows the pore water pressure-depth relationship in 2-2 section under the condition of different overburden thickness. It shows that the different overburden thickness hardly affect the pore water pressure with coverage. Anyway, the environment factors such as sunshine, weathering, artificial factors etc. may affect the efficiency and durability of geo-membrane in practical engineering. So the geo-membrane should be backfilled with a certain thickness of soil to prevent from being destroyed. Figure 10 shows the pore water pressure-depth relationship in 2-2 section under the condition of different coverage area. 2-2 section is out of the blind area of coverage, so the pore water pressure keeps stable at different time in this scheme.
temperature (C)
1293
pore pressure (kPa) -200 -160

10d-12m 20d-12m 30d-12m 10d-12m-no 20d-12m-no 30d-12m-no 10d-23m 20d-23m 30d-23m 10d-23m-no 20d-23m-no 30d-23m-no 10d-36m 20d-36m 30d-36m 10d-36m-no 20d-36m-no 30d-36m-no
pore pressure (kPa)
-120
-80
-40
40 0
-100
-80
-60
-40
-20
20
40
60 0
buried depth (m)
10d-4a 20d-4a 30d-4a 10d-4b 20d-4b 30d-4b
Figure 7. Pore water pressureburied depth relationship at different time of scheme 1.

pore pressure (kPa) -100 -80 -60 -40 -20 0 20 0
Figure 10. Pore water pressureburied depth relationship at different time of scheme 4 (X 23 m).
pore pressure (kPa) -120 -80 -40 0 40 0
10d-5a
10d-2a 20d-2a 30d-2a 10d-2a-no 20d-2a-no 30d-2a-no 10d-2b 20d-2b 30d-2b 10d-2b-no 20d-2b-no 30d-2b-no
buried depth (m)
30d-5a 10d-5a-no 20d-5a-no 30d-5a-no 10d-5b 20d-5b 30d-5b 10d-5b-no 20d-5b-no 30d-5b-no
-100 -80 -60 pore pressure (kPa) -40 -20 0 20 0
pore pressure (kPa) -100 -80 -60 -40 -20 20 0
10d-3a 20d-3a 30d-3a 10d-3a-no 20d-3a-no 30d-3a-no 10d-3b 20d-3b 30d-3b 10d-3b-no 20d-3b-no 30d-3b-no
buried depth (m)
15d-6a 30d-6a 15d-6a-no 30d-6a -no 10d-6b 20d-6b 30d-6b 10d-6b-no 20d-6b-no 30d-6b-no
However, the coverage area still should be arranged reasonably in order to prevent the key parts locating in blind area of coverage because the slope surface could not be covered with geo-membrane totally. Figure 11 shows the pore water pressure-depth relationship in 2-2 part under the condition of different soil permeability coefficient. Water loses easily when the soil has large permeability coefficient and little water accumulation. The pore water pressure keeps stable in presence or absence of geomembrane in this instance. While hysteresis effect of rainfall infiltration and evaporation exists obviously
when the soil has low permeability coefficient under the condition without coverage and the pore water pressure changes distinctly. This phenomenon exists hardly under the condition with coverage. Figure 12 shows the pore water pressure-depth relationship in 2-2 section under the condition of different rainfall intensity and duration. The effect of rainfall intensity on pore water pressure is greater than the effect of duration under the condition without coverage. While the rainfall intensity and duration has hardly effect on pore water pressure under the condition with coverage.
1294
buried depth (m)
buried depth (m)
20d-5a
buried depth (m)
safety factor
In a word, soil pore water pressure can keep stable with the coverage of geo-membrane; while soil pore water pressure is mutative due to climatic factors without coverage. Therefore, the change range of pore water pressure is decreased evidently under the condition with coverage to prevent the soil from cracking and the soil suction can maintain highvalue. Moreover, according to the analyses above, slope ratio, overburden thickness, rainfall intensity and duration have little effect on the soil pore water pressure, while coverage area and soil permeability coefficient have significant effect on pore water pressure. The results remind us the coverage area should be defined reasonably according to the practical conditions and well drainage is necessary. 4 STABILITY ANALYSIS ON EXPANSIVE SOIL SLOPE WITH THE COVERAGE OF GEO-MEMBRANE
1.8
safety factor
1.7
1.6
1 2-a 2-b 3-a 3-b 4-a 4-b 5-a 5-b 6-a 6-b
1.5 5 10 15 20 25 30 35 (Day)
(a) With coverage

1.2 1-no 2-a-no 2-b-no 1.1 3-a-no 3-b-no 5-a-no 1.0 5-b-no 6-a-no 6-b-no 0.9 5 10 15 20 25 30 35 (Day)
Based on the shear strength theory of unsaturated soil (Fredlund 1997), the software SLOPE/W is adopted to analyze the slope stability. The pore water pressure has already been obtained as above seepage analysis, and then the effective stress and suction at the middle of each soil bar can be obtained. The slope with coverage is simplified to homogeneous one. The slope without coverage was divided into four layers along the depth to simulate the soil strength decreasing due to cracks growing in slope. Both models are shown in Figure 13. The parameters are shown in Table 2. Figure 14 shows the safety factor Fs-time relationship of different schemes. The results are as follows:
d
(b) Without coverage
Figure 14. schemes.
Safety factors-time relationship of different
(a) With coverage

a b c d
Figure 13. Calculation models for stability analysis. Table 2. Soil layer a b c d Mechanics parameters of soil. d(kN/m3) 16.0 16.0 18.0 18.0 c(kPa) 6.0 8.0 12.0 18.0 () 4.0 6.0 12.0 16.0 b() 2.0 2.0 4.0 6.0
the Fs of different schemes at different time under the condition of coverage change a little and the values are high. The calculation result is shown in Figure 15a and a deep slip arc can be seen clearly. The Fs of different schemes at different time under the condition without coverage change a lot and the values are low. The Fs of some schemes are less than 1.0. The calculation result is shown in Figure 15b and a shallow slip arc can be seen clearly. For scheme 5-a, the soil pore water pressure change hardly because of high permeability coefficient and Fs is stable hereby. Fs (Fso Fsno)/Fsno defined as increase amplitude of safety factor reflects the effect under the condition of coverage of different schemes, where Fso and Fsno are the results under the condition with coverage and without coverage respectively. Figure 16 shows the Fs -time relationship of different schemes. It is shown that the safety factors of slope can be increased by covering with geo-membrane of different schemes notably. The maximum increase amplitude obtained from scheme 5-b is 0.76; the second maximum increase amplitude obtained from scheme 6-b is 0.63. While the minimum increase amplitude obtained from scheme 2-b is 0.51. That is to say, the smaller the soil permeability coefficient and the greater slope ratio, the more effective the method with the coverage of geo-membrane. The results show that safety factors decrease obviously when wetting and increase when drying due to the effect of suction under the condition without coverage. The slope might become unstable during infiltration. Contrarily, the safety factors remain high during wetting-drying cycles under the condition with coverage, which means the slopes are
1295
1.630
(a) With coverage

1.035
the practical conditions and the soil water content should be controlled by taking relevant drainage. 3. The main function of covering slope surface with coverage is to prevent the soil water content from being changed by climatic factors. That means only using this method may not maintain the slope stability notably. It should be used together with the other strengthening method to achieve the best effect. 4. The slope stability is analyzed by a simply way taking into account of cracks of soil. While the cracks problem in expansive soils is very complicated, further studies need to be done on cracks correspondingly. ACKNOWLEDGEMENTS This study was sponsored by the National Technology R&D Program of China through Grant No. 2006BAB04 A10. REFERENCES
Bao C.G. (2004). Characters of unsaturated soil and the stability of expansive soil slope. Chinese journal of geotechnical engineering, 26(1): 115. Fredlund D.G. and Rahardjo H. (1997). Soil mechanics for unsaturated soils. Translated by Chen Z. Y. et al. China Architecture and Building Press. Huang R.Q. and Qi G.Q. (2002). The effect of unsaturated soil suction on slope stability. Journal of engineering geology, 10(4): 343348. Li S.L. (1992). Studies on the engineering geology of expansive soils in China. Jiangsu science and technology publishing house. Li Z.P. and Zhang M. (2001). Effects of rain infiltration on transient safety of unsaturated soil slope. China civil engineering journal, 34(5): 5761. Liu H.Q. (2008). Study on analysis method and influencing factors of expansive soil slope stability. Nanjing: Hohai University, 2008. Wu J.H., Yuan J.P. and Lu T.H. (2008). Stability analysis of unsaturated expansive soil slope. Rock and soil mechanics, 29(supp.): 363367. Yin K.L., Wang Y. and Tang Z.H. (2002). Mechanism and dynamic simulation of landslide by precipitation. Geological science and technology information, 21(1): 7578. Ying X.D. (2007). Study on stabilizing effect to expansive soft slope by manual bored concrete pries. Jiangsu construction, 6: 2022. Yuan J.P. (2003). Generalized model of fissures distribution and slope stability analysis for unsaturated expansive soils. Nanjing: Hohai University. Yuan J.P., Wu J.H. and Yin Z.Z. (2009). Field test for expansive soil slope reinforcement by covering geomembrane on slope surface. Proceedings of the 4th Asia-Pacific Conference on Unsaturated Soils, Newcastle, Australia: 347352. Zhang K.J. (2006). Study on failure mechanism and treatment technology of expansive soil slope in landscape area. Jiangsu construction, 4: 4447.
Figure 15. Most dangerous sliding surface and safety factor of scheme 1 (t 30 d).
0.8 1 2-a increase amplitude 0.7 2-b 3-a 3-b 0.6 5-a 5-b 0.5 6-a 6-b 0.4 5 10 15 20 25 30 35 (Day)
Figure 16. Safety factors increase amplitude-time relationship of different schemes.
stable. The results prove that the method of covering the slope surface with geo-membrane is beneficial to improve slope stability by numerical calculation. 5 CONCLUSIONS
1. It is shown that the dry-wet cycles caused by repeated rainfall and evaporation lead to cracks development. This is one of the most important factors causing slope failure. Geo-membrane is impervious to water in its life cycle (normally over 50 years). It can be used as permanent coverage on the slope surface to prevent not only infiltration but evaporation. If the soil water content can be kept stable, the swell-shrinking deformation will be avoided. This can prevent the slope surface from cracking and the slope can remain integrity. The results of both field tests and numerical calculation have shown that the slope stability was improved with the coverage of geo-membrane. 2. It is shown that under the condition with coverage, slope ratio, overburden thickness, rainfall intensity and duration has little effect on the slope stability. Coverage area and soil permeability coefficient affect more on slope stability. So the coverage area should be defined reasonably according to
1296
Embankments
Investigation of the water flow in an embankment built of organic soil: Experimental results and large scale field test
E. Birle, M. Boso & D. Heyer
Zentrum Geotechnik, Technische Universitt Mnchen, Mnchen, Germany
ABSTRACT: The paper presents experimental results of the flow of water in an embankment built of clay with organic content. The organic soil used for the field test contains geogenetically elevated concentrations of arsenic. As it is planned to use it for noise and view protection embankments, the leachate to be expected at the bottom of the embankments is of great interest. For the measurement of the water balance of the test embankment, extensive measuring systems are installed. With these the flow of water in the dry and wet compacted organic clay can be investigated. Based on the measurements, the water flow in the compacted organic clay is analyzed with respect to the compaction water content. 1 INTRODUCTION for the determination of seepage quantities a large scale, 5 m high, 30 m long and 25 m wide field test embankment was constructed. For the measurement of the hydrology of the embankment including seepage flow through the compacted organic soil extensive measuring systems were installed. Initially the paper briefly describes the construction of the test embankment, the properties of the organic soil and the compaction conditions and methods. Thereafter, the instrumentation of the measuring system is presented and the water flow within the test embankment is analysed on the basis of the recorded data. 2 SOIL PROPERTIES
Contaminated soils and recycled materials can be used for earth works if specified contaminant concentrations are not exceeded and if technical safeguards are considered in the design. Technical safeguards are necessary for reasons of soil and groundwater protection. The German Association for Research on Road Construction and Traffic Affairs (FGSV) developed various construction methods for the use of such materials (FGSV, 2008). An overview of the various construction methods is given by Heyer (2007). A technical safeguard or a technically safe installation is usually understood as consisting of a sealing system. If the contaminated soils or recycled materials are of low permeability, additional clay liners or geomembranes can be dispensed with. The design principle of such earth structures is to minimize the elution of contaminants. Extensive laboratory and field tests are currently being conducted to investigate the use of slightly contaminated organic soils for earth structures. In the area of the Erdinger Moos, where the Munich Airport is located, the top soil consists of clays with organic content. These organic soils contain geogenetically elevated concentrations of arsenic. The organic top soils borrowed in the course of building activities are to be reused for noise and view protection embankments. Laboratory investigations were performed to investigate the compaction properties as well as the mechanical and hydraulic behaviour of the organic soils under compacted conditions. The results show that these organic soils can principally be used for such embankments. As the permeability of the soil is low, additional sealing layers are not necessary. For the detailed investigation of the hydro-mechanical behaviour and
According to DIN 18196, the soil used for the field test is a clay with organic content (OH), with a liquid limit of 124% and a plastic limit of 44%. The organic content determined by the ignition method lay between 25% and 30%. For the determination of the compaction behaviour a Standard Proctor test was carried out. Other than in mineral soils no distinct Proctor Optimum could be determined. Furthermore, a dry density of approximately 0.8 g/cm3 in the water content, ranging between 55% und 75% was noted. The coefficient of permeability depends strongly on the compaction water content. For a water content of 55% a coefficient of permeability of k 5 108 m/s and for a higher water content of 75% a lower coefficient of k 5 109 m/s was determined. The natural water content in the field varies, depending on climatic conditions, between 60% and 85%. The Soil Water Retention Curve (SWRC) was determined for compacted samples with an
1299
II
100000 10000 1000 100 10 1 0 0.1 0.2 0.3
Matric suction - wetting path Total suction - drying path Matric suction - main wetting path Total suction - main wetting path
"dry" compacted material "wet" compacted material

10
Suction (kPa)
extensometers
1000000
Matric suction - drying path
0.4
0.5
0.6
0.7

test areas
Figure 1.
Soil water retention curve. Figure 2.

100% 90% volumetric water content 80% 70% 60% 50% 40% 30% 20% 10% 0% 0 10
vertical clay barriers
II
10
10
initial water content of 75% and a dry density of 0.8 g/cm3. The low suction range up to 100 kPa was investigated by means of tensiometers (matric suction). For the high suction range a chilled mirror dew-point hygrometer was used (total suction). Figure 1 shows the obtained drying and wetting curves. For the determination of the drying path, the samples were not previously saturated, the drying was rather started from the initial state after compaction. Starting from the initial state after compaction also a wetting path was determined. The main wetting path of the SWRC was investigated with samples previously dried from the initial state after compaction to a hygroscopic water content of approximately 3%. To reduce the time required for the determination of the SWRC, several samples were prepared and dried to the desired water contents. Thereafter the matric suction was measured by the tensiometers and the exact water content was determined by oven drying. 3 DESIGN OF THE TEST EMBANKMENT
Plan view of the test embankment.
calibration curve determined by Topp (1980) for mineral soils experimental data of the organic clay approximation by the logarithmic function y = 0.2135 ln(x) - 0.1416
water content range in field
20
30 40 50 60 dielectric constant ka
70
80
10
90
Figure 3.
Calibration curve of the TDR-probes.
The principle design of the embankment is shown in Figure 2. As with the construction method E recommended by the German Association for Research on Road Construction and Traffic Affairs (FGSV, 2008) the compacted organic soil is covered only by a drainage and top layers without an additional sealing layer. The test embankment is 5 m high, 30 m long and 25 m wide. As the hydraulic behaviour depends strongly on the compaction water content, it was decided to divide the test embankment in two parts. On the west side the soil was compacted with an initial water content of approximately 55% and on the east side with an initial water content of approx. 75%. The compaction was performed by means of a heavy polygonal drum roller compactor (25 tons). With it a dry density of the dry and wet compacted material between 0.75 g/cm3 and 0.85 g/cm3 was achieved. After compacting the organic material the drainage mat was installed. For the non-compacted top layer, a clay with organic contents of similar properties was used.
On the south side of the test embankment, two areas of 5 m width and 12.5 m length were separated from the remaining embankment by vertical clay barriers of very low permeability (see figure 1). In these areas, two-dimensional flow conditions can be assumed. For the analysis of the seepage extensive measuring systems were installed in these test areas. The leachate at the bottom of the embankment was collected by a drainage mat and an underlying geomembrane. A tube connected to the geomembrane discharges the water to a gauge well situated on the south side of the embankment, where the leachate discharge is measured by electronic balances. The interflow in the drainage mat covering the compacted organic soil is collected at the bottom of the test areas and measured in the gauge in a like manner. The flow of water in the compacted organic soil is monitored by means of tensiometers and TDR-probes. The position of these sensors after compaction is shown in Figure 4. For the determination of the volumetric water content by means of the signals of the TDR-probes, a calibration of the TDR-probes had been performed. The soil specific calibration curve is shown in Figure 3. For comparison, the calibration curve of Topp (1980) for mineral soils is also given in Figure 3.
1300
legend: tensiometer TDR-probe south top layer 2 1: drainage mat compacted clay drainage mat and geomembrane
rainfall
radiation temperature humidity
evapotranspiration clay barrier infiltration surface run off
leachate A
ground water
interflow (drainage mat, top layer)
Figure 4. Cross section of the test embankment with the final arrangement of the tensiometers and TDR-probes (left section) and the description of the water balance (right section).
For the simulation of the water balance of the test embankment the climatic conditions have to be known. The climate is measured by a meteorological station close to the embankment. It consists of several sensors measuring the air temperature and humidity, the effective radiation, the wind velocity and direction as well as the amount of precipitation. For the evaluation of the water flow in this paper only the measured data of precipitation are used. The vertical deformations of the test embankment are measured by means of four extensometers, installed in two boreholes; one in the wet and one in the dry compacted material (see Figure 2). The water balance of the test embankment is shown schematically in Figure 4. 4 RESULTS AND ANALYSIS
The precipitation and the seepage water at the bottom of the wet and the dry side recorded between October 2008 and September 2009 are shown in the Figures 5 and 6. It can be seen that until September 2009 a significant leachate was measured only at the bottom of the wet side. Comparatively large fluxes arose directly after compaction of the clay. Taking the measured pore water pressures into consideration, a consolidation process was initiated by the compaction of the wet material. In some areas the water content of the wet compacted material was too high, so that excess pore water pressures were generated by the heavy compaction equipment. As the consolidation diminishes, the leachate at the base of the wet compacted material reduces significantly. The consolidation is assumed to be completed at the end of April and henceforth the leachate discharge takes on an approximately constant value. Between May and September 2009, when the consolidation process was already completed, a relatively constant flux of 0.03 mm/d was measured at the bottom of the wet side. This corresponds to approx. 1.3% of the precipitation measured during this period of time.
Figure 7 shows the accumulated water volumes of precipitation, the interflow in the drainage mat and the leachate at the bottom of the wet and dry compacted material. The measured leachate at the bottom of the wet side results mainly from the consolidation of the soil. As the top layer is of low permeability, the interflow in the drainage layer above the wet compacted material is small. So far no significant leachate discharge was measured at the bottom of the dry side. Similarly, an interflow in the drainage mat placed on top of the dry compacted clay was measured only after heavy rainfall. This is caused by the rather low water content of the compacted clay that absorbs the water from the drainage mat. Correspondingly, the measurements of the TDRprobes located in the dry compacted clay show an increase of the water content in time (see Fig. 8). As the water is absorbed by the clay from the drainage mat, the sensors located close to the drainage mat and thereafter those sensors placed deeper in the embankment show an increase in the water content. This is clearly visible from the TDR-probes MW1A, MW1B, MW1C, MW1D and MW1E, which are located approximately one upon the other in the section A-A indicated in Figure 4. In this way the TDR-probe MW1D located at a distance of approx. 20 cm beneath the drainage mat indicated an increase of the water content in November 2008. (Due to problems with the data acquisition system the exact date when the water content started to increase is not known.) In December a further increase of the water content of the TDR-probe MW1D was observed. A few days after the reaction of this sensor the TDRprobe MW1C located at a distance of approx. 50 cm beneath the drainage mat also displayed an increase in the water content. However, both sensors located close to the drainage mat indicated that they were not fully saturated at any time. Moreover, a volumetric water content of approx.
1301
45 40 35
measured precipitation at the test embankment precipitation at the Munich airport due to the German Meteorological Service
precipitation (mm/d)
30 25
irrigation failure of the precipitation measurement device
20 15 10 5 0 29.10.08
28.12.08
26.02.09
27.04.09
26.06.09
25.08.09
Figure 5. Measured precipitation at the test embankment and data of the German Meteorological Service for the Munich airport.
0.25
leachate at the bottom (mm/d)
0.2
temporarily failure of the measurement device
dry side
wet side
0.15
0.1
0.05
0 29.10.08
28.12.08
26.02.09
27.04.09
26.06.09
25.08.09
Figure 6. Measured leachate at the bottom of the wet and dry side between October 2008 and September 2009.
700
20 18
600 Leachate at the bottom (wet side) 500 400 300 200 100 0 29.10.08 Precipitation Interflow in the drainage mat (wet side)
16 14 12 10 8
Interflow in the drainage mat (dry side) Leachate at the bottom (dry side)
6 4 2 0
28.12.08
26.02.09
27.04.09
26.06.09
25.08.09
Figure 7. Accumulated water volumes of the wet and dry compacted clay.
53%, which corresponds to a degree of saturation of approx. 90%, was rarely exceeded. The TDR-probe MW1E located in the top layer approx. 20 cm below the ground surface, reveals the changes of water content of the top layer depending on the climatic conditions. Due to the heavy precipitations in June and July a strong increase in the volumetric water content was observed. During the following dry August and September 2009, the volumetric water content decreased considerably (20%) to a value of 30%. From other sensors it is known that the volumetric water content in the top layer, even in the lower part close to the drainage mat reached
Cumulative leachate/interflow (mm)
Cumulative precipitation (mm)
1302
60
MW1A (4.5 m beneath drainage mat)

55
MW1C (50 cm beneath drainage mat)
MW1D (20 cm beneath drainage mat)
volumetric water content (%)
50 45 40 35 30 25 20 29.10.08 28.11.08 28.12.08 27.01.09 26.02.09 28.03.09 27.04.09 27.05.09 26.06.09 26.07.09 25.08.09 24.09.09
MW1B (2 m beneath drainage mat) MW1E (top layer)
Figure 8. Volumetric water content at different depths beneath the drainage mat in section A-A (dry side).
values close to 35% in September. In comparison to this the TDR-probe MW1D in the same period of time (between August and September 2009) recorded a decrease of the volumetric water content of only approx. 5%. Since the top layer and the compacted soil have very similar hydromechanical properties (including the SWRC), a stronger decrease of the water content in the compacted clay could have been expected. The difference in the water contents between the top layer and the compacted clay is caused by the drainage mat acting as a capillary barrier between the top layer and the compacted layer thereby interrupting the capillary water rise from the compacted clay to the top surface. Thus, the water content of the compacted core decreases in principle only due to a leachate at the bottom or due to evaporation from the compacted clay into the drainage mat. The latter is supposed to be small, since no significant circulation of air in the drainage mat is assumed. On considering all sensors installed in the dry compacted clay, the water content increased only in the upper area of the compacted clay to a depth of approx. 1 m below the drainage mat. As a significant leachate discharge is expected at the bottom of the dry side only if water seeps from the top to the bottom of the embankment, making it important to know how the saturation process progresses or slows down. For this more data is needed. As the suction decreases with increasing water content, it is possible that seepage into the dry compacted material diminishes with time. The measurements of the TDR-probes MO1A, MO1B, MO1C and MO1D, which are installed in the wet compacted material in the section A-A (see Fig. 4) approximately one upon the other, are
presented in Figure 9. It can be seen that after the compaction in the upper area of the clay a volumetric water content of approx. 48% was reached (see MO1C and MO1D), which is considerably smaller than the desired volumetric water content of 60%. The desired value was reached at the TDR-probes MO1 A and MO1B. Since a volumetric water content of approximately 60% is related to an almost saturated state, no significant change of the water content was measured by the TDR-probes MO1A and MO1B until September 2009. In comparison the TDR-probes MO1C and MO1D indicated a significant increase of the volumetric water content in July 2009. In this month a maximum water content of 65% was reached. During the dry periods of August and September 2009, the water content decreased once again. This was mainly caused by seepage to the bottom and not by a capillary water rise to the top surface, since the drainage mat acts as a capillary barrier. The TDR-probe MO1B to date has shown no change of water content even in dry periods. Since the effect of the evapotranspiration on the water content of the compacted core can be neglected, in a theoretical state of equilibrium (e.g. after a long period of time without precipitation and evapotranspiration) the matric suction in the compacted clay should correspond to the elevation head. Considering the measuring point MO1B located 2.5 m above the geomembrane in a state of equilibrium a matric suction of 25 kPa should arise. Under consideration of the drying path of the SWRC it can be seen that this value still results in a volumetric water content of approx. 60%. That is why no significant change of water content has been indicated by this TDR-probe and is not expected in future.
1303
70
MO1C (60 cm beneath drainage mat)

65
MO1B (2 m beneath drainage mat)
volumetric water content (%)
60
MO1A (4.5 m beneath drainage mat)
55
MO1C (60 cm beneath drainage mat)

50
MO1D (30 cm beneath drainage mat)

45 29.10.08 28.11.08 28.12.08 27.01.09 26.02.09 28.03.09 27.04.09 27.05.09 26.06.09 26.07.09 25.08.09 24.09.09
Figure 9.
Volumetric water content at different depth beneath the drainage mat in section A-A (wet side).
CONCLUSIONS
By means of a large scale field test, it could be demonstrated that slightly contaminated soils can be used for earth constructions without additional sealing systems if they possess a low permeability. By compacting the organic clay used in the field test at water contents dry and wet of the Proctor Optimum, the influence of the compaction water content on the hydro-mechanical behaviour could be studied. The measurements of the wet compacted side of the test embankment showed that a significant leachate discharge can occur when the compaction water content is slightly higher than desired and pore water pressures are created, which lead to a consolidation of the material. As a rule, when the soil is compacted wet of the Proctor Optimum, nearly saturated conditions can be reached. In this state no more water can be stored by the compacted soil and water infiltrating due to precipitation leads to a leachate discharge at the bottom of the embankment. In comparison to the wet compacted soil the degree of saturation is much smaller when the soil is compacted at a water content dry of the Proctor Optimum. In this case the seepage water will be absorbed by the compacted soil before a leachate discharge at the bottom of the embankment takes place. In the field test such a saturation process of the dry compacted clay was clearly observed. However, during the first ten months no significant leachate discharge was measured at the bottom. Thus, for the determination of the future seepage fluxes it is important to know, how the saturation process progresses or slows down. This depends on the effects of the evapotranspiration
on the compacted clay. If a water transport from the compacted soil to the top layer due to capillary forces resulting from evapotranspiration can take place, the water content of the compacted soil will be reduced. In the field test the compacted clay is covered by a drainage mat, which acts as a capillary barrier as it interrupts the capillary flow to the top surface. As a result only a small decrease of the water content of the dry compacted clay was measured in dry periods. However, the drainage mat dissipates the seepage water infiltrating due to precipitation and thereby reduces the seepage into the compacted clay. By means of numerical investigations it shall be investigated under which conditions the application of a drainage layer is useful. In this context the hydraulic properties of the top layer and the climatic conditions have to be taken into consideration.
REFERENCES
FGSV (2008): Merkblatt ber Bauweisen fr Technische Sicherungsmanahmen beim Einsatz von Bden und Baustoffen mit umweltrelevanten Inhaltsstoffen im Erdbau (MTSE), Forschungsgesellschaft fr Straen- und Verkehrswesen, Arbeitsgruppe Erd- und Grundbau, Ausgabe 2008. Heyer, D. (2007): Construction methods of technical safeguards regarding earth constructions with contaminated soil and demolition waste, Proceedings Sardinia 2007, XI International Waste Management and Landfill Symposium, 15 October 2007. Topp, G.C., Davis, J.L., Annan, A.P. (1980), Electromagnetic Determination of Soil Water Content: Measurements in Coaxial Transmission Lines, Water Resour. Res., 16(3), 574582.
1304
Suction changes during the construction of an embankment from A10 Motorway in Portugal
R. Cardoso, E. Maranha das Neves & P. Almeida Santos
Department of Civil Engineering and Architecture, Instituto Superior Tcnico, Lisbon, Portugal
ABSTRACT: According to a research project, an embankment from A10 Motorway, in Portugal, built with marls was instrumented during its construction. The vertical displacements and the changes in the water content were recorded. Suction profile in depth was estimated through the measurement of the water content. During the construction of the embankment, readings show that the suction distribution in depth tends to a single value. CODE_BRIGHT was used to reproduce in situ readings during the construction. Experimental data from an extensive set of tests performed on samples of compacted material and realistic climate actions were used in the model. For the mechanical behaviour of the compacted marls the Barcelona Basic Model was adopted. The results from the simulation reproduce quite well the values measured in the embankment. 1 INTRODUCTION (UPC-DLT 2002) was used to simulate the construction of the embankment. Climate actions were considered as the major cause of the deformations. Barcelona Basic Model was adopted as constitutive model for the materials. The numerical results concerning the evolution during the construction of suction and vertical displacements show a good agreement with the values measured in the field. 2 DESCRIPTION OF THE EMBANKMENT
Marls in general are classified as hard-soil/soft rocks and exhibit evolutive behaviour since their mechanical and hydraulic properties change markedly due to wetting-drying cycles and to other weathering processes. Crack opening and/or loss of bonding are associated with these processes, having negative impact on the strength and compressibility of the material. Several embankments from A10 Motorway (Portugal) were built with marls. Relatively large fragments remain after the compaction processes. The degradation of their mechanical properties can have strong effect on global behaviour and be responsible for large deformations. The particularity of the embankments investigated is that the marls used to build the shoulders were treated with lime (3.5% lime, in weight). The treatment intended to reduce the swelling potential of the marls, protecting them from atmospheric actions therefore reducing the influence of these main causes of marl degradation. Vertical deformations and the evolution of the water content in embankment AT1 were measured with instruments installed during the construction. Suction was obtained from water content values through the water retention curve. The water retention curve and other useful data were obtained from an extensive set of laboratorial tests performed to characterize the hydro-mechanical behaviour of the marls with and without lime treatment. Experimental data provided the information for model calibration. Program CODE_BRIGHT
A simplified profile of embankment AT1 is presented in Figure 1. It has the particularity of being built in a slope and is relatively high (near 10 m at the pavement axis). As shown in Figure 1, a zoned profile was adopted with compacted marls in the core and compacted marls treated with lime in the shells (with 5 m thickness). Placement with a vibratory sheepstoot roller reduced the dimension of the marl fragments (specific energy equivalent to heavy compaction) and compaction in the wet side of optimum (interval [wopt,wopt2%]) was adopted
service access inclinometer
2 3
O +
pavement
colector collector
Marls treated with lime Solo-cal Rockfill toe 2 3 collector
Marls geotextile collector

0 5 10 20 25 m
Figure 1. AT1 embankment.
1305
in the construction of the embankments. Design provisions also established a minimum of 95% of relative compaction. The compaction process was prescribed to accelerate the physical degradation of the marl during the construction, therefore reducing the settlements that could affect the serviceability of the entire structure. Other usual provisions were adopted to prevent water access into the embankment core, such as the installation of adequate drainage systems in the foundation and at the pavement levels (Fig. 1). The instruments were installed in AT1 during its construction. Data collected concerns this period and the two years after. Two vertical sections were instrumented at the locations PK1 150 and PK1 250 identified in Figure 2 (100 m distance between them). One gutter for measuring displacements and seven sensors (Figure 3) for measuring water content were placed in each section. To avoid interferences between the different instruments, the sensors for soil suction measurement are installed at 3 meters distance (in longitudinal direction) from the gutters.
SENSORS INSTALLED
For measuring the vertical deformations with large precision it was adopted INCREX instruments (INCRemental EXtensiometers, Interfels 2004) (Figure 4). Large precision was required to detect small deformations due to suction changes. The INCREX system is based on the measurement of the relative vertical displacements of magnetic rings spaced exactly one meter in the installation (Figure 4, left). The rings are installed outside the gutter and are free to slide. Since they are connected to the soil (Figure 4, right), their sliding displacement is equal to the one of the soil. The system required to measure the magnetic field allows a precision higher than the one obtained when standard measuring systems are used (0.02 mm for INCREX, and 0.5 mm for more general measurement equipment where tape is used). The evolution of soil suction during the construction was determined by measuring the water content and the relative humidity of the soil. Sensors ECH2O (ECH2O 2004) (Figure 5.a) were used, which measure water content. Vaisala HMT 337 sensors were also installed (Figure 5.b), which measures relative humidity and temperature (HMT337 2004). They were installed in the soil
gutter
PVC protection while the embankment was not built
1m
Figure 2.
Location of the intrumented profiles.
1m
rings
Built zone: Cement and bentonite around the ring (0.8m) and soil compacted manually between the rings
Figure 4.
INCREX sensors.
Figure 3. Profile instrumented with sensors ECH2O (DECAGON) and HMT330 (VAISALA).
Figure 5. Sensors installed: a) ECH2O (water content); b) HMT330 (relative humidity and temperature).
1306
1700 1600
V = 19,112w + 1353,2 2 R =1
18.2 kN/m3 17.2 kN/m3

V = 20,465w + 1224,9 R2 = 0,9998
Measurement (V)
1500 1400 1300 1200 1100 1000 900 0 5 10 15 20

V = 20,103w + 907,84 R2 = 0,9913
16.0 kN/m3 14.7 kN/m3 13.1 kN/m3 12.0 kN/m3
Figure 6.
Calibration of sensors ECH2O.
dry volumetric weight (kN/m )
treated with lime at 1.5 m depth. Temperature was measured because its control is very important to calculate suction from RH values. The ECH2O sensors were placed in depth, five in the core made of marls and two in the shoulders (Figure 3). Automatic recording of the water content readings (3 readings/day) was performed and recorded in the dataloggers (Figure 5.a right). The sensors were calibrated in the laboratory before being installed. Their calibration was fundamental because the readings depend on the resistivity of the surrounding soil which depends, among other factors, on the density of the material. Calibration was done only for the compacted marls. For the ECH2O sensors, calibration was done by measuring their voltage when installed in samples compacted with different water contents for the same dry unit weight. The relationship between voltage and water content (Figure 6) for each dry unit weight shown to be linear with similar slope (average value 19.9). The voltage for null water content increases with the dry unit weight. It was calibrated with in situ measurement of voltage to find the real dry unit weight of the soil when each sensor was installed. Unfortunately, Vaisala sensors were saturated accidentally during the installation in the embankment and the accuracy in the readings of the relative humidity was permanently lost. Data collected during the construction allowed identifying an initial period where water content of the soil surrounding the sensor was in equilibrium with the one of the layer where it was installed. After this period it was possible to detect the homogenization in depth of the water content (between 12% and 16%). 4 CHARACTERIZATION OF THE MATERIALS
was determined (Maranha das Neves & Cardoso 2006). Carbonates (calcite mainly) are present, as also as quartz, mica, dolomite, feldspar, clays and a very small percentage of organic matter (02%). Among the clay minerals there is almost no smectite and the existing expansive minerals are chlorite and gypsum. The presence of these minerals may explain the degradation (mainly cracking and disaggregation) and the moderate expansive behaviour observed in this material when fully saturated. Laboratory tests on marl samples with different weathering degrees were performed (Maranha das Neves & Cardoso 2006). The samples were compacted in conditions similar to those adopted in the field. For all cases, the reconstituted marl (fraction with dimensions D 0.20 mm) can be classified as low plasticity clay (CL), which is in accordance with the nature of the clay minerals found. The Atterberg limits that allowed this classification and other relevant geotechnical properties of Abadia marls are presented in Table 1. Suction was measured in samples prepared with different water content and dry density (Maranha das Neves & Cardoso 2006) and allowed defining the suction chart presented in Figure 7. In the interval of compaction prescribed (rectangle identified
Table 1. Some geotechnical properties of Abadia Marls. Marls without treatment Liquid limit, wL Plasticity index, PI Weight density of solid particles In situ void ratio In situ porosity Saturated water content, wsat Marls treated with 3.5% of lime Liquid limit, wL Plasticity index, PI 36%52% 20%25% 27.5 kN/m3 0.220.55 18%33% 8%22% 36% 7%
20 19 18 17
3
Standard comp. Modified comp. Full saturation curve Analyzed points Construction conditions Sr=100%
16 15 14 13 0 2
s=56MPa s=2MPa s=10MPa s=5MPa Sr=20% Sr=40% s=0,5MPa s=0,8MPa s=1MPa Sr=60% Sr=80% Sr=90%
8 10 12 14 16 18 20 22 24 26 28 30 32 34 36
Abadia marls, Upper Jurassic in age, were used in the construction of the embankments. The mineralogical composition of unweathered marls
water content, w (%)
Figure 7. Suction measured in tests of compacted marls.
1307
in the figure as Construction conditions) the soil has high degree of saturation because low suctions were measured (average value 0.8 MPa). The effects of lime addition on the hydromechanical behaviour of Abadia marls was investigated by comparing the results of tests performed on samples with and without treatment. The study was extended to different curing periods but this paper only presents the results in tests performed in samples with a curing period of one week, assumed to be valid during the construction of AT1. According to Table 1, the treatment reduces the plasticity of the marls (due to the reduction of the wL thus on PI), improving the workability of the material and changing the classification of the fine fraction from CL to ML. Swelling tests were performed on samples of compacted marls after the treatment. As expected, the swelling deformations were lower after the lime treatment. The elastic compressibility for suction changes, s, was also reduced with the treatment (from 0.008 to 0.003). Several other tests were performed to compare the hydro-mechanical properties of the compacted marls before and after the treatment (Lynce de Faria 2007 and Godinho 2007) and to find the constants for BBM calibration for the marls. They show that the treatment decreased the elastic compressibility for isotropic compression, (in Table 3) as well as the elastoplastic compressibility measured in oedometer tests. These results can be explained by the formation of cementation minerals after the lime hydration (Locat et al. 1990, for example). This cementation is also responsible for the increase of the measured saturated shear stiffness and strength. Changes in soil structure and the presence of crystals or cements generated in the hydration of the lime are reported by several authors (Clare & Cruchley 1957, among others). They could be identified in other tests reported by Maranha das Neves & Cardoso (2006) such as mercury intrusion porosimetry test, saturated permeability and changes in the water retention curve (Figure 8). The curves are presented in Figure 8 were fitted by Equation (12) (van Genuchten 1980),
Pg Pl 1 Se 1 P
1000.00 Total suction (MPa) 100.00 10.00 1.00 0.10 0 2 4
Drying - Marl Wetting - Marl WRC-Drying WRC-Wetting Drying - Marl treated Wetting - Marl treated WRC-Drying WRC-Wetting
6 8 10 12 14 16 18 20 22 water content, w (%)
Figure 8. Water retention curves of the marls with and without the treatment.
Figure 9. Model of AT1 (Almeida Santos 2009): a) geometry and b) FE mesh.
for the drying branch and P 0.08 MPa, 0.20 for the wetting branch). 5 NUMERICAL MODEL AND RESULTS
(1)
where Se is the saturation degree at the current liquid pressure, Pl, Pg is the gas pressure (assumed to be the atmospheric pressure, 0.1 MPa), P is the pressure corresponding to the air entry value and is a fitting parameter (marls without treatment: P 0.3 MPa and 0.23 for the drying branch and P 0.18 MPa, 0.23 for the wetting branch; marls after the treatment: P 0.51 MPa, 0.25
The construction of the embankment was simulated using CODE_BRIGHT (UPC-DLT 2002). Thermo-hydro-mechanical coupled analysis was performed to allow the inclusion of climate as input action besides the increment of vertical stress during the construction of successive layers. The inclusion of suction as a state variable requires the use of unsaturated soils constitutive models. Barcelona basic model, BBM (Alonso et al. 1990) was adopted. For the marls the complete elastoplastic framework was used. Only the elastic part was used for the marls treated with lime due to the higher yielding stress found in the tests for this material, higher than the stresses installed in the field. Figure 9 presents the geometry used in model. The points where vertical displacements and
1308
Table 2.
Initial conditions for each construction layer. Soil 1 2 17.0 18.9 0.62 0.20 3 17.0 18.9 0.62 0.20 4 17.1 14.4 0.61 0.67 5 15.4 13.5 0.79 2.15 6 16.2 13.6 0.70 1.38 7 16.4 14.0 0.68 1.11 8 16.4 14.0 0.68 1.11 9 16.4 14.0 0.68 1.11 Soil treated Identical in all layers 16.3 14.5 0.68 1.07
Dry unit weight (kN/m3) Water content (%) Void ratio Suction (MPa)
17.0 18.9 0.62 0.20
0 -1 -2 -3 -4 -5 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15 -30 20-05-2005 14-06-2005 20-06-2005 23-06-2005 15-07-2005 21-07-2005 -20 -10 0
-30
12-05-2005 20-05-2005 14-06-2005 15-07-2005 21-07-2005 31-08-2005
0 -1 -2 -3 -4 -6 -7 -8 -9 -10 -11 -12 -13 -14 -15 -20 -10 0 depth of the rings (m) -5
a)
vertical displacement (mm)
b)
vertical displacement (mm)
Figure 10. Vertical displacements (cumulative values): a) calculated; b) measured.
suctions were measured are the vertical alignment shown. The bed rock where the embankment was founded was also included in the model. The different steps of construction (10 steps in total, corresponding to the construction of 9 layers 2 m thick and the construction of the pavement) are also in Figure 9. It was assumed one weak for each construction step. Loading included the weight of the soil, of the rockfill at the toe (50 kN/m2) and of the bituminous pavement (1.2 kN/m2). The bituminous pavement was assumed to be impervious. The initial porosity and suction installed in each construction layer was defined with the dry unit weight and water content measured in the installation of each ECH2O sensor (Table 2). Realistic climate data were obtained through a weather station located near the embankment (Almeida Santos 2009). The parameters necessary for the model are presented in Table 3. They were obtained from the laboratory tests previously presented. The vertical displacements calculated during the construction are presented in Figure 10.a. These values can be compared with the vertical
displacements measured in the embankment during the construction presented in Figure 10.b (PK1 150). Similar path was measured in both profiles and for both instruments, the readings were interrupted at the same construction stage because of one accident that destroyed the gutters. For the two cases analysed, the last four meters correspond to the foundation. The major differences found in the plots are in the shape of the curves mainly in the construction of the first layers. The smaller displacements measured are explained by the absence of readings in these layers because of the process adopted for the installation of the gutters. Nevertheless these differences the comparison of the plots from Figure 10 shows that the maximum displacement occurs at 21st July 2005 (after the end of the construction), approximately at 5 meters depth. The maximum values are very similar (21 mm measured and 18 mm calculated). Suction evolution during the construction was quantified through changes in the water content. The suction values from output were converted into water content values using the water retention curve of the material (Fig. 8). They are in Figure 11.a. The water content values measured are in Figure 11.b (also in PK1 150). In both plots it is evident that the distribution in depth belongs to similar intervals (13% and 19%). In situ, the evolution of the values tend to a final homogeneous distribution, expected after equilibrium. This is a slow process because of the low permeability of the materials and for this reason equilibrium was not observed at the end of the period studied. In the model the convergence into a single value is not so evident. This is possible because the model considered that the entire layer at each construction step had the same water content while this is not necessarily true in the field since sensors were installed in a zone manually compacted therefore different than the surrounding soil. Similar results are found if suction values are compared instead of water content (Almeida Santos 2009).
depth (m)
1309
water content, w (%) 12 0.0 -1.5 -3.0 depth of the sensor (m) -4.5 -6.0 -7.5 -9.0 -10.5
t=0 (inst #1) t=7 days (inst #2) t=18 days (inst #3) t=22 days (inst #4) t=28 days (inst #5) t=33 days (inst #6) t=42 days (inst #7) t=1month (72 days) t=2 months (102 days) t=3 months (132 days) t=4 months (162 days) t=5 months (192 days) t=6 months (222 days) t=7 months (252 days) t=8 months (282 days)
14
16
18
20
0 0.0 -1.5 -3.0 depth of the sensor (m) -4.5 -6.0 -7.5 -9.0 -10.5
8 10 12 14 16 18 20 22
t=0 (inst #1) t=7 days (inst #2) t=18 days (inst #3) t=22 days (inst #4) t=28 days (inst #5) t=33 days (inst #6) t=42 days (inst #7) t=1month (72 days) t=2 months (102 days) t=3 months (132 days) t=4 months (162 days) t=5 months (192 days) t=6 months (222 days) t=7 months (252 days) t=8 months (282 days)
line connecting the last readings of the sensors before they stop working
a)
-12.0
b)
-12.0
Figure 11. Table 3.
Evolution of the water content in the embankment: a) calculated; b) measured. Calibration of the constitutive models. Meaning Elastic compressibility index for isotropic stress changes Elastic compressibility index for suction changes Elastoplastic compressibility index for isotropic stress changes Parameter defining maximum stiffness with suction Parameter controlling the rate of stiffness increase with suction Saturated isotropic yielding stress Reference isotropic stress Parameter describing the increase in cohesion (apparent) with suction Tension resistance for saturated conditions Shear modulus Poisson coefficient Slope of the critical state line Adjustment parameter of the water retention curve Adjustment parameter of the water retention curve Intrinsic permeability (isotropic) Thermal conductivity in dry and in full saturated conditions Marl 0.009 0.0013 0.20 0.002 MPa1 100 77 0.001 0 20 0.3 1.0 0.23 (wetting) 0.18 MPa 9 1021 m2 2 WmK1 Marl treatead 0.003 0.0003 0.001 100 kPa 70 0.3 1.1 0.20 (wetting) 0.08 MPa 5 1021 m2 2 WmK1
Parameter s (0) r po* pc K c0 G M P ko DRY SAT
CONCLUSIONS
The evolution of water content (suction) and the deformations of the embankment during its construction calculated with BBM were relatively well predicted as they were similar to those measured in the field. Further study is required to estimate the deformations during service, based in the comparison with in situ readings. For a more realistic analysis to estimate the long term deformations of the embankment, a different
model must be used. As a matter of fact, the model does not cover the calculation of creep settlements mainly due to progressive degradation of the fragments of marl. ACKNOWLEDGEMENTS The authors would like to thank to BRISA for allowing the publication of the results presented, to Professor Eduardo Alonso, Dr Enrique Romero
1310
and Dr Sebasti Olivella from the Universidad Politecnica de Catalua, Barcelona, for allowing some of the experimental tests presented and the use of Program Code Bright. Acknowledgment is also due to Dr Ferrer (Lab Ferrer, Spain) for helping during the installation of ECH2O sensors and to the Portuguese Foundation for Science and Technology, FCT, for the financial supports that allowed this study (SFRH/BD/25846/2005, POCTI/ECM/59320/2004). REFERENCES
Almeida Santos, P. (2009). Behaviour of an embankment built with compacted evolutive material. Bologne MSc thesis, Instituto Superior Tcnico (in Portuguese). Alonso, E.E., Gens, A. & Josa, A. (1990). A constitutive model for partially saturated soils. Gotechnique, 40(3), pp. 405430. Clare, K.E. & Cruchley, A.E (1957). Laboratory experiments in the stabilization of clays with hydrated lime. Geotechnique, vol 7, pp. 97110. ECH2O (2004). Eco-sensors ECH2Osoil moisture measurement. Decagon Devices, Inc., Pullman, USA. Fredlund, D.G. & Rahardjo, H. (1993). Soil Mechanics for Unsaturated Soils. John Wiley and Sons, New York.
Godinho, N. (2007). Study on the hydro-mechanical behaviour of marls treated with lime. Bologne MSc thesis, Instituto Superior Tcnico (in Portuguese). HMT337 (2004). Vaisala HUMICAP Humidity and Temperature Transmitter, Series HMT330, USERS GUIDE. Interfels (2004). Boart Longyear Interfels CatalogueGeotechnical Instrumentation, Interfels. Locat, J., Brub, M-A. & Choquete, M. (1990). Laboratory investigations on the lime stabilization of sensitive clays: shear strength development, Can. Geotec J., 27, pp. 294304. Lynce de Faria, F. (2007). Wetting effects on the behaviour of an embankment from A10 Highway. Bologne MSc thesis, Instituto Superior Tcnico (in Portuguese). Maranha das Neves, E. & Cardoso, R. (2006). Research Project for BRISA on the mechanical behaviour of embankments from A10 Motorway. Dep. of Civil Engng, Report ICIST EP 23/06 and EP 13/08. Instituto Superior Tcnico. UPC-DLT (2002). CODE_BRIGHTUsers Guide. Dep. Ing. del Terreno, E.T.S. Ingenieros de Caminos, Canales y Puertos de Barcelona, Univ. Polit. de Catalua, Spain. van Genuchten, M.T. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44, pp. 892898.
1311
Influences of spatially varying porosity on unsaturated flow through earth structures

T.M.H. Le, D. Gallipoli & S. Wheeler M. Snchez
Department of Civil Engineering, University of Glasgow, Glasgow, UK Zachry Department of Civil Engineering, Texas A&M University, Texas, US Formerly Department of Civil Engineering, University of Strathclyde, Glasgow, UK
ABSTRACT: Soil properties of both natural deposit or manmade fills are inherently variable in space. However, in geotechnical practice, the assumption of homogeneity within individual material zones is almost invariably introduced to simplify analysis. This study demonstrates the application of a stochastic approach to model random soil heterogeneity and to investigate its influence on hydraulic properties by analysing the case of seepage through a flood defence embankment with a random spatial distribution of porosity. In particular, a dependency on porosity is introduced inside the van Genuchten model, which is then used to describe water retention and permeability in the unsaturated domain. A Monte Carlo analysis is undertaken by using the finite element program CODE_BRIGHT to predict seepage through embankments with different random porosity fields. The study demonstrates that flow within a heterogeneous soil mass shows a preference to follow more porous paths. As a result, the unrealistic assumption of material homogeneity can lead to over-prediction of the time to reach steady state and under-prediction of the corresponding total flow rate across the embankment during wetting from an initial unsaturated state. 1 GENERAL INSTRUCTIONS unsaturated flow when the degree of saturation is equal to one at very low values of suction. Research into unsaturated flow was further developed by various authors (Gardner, 1958; Taylor and Brown, 1967; Freeze, 1971; Neuman, 1973; Lam and Fredlund, 1984). In a paper by Freeze (1971), the author developed a finite element model where ground surface, instead of free surface, was treated as the border of the flow region. This work clearly demonstrated the importance of accounting for unsaturated flow and showed that the predicted position of the free surface (defined as the locus at zero suction) could be rather different if absence of flow through the unsaturated region was assumed. If wetting from an initial unsaturated state, the rate of advancement of the wetting front depends strongly on the water retention characteristics, unsaturated hydraulic conductivity and initial soil moisture, which also influence the time it takes for seepage to reach steady state. The study of groundwater seepage is also usually simplified by assuming homogeneous hydraulic properties in each material zone of the soil domain. It is, however, well known that both natural soils and manmade fills can be inherently variable in space. Porosity, for example, is easily recognised as non-uniform over individual material zones with a spatial variation that often incorporates a strongly
Saturated and unsaturated soil regions exist in all earth structures exposed to the atmosphere such as embankments, earth dams and canal dykes. However, when studying seepage through earth structures, engineering analysis often focuses on the saturated region while flow through the superficial unsaturated layer is frequently ignored due to two reasons. For a given soil, the saturated permeability is much higher than the unsaturated permeability, thus is often assumed that most of the water flows through the saturated region. Also, unsaturated seepage models require handling of highly non-linear equations and, hence, advanced numerical techniques and computational resources. The extent of the unsaturated region can, however, be dominant especially in dry environments where earth structures are subjected to high suctions at the boundary; in such cases, the analysis cannot ignore unsaturated flow without resulting in erroneous predictions. Flow through unsaturated soils has been investigated as early as 1930s, when Richards (1931) suggested an extension of Darcys Law for saturated soils to take account of the dependency of hydraulic conductivity on degree of saturation. In this context, flow in saturated soils is a special case of
1313
random component. Porosity is also related to hydraulic permeability through the well-known Kozenys equation and, hence, any variation of porosity will also lead to a variation of soil permeability. Given that soil permeability has a strong influence on the flow regime, it is evident that the introduction of the simplifying assumption of porosity homogeneity can cause additional errors in the predictions of seepage. Several authors have attempted to tackle this issue by solving equations representing seepage problems to obtain analytical approximations and/or by experimental numerical methods using a MonteCarlo simulation. Gaussian closure analytical solutions of steady state and transient seepage problems in heterogeneous soil masses have been derived by Amir and Neuman (2001, 2004). Their analytical approximation agrees well with their validation by a Monte Carlo analysis (Amir and Neuman, 2001; Amir and Neuman, 2004). The approach however assumes that the entire flow domain is unsaturated which is unlikely to be the case in most real seepage problems. Fenton and Griffiths (1996) conducted Monte Carlo simulations to predict seepage flow through randomly heterogeneous earth dams with various geometries, monitoring in particular the total flow rate across the dam and the amount of drawdown of the free surface on the downstream face of the dam. The authors selected a simple saturated Darcy flow model and iteratively adjusted the nodal elevations along the free surface to match the calculated potential heads. The study only accounts for seepage through the saturated zone and assumes no flow above the free surface (Fenton and Griffiths, 1996). The present study demonstrates a stochastic study of unsaturated seepage through heterogeneous flood defence embankments with randomly varying porosity. The study employs a Monte Carlo approach based on the finite element analysis of seepage through embankments with same geometry and boundary conditions but different random porosity fields (each finite element analysis is referred to here as a realization). A statistical distribution of results is calculated from all realizations and compared with the deterministic solution for the homogeneous case with porosity equal to the mean porosity of the stochastic study. 2 2.1 METHOD Creating the random field
the field is correlated with values within a defined distance from it. The generation of a random field normally requires a mean (), a standard deviation () or coefficient of variation (cov) and a correlation length or scale of fluctuation (). The mean and standard deviation of the random field in this study is equal to the mean and standard deviation of the porosity field which can be estimated from field data, provided that a representative number of experimental observations is available, by using the following equations:
( )
1 n i n i 1 1 n 1
(1)
n
( )
(i ( ))2

i 1
(2)
To create each random porosity field, a rectangular random field is first generated which provides the pool to select the random porosity values from. A random field is a list of random values, which can be mapped on to space. Each random value of
where n denotes the number of observations and denotes the porosity variable. Some authors (Vanmarcke, 1984; Phoon and Kulhawy, 1999) suggest that the correlation length () can be estimated by adding up all individual scales of fluctuation over the field, then take the average and multiply by 80%. Porosity is assumed in the current study to follow a log-normal distribution with mean equal to 0.3 and standard deviation equal to 0.1. Reported values of are very limited in literature and highly depending on the history of the site where measurements are conducted. In the absence of data, a value of 1 m is assumed for the correlation length during the present work considering that materials used in constructing earth embankments are likely to have undergone disturbance and, hence, to have shorter correlation length. The Local Average Subdivision (LAS) technique is used to generate the random porosity field (Fenton and Vanmarcke, 1990). Griffiths and Fenton (2004) argued that the LAS technique is probably the most appropriate technique for generating random fields of geotechnical variables because it is based on an averaging process (Griffiths and Fenton, 2004; Fenton and Griffiths, 2008) that mimics the spatial averaging implied in the measurement of geotechnical properties. The actual value of the variance can be calculated at any stage during random field generation from the global variance, correlation length and correlation function using the local averaging theory. The present study creates a random field of porosity () following a log-normal distribution by first generating a random field of ln, which follows a normal distribution with (ln) and (ln) directly related to () and () (Fenton and Vanmarcke, 1990; Fenton and Griffiths, 2008). Figure 1 presents a sample random field of the natural logarithm of
1314
Table 1. Sensitivity of the time to steady state (Tss) and corresponding total flow rate (Q) to mesh size. Mesh size (Number 554 988 2005 4050 of elements) CPU time 5 7 20 85 (minutes) 156.0 136.7 133.7 133.2 Time to steady state Tss (days) 0.306 0.289 0.293 0.292 Total flow rate Q (m3/day)
Figure 1. A sample random field of ln with () 0.3, () 0.1 and 1 m.
porosity generated using the LAS technique with a Markovian correlation function. 2.2 Finite element model
Figure 2. Frequency histogram of assigned to the mesh of a realization plotted with the probability distribution function of the simulated random porosity field and its targeted distribution.
The Finite Element program CODE_BRIGHT (Olivella et al. 1996) is used to analyze a twodimensional model of a 28 m wide and 7 m high flood defense embankment, with side slopes at 30.3 degrees. The adoption of a two-dimensional model implies that streamlines remain on the plane of analysis, which occurs in reality if the out-of-plane scale of fluctuation is infinite and the embankment ends are impervious (Fenton and Griffiths, 1996). A preliminary mesh sensitivity study is conducted for the homogeneous embankment to decide the appropriate level of mesh refinement to be used in all finite element realizations. Four meshes with numbers of elements equal to 554, 988, 2005 and 4050 respectively are used and the computed values of the time to reach steady state (Tss) and corresponding total flow rate across the embankment (Q) are given Table 1 for all four cases. As it can be seen, the values of Tss and Q differ less than 3% for the three finer mesh sizes. However, the computation time for the mesh with 988 elements is about one-third of that for the mesh size with 2005 elements and less than one-tenth of that of the mesh size with 4050 elements. Since the Monte Carlo simulation requires a large number of realizations, the mesh size with 988 elements is considered optimal to give an acceptable level of accuracy within a manageable computational time. Exactly the same mesh is used for all realizations to avoid introducing any unwanted difference between results caused by a particular meshing choice.
For each realization, a rectangular random field having same width and height as the embankment is generated. The field consists of a regular grid of 3136 cells (i.e. 112 28) with individual random values. This grid is superimposed on the finite element mesh in such a way that their bottom left corners coincide. Each of the 988 elements of the finite element mesh is then assigned the porosity corresponding to the cell having the closest centroid to the element centroid. If the closest porosity value is larger than one, the element takes the next closest value smaller than one. Figure 2 shows the frequency histogram of the mapped porosity field for the finite element mesh of a typical realization. Although less than onethird of the random values are assigned to the mesh elements, the frequency histogram follows relatively closely the distribution of the simulated random field consisting of 3136 values. The simulated field also deviates insignificantly from the targeted Probability Distribution Function (PDF) which is calculated from the mean and standard deviation of the simulated results. 2.3 Boundary conditions and constitutive laws
Pore air pressure is fixed constant and equal to zero across the entire embankment while initial
1315
pore water pressures are set to 0.4 MPa. Liquid pressures at the boundaries are modeled by means of a leakage coefficient (i.e. Cauchy type) according to Olivella et al. (1996). Pore water pressures at the boundaries are then raised at a constant rate over a period of 10 days from 0.4 MPa until reaching a constant water head of 6 m on the left hand slope of the embankment and a constant pore water pressure of 50 MPa on the left hand slope above 6 m height, on the crest and along the right hand slope. The bottom boundary is assumed to be impermeable. The simulation then continues until the steady state is achieved. The van Genuchten model is adopted to describe the change in degree of saturation with suction as well as to predict the variation of relative permeability (kr) with degree of saturation. Unsaturated flow is calculated using generalized Darcys law: u q ks kr w 1 w g (3)
100 0 Total flux x 10 -1 (m3/day) -1 -2 -3 -4 -5 -6 -7 10
200
300
400
Time (days) 500 600 700
800
900 1000 2 1 0.5 0 -1 -2 Total flux x 10 -5(m3/day) 1.5
-0.5 Time 10 to 100 days Time 100 to 1000 days 20 30 40 50 60 70 Time (days) 80 90 100 -1.5
Figure 3. Fluctuation of the net flow rate into/out off the embankment over the simulated period. Negative values indicate inflow.
where q is the water flux, uw is the pore water pressure, w is the water density, g is the gravitational acceleration and ks is the intrinsic permeability calculated through Kozenys equation: ks ko
3 (1 o ) 2 3 (1 ) 2 o
(4)
where ko is the intrinsic permeability at the reference porosity o. The parameters ko and o are set in this work equal to 1012 m2 and 0.3, respectively. According to Equation 4, it is evident that the random spatial variation of soil porosity directly leads to a similar spatial variation of intrinsic permeability. Because of this, the spatial variation of unsaturated intrinsic permeability, which is the product of the intrinsic and relative permeability (Eqn. 3), will also include a random component. In addition, relative permeability depends on the degree of saturation according to the van Genuchten model and, hence, a further influence on the spatial variability of unsaturated permeability comes from the predicted distribution of degree of saturation during both transient and steady state. Experimental evidence has also shown that porosity influence the air entry value of the soil (Romero and Vaunat, 2000), however this influence is the subject of future investigation and not included in this study. 2.4 Defining the steady state During each realization, the time (Tss) from the start of the analysis until the achievement of steady state is recorded. Formally, steady state is defined as the state at which total inflow is equal
to total outflow, thus there are no further changes in the amount of water stored in the embankment. In order to identify the time to reach steady state, the net flow rate into the embankment is calculated by adding the flux along all boundaries. This sum is monitored throughout the analysis until it becomes approximately constant and equal to zero, which is the condition identifying the achievement of steady state. Figure 3 shows that, for a given realization, water accumulates in the embankment over the first 100 days. After that, the net flow rate fluctuates and eventually converges to approximately zero. For this particular example, the fluctuations from around day 190 onwards tend to be abrupt and likely to be caused by numerical artifacts rather than the dynamic of the actual water flow. This behaviour has been observed for a number of realizations. The time at which the net flow rate becomes less than a prescribed tolerance of 0.864 105 m3/day (or 107 kg/s) is recorded as the time when steady state has been attained (Tss). At this point, the total flow rate across the embankment (Q) is calculated by adding the flux along the constant head boundary. Note that, because the boundary of the embankment exposed to atmosphere remains unsaturated throughout the analysis, water can only exit the embankment as unsaturated flow. 3 3.1 RESULTS & DISCUSSION Number of realizations
During any Monte Carlo analysis, a sufficiently large number of realizations should be performed so that the statistics of the sample are not influenced by single results that are either too large or too small, thus being representative of the
1316
population investigated. In this study, a total of 1015 finite element realizations was performed, each corresponding to a different random porosity field generated from the same set of statistics, i.e. () 0.3, () 0.1 and 1 m. The average time to reach steady state (Tss) and the corresponding average total flow rate across the embankment (Q) are plotted in Figure 4 for increasing numbers of realizations, together with their upper and lower bounds of the 95% confidence intervals. The graph shows that the results for both (Q) and (Tss) fluctuate very slightly when the number of realization exceeds 500. Thus the number of realizations performed in this study can be considered larger than the minimum required, which reinforce confidence in the accuracy of the stochastic analysis. 3.2 Comparison with homogeneous embankment
Figure 5. Positive pore water pressure contours and flow vector field for homogeneous embankment with 0.3.
A finite element analysis of a homogeneous embankment, with uniform porosity equal to the mean value of porosity of the heterogeneous cases, is performed to provide a baseline for comparison. Figure 5 show the (positive) hydraulic pressure contours over the saturated zone of the embankment, together with the corresponding flow vector field at steady state for the homogeneous case. Unusually large flow vectors can be noticed in the area close to the top corner of the saturated region on the left hand slope. This is due to a discontinuity in the boundary condition causing a large hydraulic gradient between adjacent nodes. A corresponding prediction of the flow field for one heterogeneous realization is shown in Figure 6, together with the corresponding porosity field in Figure 7. Simultaneous inspections of
Figure 6. Positive pore water pressure contours and flow vector field for a heterogeneous realization.
Figure 7. Porosity distribution showing more porous areas corresponding to preferential flow paths.
1200 0.90 0.80

3 (Q) [m /day]
1000
Number of realisations 800 600 400 95% Confidence Interval of (Q)
200
0 230 210 190 170 150 130 110 90 70 1200
0.60 0.50 0.40 0.30 0.20 0 200 400 600 800 Number of realisations 1000 95% Confidence Interval of (Tss)
Figure 4. Convergence of (Q) and (Tss), together with their upper and lower bounds of the 95% confidence interval, as the number of realizations increases. (Q) is plotted on the left and bottom axes and (Tss) is plotted on the top and right axes.
(Tss ) [days]
0.70
Figure 6 and Figure 7 confirm that, in the heterogenous case, preferential flow occurs along the most porous paths within the soil mass, which results in an irregular flow field. Comparison of Figure 5 and 6 also indicates that the free surface is a smooth curve for the homogeneous case while it is non-smooth for the heterogeneous case. Water seepage has also potential implications for the stability of the embankment. As suction drops following the advancement of the wetting front, the apparent cohesion caused by capillary tension vanishes and shear strength reduces accordingly. The irregularity of the wetting front in the heterogeneous embankment, and the existence of preferential flow paths, causes an irregular distribution of strong and weak soil regions. Because of this, any potential failure is likely to follow a different slip surface from that predicted by standard methods of analysis for homogeneous embankments. From all finite element analyses performed in this work it also emerges that flow through the unsaturated regions can be significant and should therefore be taken into account. Note that, if unsaturated flow is neglected, no infiltration could take place from the constant head boundary into the embankment.
1317
3.3
Probability distribution of results
Visual inspection of the frequency histogram of the results shows that Q and Tss appears to follow a log-normal distribution. The method of moments is then used to estimate the statistical parameters from the results, which are used in computing the theoretical probability distribution function to facilitate comparison. The frequency histograms of Q and Tss are plotted with their theoretically fitted log-normal probability distribution functions in Figure 8 and Figure 9 respectively. For both cases, log-normal probability distribution functions capture reasonably well the frequency distribution of the computed variables. The probability distribution functions of Q appears to be slightly underestimated along the short-tail and overestimated along the long-tail (Figure 8) of the curve while the probability distribution functions of Tss fits remarkably well the frequency histogram of the computed values (Figure 9). The values of Q and Tss for the homogeneous case are also indicated on both figures for comparison. Inspection of Figure 8 and Figure 9 indicates that the total flow rate across the homogeneous embankment is smaller than the average value
of flow rate calculated from all heterogeneous realizations while the opposite is true for the time to reach steady state. Assuming Q and Tss follow the fitted distribution, the probability of the total flow rate across the embankment exceeding that of the homogeneous embankment (0.289 m3/day) is approximately 82%. On the other hand, the probability of the time to reach steady state being less than that of the homogeneous case (137.6 days) is around 61%. A possible explanation is because, in the heterogeneous case, water can bend its routes to avoid low porosity regions and follow higher porosity paths, which have larger permeability than the permeability in the homogeneous embankment. This leads to a higher average flow rate in the heterogeneous embankment compared to the homogeneous one. Since, in the heterogeneous case, water tends to flow through highly permeable routes cutting across the embankment, it also takes less time to wet the soil mass up to its storage capacity, resulting in a shorter time to reach steady state. 4 CONCLUSIONS
Figure 8. Frequency histogram of the total flow rate Q (from the Monte Carlo analysis) together with estimated log-normal probability distribution function.
Figure 9. Frequency histogram of the time to steady state Tss (from the Monte Carlo analysis) together with estimated log-normal probability distribution function.
The study confirms the importance of taking into account both unsaturated flow and random material heterogeneity during the analysis of seepage through flood defence embankments. When high suction values are imposed to the embankment surface, unsaturated flow can become significant and should not be ignored in these cases. Traditional engineering methods, which assume no flow in the unsaturated region, can lead to mis-calculation of the free surface and inaccurate prediction of the flow regime. The effect of random material heterogeneity is investigated by following a Monte Carlo approach based on the stochastic analysis of a large number of finite element simulations of seepage through embankments characterized by different random porosity fields. A log-normal probability density function is assumed for the generation of the random porosity fields across the embankment. The time to reach steady state and the corresponding total discharge are computed for the different random porosity fields and a statistical analysis of the whole set of results indicates that both these quantities also appear to follow a log-normal probability distribution function. The study also demonstrates that the assumption of material homogeneity could lead to under-prediction of the total flow rate and overprediction of the time to reach steady state during wetting from an initial unsaturated condition. Since both these quantities are frequently used in engineering design, this tendency highlights risks
1318
and confirms the importance of including random material variability in modelling soil structures. The qualitative comparison between the analysis of seepage through homogeneous and randomly heterogeneous embankments indicates that, in the latter case, the shape of pore pressure contours and free surface tends to be non-smooth and more irregular compared to the former case. Given the dependency of soil strength on suction, this might also influence potential occurrence of slope instabilities, which are likely to differ from the failure mechanisms predicted by using conventional engineering methods for homogeneous slopes. REFERENCES
Amir, O. & Neuman, S.P. 2001. Gaussian Closure of One-Dimensional Unsaturated Flow in Randomly Heterogeneous Soils. Transport in Porous Media 44: 355383. Amir, O. & Neuman, S.P. 2004. Gaussian Closure of Transient Unsaturated Flow in Random Soils (Part II). Transport in Porous Media 54: 5577. Fenton, G.A. & Griffiths, D.V. 1996. Statistics of free surface flow through a stochastic earth dam. Journal of Geotechnical and Geoenvironmental Engineering 122(6): 427436. Fenton, G.A. & Griffiths, D.V. 2008. Risk assessment in geotechnical engineering. Hoboken, New Jersey: John Wiley & Sons, Inc. Fenton, G.A. & Vanmarcke, E. 1990. Simulation of random field via Local Average Subdivision. ASCE J. Eng. Mech. 116(8): 17331749.
Freeze, R.A. 1971. Influence of the unsaturated flow domain on seepage through earth dams. Water Resources Research 7(4): 929. Gardner, W.R. 1958. Some steady-state solutions of the unsaturated moisture flow equation with application to evaporation from a water table. Soil Science 85(4). Griffiths, D.V. & Fenton, G.A. 2004. Probabilistic Slope Stability Analysis by Finite Elements. Journal of Geotechnical and Geoenvironmental Engineering 130(5): 507518. Lam, L. & Fredlund, D.G. 1984. Saturated-unsaturated transient finite element seepage modal for geotechnical engineering. Advance Water Resources 7: 132136. Neuman, S.P. 1973. Saturated-unsaturated seepage by finite elements. ASCE Hydraulics Division 12: 2233. Olivella, S., Gens, A., Carrera, J. & Alonso, EE. 1996. Numerical formulation for a simulator (CODEBRIGHT) for the coupled analysis of saline media. Engineering Computations; 13(7): 87112. Phoon, K. & Kulhawy, F. 1999. Characterization of geotechnical variability. Can Geotechn J(36): 612624. Richards, L.A. 1931. Capillary conduction of liquid through porous medium. Journal of Physics: 318333. Romero, E. & Vaunat, J. 2000. Retention curve of deformable clays. Experimental Evidence and Theoretical Approaches in Unsaturated Soils, Balkema, Rotterdam. Taylor, R.L. & Brown, C.B. 1967. Darcy flow with a free surface. ASCE Hydraulics Division 93: 25. Vanmarcke, E. 1984. Random Fields: Synthesis and Analysis: MIT Press.
1319
Improvement and controlling deformation of the expansive soil ground

Linchang Miao, Fei Wang & Ying Cui
Transportation College, Southeast University, Nanjing, China
ABSTRACT: Unsaturated and swell-shrinking are two important features of expansive soils. Moisture content increasing can cause the swelling deformation in expansive soils which may consequently induce the heave of the expansive soil ground. Thus, it is important for highway projects to control or decrease the swelling deformation or heave of the road pavement in expansive soil area. An engineering case of improvement and controlling deformation of the expansive soil ground in Huaian where the expansive soil distributes extensively is introduced in the paper. First, the physical characteristics and mechanical behaviors of Huaian expansive soil were investigated. Second, the improvement magnitude was determined based on the swelling pressures of the expansive soil and embankment load. Finally, controlling deformation of the expansive soil ground was analyzed. Research results show that the method of the quicklime-stabilized expansive soil is applicable in expressway embankment projects and the embankment settlement can meet the design requirement. Keywords: expansive soil ground, controlling swelling deformation, quicklime-stabilized expansive soils 1 INTRODUCTION Index (PI) of expansive soils can apparently decrease after the soil is stabilized using quicklime which also indicates that the swell potential of expansive soils has been significantly reduced. The quicklime chemically reacts with the soils, and thus reduces the activity of the soil as well as providing bonding of the soil particles that can reduce the swell potential. Stabilization using quicklime also increased the overall strength of the stabilized soil. The performance of the treated expansive soils using quicklime in the presented highway project demonstrates that quicklime stabilization is a good method for the expansive soil ground and can be used to control the heave deformation of the embankment. 2 PROPERTIES OF HUAIAN EXPANSIVE SOILS
Expansive soils are generally defined as clay layers containing minerals, which are sensitive to the moisture content change in soil. The increasing of moisture content in expansive soils can cause the swelling which represents as vertical movements in expansive soil layers, while the decreasing of moisture content induces the shrinkage in expansive soils. The expansive soil problem has become a world problem. When a road or light structure is constructed over expansive soils, the swelling or shrinkage in expansive base course due to moisture content change may cause extensive deformation and/or cracking in the road pavement or structure. If the moisture equilibrium (i.e., the volume of inflow is equal to the outflow) in expansive soils can be reached, the swelling and shrinkage will never happen. An important parameter in analysis of predicting free-field heave is the migration of the wetting front during the service life of the structure. Conservative analysis assumes that the entire depth of potential heave is wetted. The depth of potential heave is defined as the depth to which the overburden vertical stress equals the swelling pressure of the soil (Nelson, et al., 2001). The volume change constitutive relations are particularly useful for heave predictions in expansive soils (Fredlund et al. 1980; Yoshida et al. 1983). The paper analyzes the swelling ratio and swelling pressure of a typical expansive soil in Huaian, China. The Liquid Limit (LL) and the Plasticity
The site investigations reveal that typical expansive soils extensively distributes in Huaian area, and some highway embankments in this area were damaged by distortion and cracking in pavements because of shrink-swell characteristic of expansive soils. The undisturbed expansive soil samples were borrowed at a project site in Huaian, and the physical properties of Huaian expansive soil are summarized in Table 1. Test data shows that Huaian expansive soil is mainly weak and medium swelling soils and the plasticity index is high. The soil minerals, which are assessed by X-ray diffraction, are presented in Table 2.
1321
Table 1.
Physical properties of Huaian expansive soils. Free swelling (%) 20.0 91.0 Gradation (%) 0.005 mm 18.3 64.3 0.0050.002 mm 10.7 36.4 0.002 mm 15.9 51.9
Gs 2.68 2.71
WL (%) 36.7 71.9
WP (%) 21.3 29.1
Ip (%) 17.8 42.8
Table 2. Smectite 6.012.0
Mineral composition of Huaian expansive soils (%). Illite 15.030.0 Quartz 10.030.0 Feldspar 5.010.0 Other 20.030.0
SR = -0.75 + 0.11PI R = 0.675
SR = 0.892 + 0.132PI R = 0.659
Linear swelling ratio, %
Linear swelling ratio, %

10 20 30 40 50
0 10 20 30 40 50
Plasticity index, %
Plasticity index, %
(a) The degree of compaction is 0.9

Figure 1. Swelling ratio at different degree of compaction.
(b) The degree of compaction is 0.95.
It indicates that Huaian expansive soil is consisted of swelling mineralogy, such as smectite and illite. Figure 1 shows the swelling ratio of the Huaian expansive soil at different degree of compaction, and the swelling ratio increases with the increasing of plasticity index. Figure 2 shows the relationship between swelling pressures and plasticity index, and swelling pressures increases when the plasticity index increases. Figure 3 shows that the plasticity index of expansive soils stabilized with quicklime apparently decreases which also demonstrate that the swell potential is significantly reduced. 3 EXPANSIVE SOILS IMPROVEMENT
Nelson and Miller (1992) pointed out that the two major factors must be identified in the
characterization of a site for a building or a highway where potential shrink-swell problems: (1) the expansive or shrink-swell properties of the soil; (2) environmental conditions that contribute to moisture changes in the soil. The heave or swelling problem of Huaian expansive soils was encountered in Nanjing-Lianyungang expressway project which had to cross through the Huaian area, and the more serious is the long rain season in this area. Therefore, it is important to study and control the problem of heave of expansive soil ground to protect the serviceability of the Nanjing-Liangyungang expressway. Authors suggested adding quicklime to improve Huaian expansive soil ground to decrease shrinkswell characteristics or swelling ratios of Huaian expansive soils. Figure 4 shows swelling pressures at different degree of compaction of Huaian
1322
400
SP = -9.28 + 4.25PI R = 0.596
400
SP = 3.656 + 6.72PI R = 0.586
Swelling pressure, kPa
200

20 30 40 50
300
300
200
100
100
10
0 10
20
30
40
50
Plasticity index, %
Plasticity index, %
(a) The degree of compaction is 0.90. Figure 2.
(b) The degree of compaction is 0.95.
Swelling pressures versus plasticity index at different the degree of compaction.
35
500
The degree of compaction is 0.90. The degree of compaction is 0.95.
30
400
Plasticity index, %
25
300
20
200
15
100
10
0 0 2 4 6 8 10
5 0 2 4 6 8 10
Addtive lime, %
Additive lime, %
Figure 3. Plasticity index of Huaian expansive soils after stabilized.
Figure 4. Swelling pressures versus additive limes of Huaian expansive soils at different degree of compaction.
expansive soils for different proportions of additive limes. The test results show that swelling pressures increase with increasing of the degree of compaction and decrease with increasing of the additive lime. Figure 4(a) shows that the swelling pressure could decrease to 50 kPa and even smaller when the additive lime is 6% and the degree of compaction is 0.9. Those results demonstrate that the quicklime can be used to decrease or limit the swelling pressures and control the heave or swelling of the road pavement in expansive soil area.
ENGINEERING CASE
In general, the shrink-swelling characteristics of the expansive soil ground are determined by the position of water table and rainfall variations. The road pavement over the expansive soil ground cracking and/or heaving always happen when the moisture content change in expansive soils. The problem of expansive soils has to be solved properly to avoid the damage of the structures. According to research results of the quicklime stabilized expansive soils, the heave problem of the
1323
400 350 300
250 200 150 100 50 0
expressway embankment in Huaian the stabilized expansive soil area during embankment construction. During rainfalls, there is tiny heave phenomenon in early filling stage, which the fill loading is smaller than the swelling pressure of expansive soil ground because the moisture content change of expansive soil ground increases, but the final settlement of the embankment is acceptable when the height of embankment is greater than 2.5 m. 5
0 50 100 150 200 250 300 350
Filling height, cm
CONCLUSIONS
Time, d
(a) Filling height curve.

0 0 50 100 150 200 250 300 350
Time, d
-50
Rainfall
Settlement, mm
The paper discusses the swelling behavior of Huaian expansive soils. The research results show that swelling pressures increase as plasticity index increases and quicklime can be used to decrease or limit the swelling potential of expansive soils. Additive lime amount can be determined based on the overlying loading of expansive soils ground. A case history verifies the effectiveness of the proposed method in expressway embankment projects. ACKNOWLEDGEMENT
-100
Rainfall
-150
Rainfall
-200
Rainfall
(b) Settlement curve of the quicklime stability foundation.
Figure 5. Settlement-time curve of the road embankment in Huaian expansive soil area during construction.
The Projects No. 50878051 and 40374047 are supported by National Natural Science Fund of China (NSFC). The work presented in this paper is being carried out as a part of the projects No. 50878051 and 40374047. The authors wish to express their gratitude for the support given to this work by National Natural Science Fund of China (NSFC). REFERENCES
Fredlund, D.G., Hasan, J.U. & Filson, H. 1980. The prediction of total neave. Proceedings, 4th International Conference on Expansive Soils, Denver, CO. Vol. 1. 117. Fredlund, D.G. & Xing, A. (1994). Equation for the Soil-Water Characteristic Curve. Canadian Geotechnical Journal, 31 (3): 521532. Nelson, J.D. & Miller, D.J. (1992). Expansive Soils. Problems and Practice in Foundation and Pavement Engineering. John Wiley and Sons, New York. Nelson, J.D., Overton, D.D. & Durkee, D.B. (2001). Depth of Wetting and the Active Zone. Expansive Clay Soils and Vegetative Influence on Shallow Foundations, ASCE, Houston, Texas, 95109. Yoshida, R., Fredlund, D.G. & Hamilton, J.J. 1983. The prediction of total heave of a slap-on-ground floor on regional clay. Canadian Geotechnical Journal, 20(1): 6981.
expansive soil ground can be controlled or limited if the height of embankment is over 2.5 m because the embankment load is greater than the swelling pressure of the expansive soil ground. Based on the test results, quicklime was used to stabilize the Huaian expansive soil in Nanjing-Lianyungang expressway project, the treating method is forming a lime stability layer with thickness of 60 cm at the degree of compaction of 0.90, and additive quicklime is 6% for the stabilized expansive soil when the height of pavements is greater than 2.5 m. The additive quicklime is 8% in local section when the height of pavements is less than or equal to 2.5 m. The improvement design method of expansive soils realizes the performance design. Figure 5 shows the settlement-time curve of the
1324
Failure of peat dykes by drought

D. Nterekas, J.P. Oostveen, F.A.J.M. Mathijssen, R.B.J. Brinkgreve, M.A. Hicks & F. Molenkamp
Delft University of Technology, CITG, Delft, The Netherlands
ABSTRACT: In The Netherlands during the dry warm summer of 2003 some peat dykes failed. To gain some insight into the severity of various phenomena potentially contributing to such instabilities, a computational parametric study has been performed. The geometry and hydraulic conditions are combined with the approximate data on soil stratification and mechanical and hydraulic soil properties. In the parametric study the approximate soil data are varied, taking into account both the estimated mean and standard deviation of all mechanical and hydraulic soil properties. 1 1.1 INTRODUCTION History and current status of peat dykes 1.3 Research approach The study is aimed at clarifying which of these factors can be dominant for the stability of a peat dyke under drought. The main available computational tool for analysis is the combined PlaxFlow & Plaxis model; Plaxflow for the unsaturated pore water flow and evapotranspiration and Plaxis for deformation and failure. This model enables the finite element analysis of two-dimensional transient coupled saturated and unsaturated groundwater flow and soil deformation for geotechnical problems. Weather conditions can be simulated by prescribed infiltration and exfiltration of water along the contact surface between the geo-structure and the atmosphere. The input data for the research is based on the soil investigation performed by GeoDelft after the failure of the peat dyke at Wilnis. This failure involved a section of the dyke with a length of about 50 m, being pushed by the water in the upper canal towards the polder over a length of about 10 m. This motion was stopped by the driving water level in the canal decreasing due to the water in the canal segment flushing into the polder through the gaps induced by the failure on both ends of the breach. It is understood that the applied simulation model only involves static equilibrium with relatively small displacement, thus not allowing the simulation of the development of the post-stability large deformation as observed. Consequently the calculated instability mode cannot resemble the observed large motion, but is expected to represent the initiation of the drought-induced pre-instability behaviour.
The essential flood defences in The Netherlands are divided into primary and secondary types, the latter separating the polders from the main canal system. Of the 14,000 km of secondary dykes about 3,500 km consist of peat dykes. In the middle ages many polders were created by the excavation of peat for fuel and the subsequent reclamation of these areas. Before the dry and warm summer of 2003 peat dykes were generally considered to be safe. Following the 2003 failures the quality of the understanding of the behaviour of these dykes is being questioned. 1.2 Recognised potential causes of instability
As potentially important influences on the behaviour of peat dykes under drought the following aspects are recognised: the anisotropic material properties of peat (e.g. Hobbs, 1986, Cola & Cortelazzo, 2005, Den Haan & Kruse, 2007), the seepage from the upper canal through the dyke towards the ditch at the toe, the lowering of the groundwater table in the dyke by evapotranspiration (e.g. Evans et al. 1999, Holden et al. 2001, Lafleur et al. 2005), the buoyancy of the dyke (e.g. Warburton et al. 2004, Long & Jennings, 2006) by the water pressure in the underlying permeable sand layer, the deformation of the dyke due to the redistribution of shear stress along the sole of the dyke towards its heel as a consequence of increasing buoyancy and the development of cracks and shear bands in both the bottom of the canal and the heel of the dyke.
1325
Table 1. Mean input parameters of soil model properties. Name Permeability Cohesion cr Friction phi Coeff. K0 Stiffness E50 Stiffness Eod Stiffness Eur Power m m/d kPa kPa kPa kPa Holland peat 0.04 4.58 26.4 0.30 2200 425 5.E3 1 Clay 5.E-4 8.15 18.2 0.44 2.E3 600 6.E3 1 Base peat 4.E-3 9.7 20.0 0.30 250 340 1.8E3 1 Sand 5.0 1.8 31.0 0.23 1.2E4 8.E3 5.E4 0.5
Figure 1. section.
Estimated pre-failure geotechnical cross
1.4
Overview of soil data
GeoDelft (20032004) performed a number of field and laboratory tests in order to determine the geotechnical characteristics of the area. These tests included 67 CPTs, 48 borings, 15 K0-CRS tests, 12 triaxial tests and 13 simple shear tests. Based on the CPTs and borings the soil stratification is determined. Four different soil types are distinguished, namely Holland peat, clay, base peat and sand. The geotechnical cross section as applied for the analysis can be seen in Figure 1. Also visible is the hydraulic head of the deep sand layer which has been measured by observation wells and found to be on average at 6.00 m NAP with 0.04 m standard deviation. 2 2.1 PLAXIS PARAMETERS Hardening Soil model
Only the mean values are indicated. The estimated standard deviations vary per parameter based on the experimental data. 2.3 Determination of unsaturated weight unsat of Holland peat
Due to the evapotranspiration of groundwater along the ground surface, including the slope of the dyke, the groundwater table will drop and the upper Holland peat layer will become more unsaturated. The corresponding loss of weight may deteriorate the stability of the dyke. Consequently the unsaturated volumetric weight needs to be taken into account appropriately. The applied unsaturated self weight is calculated based on the calculated distribution of the degree of saturation above the groundwater table at the end of a period with severe drought. The definitions of saturated gravimetric water content wsat and unsaturated volumetric weight unsat are: wsat n w (1 n ) m (1) (2)
The material model chosen for the analysis is the Hardening Soil model (Schanz et al. 1999), which is considered to allow only a rough approximation of the strength and deformation properties of peat. The main strength parameters, i.e. cohesion and angle of internal friction, can be taken directly from the triaxial tests or the SS tests performed The stiffness is defined with three different moduli of elasticity, the triaxial loading stiffness, E50, the unloading-reloading stiffness, Eur and the oedometer stiffness, Eod. Stress dependency is taken into account by the power m. For the determination of the stiffness parameters of the Hardening Soil model for the 2 different peat layers, simulations of the K0-CRS tests have been performed. 2.2 Input parameters
unsat [(1n)m Sw n w] g
where n is the porosity, w is the density of water, m is the density of the soil material, Sw is the degree of saturation and g is the acceleration due to gravity. From equations (1) and (2) it follows for saturated soils: n wsat sat w g (1 wsat ) (3) (4)
m
sat w g (1 wsat ) sat wsat

w
The applied parameter sets for the Hardening Soil model and the 4 soil types occurring in the peat dyke (see Figure 1) have been collected in Table 1.
where concerning the saturated weight the average sat 9.8 kN/m3 and the standard deviation sat 0.35 kN/m3 have been measured, while concerning the gravitational water content the average
1326
wsat 6.58 and the standard deviation wsat 1.66 have been found. Substituting these values in (3) and (4) leads to the corresponding porosity n and material density m. The unsaturated weight unsat can be calculated from (2). 3 3.1 PLAXFLOW & PLAXIS ANALYSES Initial state
Next the three consecutive calculation phases are described in general terms, each phase involving both groundwater flow and the corresponding deformation. In the described calculations the average parameters as collected in table 1 have been applied. 3.2.1 Phase 1Removing the soil In this phase sections of soil are removed in order to lead to the actual geometry of the situation with the canal and the dyke. Also the sheet piling and anchors along the canal are activated. The groundwater levels remain the same as in the initial phase thus leading to almost the entire model being submerged under water, as indicated in Figure 3. A steady state groundwater flow calculation is performed. 3.2.2 Phase 2Lowering the groundwater table The groundwater level in the upper cohesive layers (Holland peat, clay and base peat) is initially set to just below the ground surface while along the right boundary of the sand layer a hydraulic head of 6.00 m NAP is prescribed, as depicted in figure 3. A steady state flow analysis is then performed in order to find the actual groundwater table. The calculated flow field includes the lowering of the phreatic level in the slope of the dyke, the mainly horizontal groundwater flow from the canal towards the ditch through the Holland peat layer and the mainly downward seepage from the canal towards the deep sand layer. The corresponding distribution of the hydraulic head, illustrated in Figure 4, shows a homogeneous hydraulic head
The two dimensional geometry of the model can be seen in figure 1. The model in its initial form includes a flat surface at 1.50 m NAP, which is the eventual dyke crest level. This enables the use of the K0 procedure in order to calculate the initial effective soil stresses, allowing for the soil effective stress history. The model extends 20 m beyond the important elements of the structure on both sides. The dimensions of the model are 100 m wide and 23.5 m deep. The bottom boundary of the model is fixed. The left and right boundaries are free to move vertically and have horizontal fixity. For the pore pressure a simple phreatic level calculation is conducted, based on a groundwater level of 2.15 m NAP, the eventual water level of the canal. All boundaries are set as impermeable except the right boundary of the lower sand layer, where a constant hydraulic head is described. The geometry of the model in Plaxis can be seen in Figure 2. The discretization of the mesh was set to very fine and the 3 top layers (Holland peat, clay and Base peat) were refined further around locations with concentrated interactions. 3.2 Calculation procedure for all phases
For the applied approach 3 calculation phases are defined. The first 2 phases are performed in order to simulate the soil effective stress history, following the sequence by which the dyke was created through excavation on the right (north) side in figure 2. Consequently the end of phase 2 represents the situation at the beginning of the dry summer of 2003. Phase 3 simulates the effect of drought that eventually led to the failure of the dyke.
Figure 3. Geometry and boundary conditions for phase 2 with the input water table.
Figure 2. Geometry and boundary conditions for initial state model.
Figure 4. Groundwater head of phase 2.
1327
distribution below the clay layer and a decreasing hydraulic head away from the bottom of the canal above the clay layer. The calculated degree of saturation above the phreatic level in the upper Holland peat layer on both sides of the canal remains larger than 75%, including in the upper part of the slope. The calculated displacements, depicted in Figure 5, are maximum behind both sheet pile walls, indicating their lengths of 3 m to be relatively short. The corresponding principal strains tensors are spreading out downwards from behind both the sheet pile walls, through the upper Holland peat layer, also reaching the base peat layer even in the region below the canal, as indicated in Figure 6. The effective stress state in the soil before the start of the dry summer of 2003 is depicted in Figure 7 in the form of principal stress tensors. The rotated orientation of these tensors below the lower part of the slope is consistent with the locally expected maximum mobilized friction for horizontal simple shear deformation. Having arrived at the effective soil state just before the start of the dry summer of 2003, next the corresponding safety factor of the dyke is calculated by
applying the phi-c reduction method. However, that stability analysis is found not to calculate an overall failure mechanism of the dyke, but is governed by the low safety of the sheet pile walls. 3.2.3 Phase 3Drought In phase 3 the effect of drought is simulated through a transient unsaturated consolidation analysis. The time interval is set to 60 days. Drought is simulated by outflow of water from the entire ground surface. The amount of outflow is set to 5 mm/day, which is a realistic value for a dry summer in The Netherlands. The resulting groundwater table together with the corresponding distribution of the hydraulic head are shown in Figure 8. The phreatic surface in both the slope of the peat dyke and in the polder above the clay layer is lowered as expected. The region with significantly larger hydraulic head than in the deeper sand layer remains limited to the vicinity of the bottom of the canal above the clay layer. From the results of the flow field of the groundwater it is noted that the ditch supplies infiltrating groundwater for both the toe of the slope and the polder. In addition the downward seepage from the bottom of the canal towards the deeper sand layer also remains significant. In the unsaturated region above the water table the degree of saturation along the ground surface, including the surface of the slope, reduces to about 30%. The distribution of magnitude of the displacement increments due to 60 days of drought are depicted in Figure 9. The region with significant displacements occurs behind the sheet pile wall and the neighbouring upper part of the slope of the peat dyke, while stretching downward and reaching the base peat layer. The calculated lateral displacement increment of the sheet pile wall against the peat dyke is of the order of 15 cm. The corresponding incremental strain tensors, shown in Figure 10, show a region with relatively large strains starting from the tip of the sheet pile wall and propagating about 45 downwards
Figure 5
Displacements at the end of phase 2.
Figure 6.
Principal strain tensors at the end of phase 2.
Figure 7. phase 2.
Effective principal stress tensors at end of
Figure 8.
Groundwater head at end of phase 3.
1328
Figure 9.
Displacement increments during phase 3.
Figure 10. phase 3.
Principal strain increment tensors during
3.2.4 Safety factor at end of drought The above results indicate that the applied finite element model of the peat dyke does not simulate failure by drought. To clarify the proximity of the effective soil state at the end of drought from actual failure, subsequently the phi-c reduction method (Brinkgreve & Bakker, 1991) is applied, similar to the failure analyses at the end of phase 2. For the effective stress state at the end of the period of drought a safety factor of 1.22 is obtained. The calculated strain increments in terms of principal strain tensors are illustrated in Figure 12, showing a failure mechanism, starting from the toe of the right sheet pile wall, going downwards at an angle of approximately 20 to the horizontal, continuing straight just above the clay layer and at the end going upwards towards the ground surface at about the middle of the slope of the dyke. There is also some vertical strain visible in the top of the sand layer in the right hand side of the model. 3.3 Parametric study
Figure 11. Distribution of deviatoric invariants of the strain increments during phase 3.
through the Holland peat and clay layers and becoming practically horizontal in the base peat layer, where it decreases in magnitude half way towards the toe of the peat dyke. Figure 11 depicts the distribution of the deviatoric invariants of the strain increments due to drought. A diffuse distribution of locally concentrated narrow regions with maximum deviatoric strain of the order of 0.01% is calculated. Amongst this diffuse marginal pattern one deformation mechanism starts at the tip of the sheet pile wall, dips about 45 degrees downwards towards the polder and stretches mainly horizontally in the base peat layer before disappearing and reappearing further within this layer. In addition, in the Holland peat layer composing the slope, local spots and a few vertical narrow regions with larger deviatoric strains can also be observed, but no clear failure mechanism.
The above-described calculation procedure is repeated in a parametric study in order to check the consistency of the calculated responses and to get insight into the sensitivities for parameter variations, while taking account of the measured variations of the experimental data. In this parametric study the permeability values of both peat layers and the clay layer are increased and decreased by a factor 10 around the mean value. Furthermore the strength is increased and decreased by one quarter of the measured standard deviation. The effects of the variation of the permeability on the location of the groundwater table in the slope of the peat dyke, the distribution of the degree of saturation above the water table and the volumetric weight of the unsaturated peat are all consistent and significant, but do not affect the overall pattern of deformation during drought. The effects of the variation of the strength parameters still need further attention as no clear consistent effect on the stability is obtained, possibly due to the simultaneous effect on the critical stability of the relatively short sheet pile wall.
Figure 12. Principal strain increment tensors due to phi-c reduction-induced failure following phase 3.
1329
CONCLUSIONS
REFERENCES
Brinkgeve, R.B.J. & Bakker, H.L. 1991. Non-linear finite element analysis of safety factors. In G. Beer, J.R. Booker & J.P. Carter (eds), Computer Methods and Advances in Geomechanics: 11171122. Rotterdam: Balkema. Cola, S. & Cortelazzo, G. 2005. The shear strength behaviour of two peaty soils. Geotechnical and Geological Engineering 23(6): 679695. Den Haan, E.J. & Kruse, G.A.M. 2007. Characterisation and engineering properties of Dutch peats. Characterization & Engineering Properties of Natural Soils 4: 21012133. Evans, M.G., Burt, M.G., Holden, J. & Adamson, J.K. 1999. Run-off generation and water table fluctuations in blanket peat: evidence from UK data spanning the dry summer of 1995. Journal of Hydrology 221(34): 141160. GeoDelft. 20032004. Investigation of Wilnis peat dyke failure (in Dutch), reports 17, CO-411242. Hobbs, N.B. 1986. Mire morphology and the properties and behaviour of some British and foreign peats, Quarterly Journal of Engineering Geology and Hydrogeology 19: 780. Holden, J., Burt, T.P. & Cox, N.J. 2001. Macroporosity and infiltration in blanket peat: the implications of tension disc infiltrometer measurements. Hydrological Processes 15(2): 289303. Lafleur, P.M., Hember, R.A., Admiral, S.W. & Roulet, NT. 2005., Annual and seasonal variability in evapotranspiration and water table at a shrub-covered bog in southern Ontario, Canada, Hydrological Processes 19(18): 35333550. Long, M. & Jennings, P. 2006. Analysis of the peat slide at Pollatomish, County Mayo, Ireland. Landslides 3(1): 5161. Schanz, T., Vermeer, P.A. & Bonnier, P.G. 1999. The hardening soil model: Formulation and verification. In: R.B.J. Brinkgreve (ed.), Beyond 2000 in Computational Geotechnics10 Years of Plaxis: 281296. Rotterdam: Balkema. Warburton, J., Holden, J. & Mills, A.J. 2004. Hydrological controls of surficial mass movements in peat. Earth Science Reviews 67(12): 139156.
The applied computational Plaxflow & Plaxis model for unsaturated consolidation analysis satisfies static equilibrium and includes the updatedLangrange formulation. However the observed failure mechanism of the peat dyke involves post-stability very large, possibly partly dynamic, displacements. Consequently the computational model is understood only to enable the simulation of the initiation phase of the failure rather than the forensic large-deformation evidence collected at the breached area. In the simulations the maximum shear strains occur at the toe of the right sheet pile wall. The propagation of the failure plane away from the tip of the sheet pile wall remains unclear. During drought a deformation mechanism reaching the deeper base peat layer seems to be developing, but does not completely substantiate during the period of drought. On the other hand, the stability analysis at the end of the drought period, involving phi-c reduction, indicates a shallow mechanism, already surfacing near the middle of the slope. Therefore neither of the calculated mechanisms ends in the vicinity of the ditch near the toe of the dyke as observed. The simulation of drought with outflow of pore water, representing the evapotranspiration from the ground surface, produces realistic results in terms of the distributions of pore water suction, degree of saturation and volumetric weight. It is recommended to also consider the sensitivity for variation of the hydraulic head in the deep sand layer. Retrospectively the mean length of the sheet pile wall should have been longer to improve its stability. Further research is needed to acquire more insight into the ways peat dykes work and efficient mitigation can be achieved. It is recommended to develop and apply a fullycoupled consolidation analysis for unsaturated soil behaviour and post-stability dynamic and large deformation behaviour.
1330
Numerical analysis of the behaviour of embankments constructed with expansive soils

J.I. Ortega
Geopayma, Barcelona, Spain
A. Lloret, S. Olivella & E. Romero
ABSTRACT: The recent advances made in modern Unsaturated Soil Mechanics have permitted the development of new analysis methods for the prediction of the behaviour of embankments and other earth structures subject to changes in atmospheric conditions (precipitation, temperature, relative humidity) and under the action of traffic loading. One of the breakthroughs of greatest interest is the possibility to numerically model the processes of volume change in expansive soils when subjected to variations in humidity. In this paper, the analysis employed, based upon the Barcelona Basic Model constitutive model (Alonso et al., 1990), to study the viability of the embankments of the speed railway Montorns del Valls-La Roca del Valls (LAV Madrid-Zaragoza-Barcelona-French Frontier), is presented. A coupled flux-deformation analysis has been undertaken to model the water-infiltration process through the embankment, and the associated volume change deformations have been evaluated in terms of rainstorm duration. 1 INTRODUCTION problem of expansive soils in a greater detail than the usual experience-based approach. We may say that, at least in Spain, theres no significant published experience concerning numerical simulation of humidity variation processes in embankments, even thought it seems clear that theres a great potential for the application of this kind of techniques (i.e. Alonso et al., 2000). In this paper, the conclusions obtained after a practical case study, in which the analyzed materials, excavated in adjacent cuttings corresponding a high speed railway work, exhibit an expansive potential above that prescribed for the construction of embankments, are presented. The detailed stress-strain study carried out, including numerical modelling of the construction of the embankment and rainfall events has permitted a greater knowledge of the influence of the different variables taking part in the problem. 2 MAIN ASPECTS OF THE MONTORNS DEL VALLS HIGH SPEED RAILWAY EMBANKMENTS DESIGN
Reutilization of excavated materials in road and railway works for the construction of embankments is a subject of great interest for the engineer in order to avoid excessive environmental impact and also to achieve economical efficiency. In Oteo (2004) reference is made to the need of reviewing traditional earthworks acceptance criteria, and also to study in situ treatment strategies in order to achieve maximum profit of excavated materials. In this scenario, one of the classical problems that an engineer must confront is the reutilization of materials that exhibit a certain grade of expansive potential. In Spain, normative texts such as the well-known Pliego de Prescripciones Tcnicas Generales para Obras de Carreteras y Puentes (PG-3) prescribe that only materials with a volume change of less than 3% after free saturation in oedometer, when specimens are compacted with Normal Proctor energy at the optimum humidity, can be used for the construction of embankments. Recent Technical Prescriptions drafted by official agencies, in order to specify criteria for the design of high speed railway projects, are even more restrictive, as they only admit materials under 1% free-swelling in California Bearing Ratio (CBR) test to be used in the embankments. Recent developments in Unsaturated Soil Mechanics give the engineer new calculation tools (Olivella et al., 1996) that enable the analysis of the
As previously stated, the case study presented in this paper refers to the embankments of the Montorns del VallsLa Roca del Valls high speed railway project, which is included as part of the Spanish high speed railway network. The site is situated about 20 km from the city of Barcelona.
1331
2.1
Geological context
The area concerning this study is part of the Catalnides geological domain, which is divided into three units, aligned in a SW-NE direction: the Cordillera Costera Catalana, the Valls-Peneds Depression and the Prelitoral Catalonian formation. The main part of the railway affects the sedimentary materials that fill the Valls-Peneds Depression and the fluvial sediments related to the superficial drainage. 2.2 Geotechnical characterization in project phase
Figure 3.
Design of the embankments (cross-section).
The Tertiary materials (referred as T unit) conform a very heterogeneous unit, characterized by alternations of clayey sand and sandy clay surrounding lenticular-shaped conglomerates, and zones where lutitic materials are predominant. Figure 1 shows the plasticity characteristics of these materials. The problem in reutilizing these materials is due to the fact that they demonstrate swelling indexes above 1% in CBR tests (1% represents the maximum allowed by the Spanish Speed Railway Technical Prescriptions). Figure 2 presents some
45 40 35 30 25 20 15
results of these swelling tests, which reveal a high dependence on the specimens fines content and on the dry specific weight achieved in compaction. As can be seen, specimens with a fines content above 20%, compacted achieving 95% of the maximum dried unit weight of Modified Proctor tests, demonstrate free swell values in excess of the specified limits. As a result, embankments were designed with a inner core composed of in-situ excavated materials, encapsulated, to prevent expansion, by non-expansive materials, brought to the site from borrow areas situated far from the site. Typical cross-section is presented in Figure 3. 3 EMBANKMENT MODELLING
Plasticity index (%)
CH CL
MH-OH
10 5 CL-ML 0 20 25 30 35 40 Liquid limit (%) 45 50 55 60
ML-OL
Due to the important economic repercussion of the previously described design, numerical modelling of the embankment construction and subsequent response to rain events was undertaken. The main purpose of this work was to identify the influence of each of the principal variables involved in the problem, and to get the magnitude of the volumetric strains, in terms of movement development in top of the embankment. 3.1 Conceptual framework
Figure 1. materials.
Plasticity characteristics of the Tertiary
The constitutive model used to characterize the materials that constitute the embankment core was the well-known Barcelona Basic Model (BBM, Alonso et al., 1990), slightly modified incorporating Equation 1 in order to model the dependence of the swelling potential on the confinement stress in the elastic path. This kind of non-linear elasticity is described in further detail in Olivella et al. (1996). More complex interaction models, including double-structure formulations, can be found in Sanchez et al. (2005), but Equation 1 was considered accurate enough for our analysis. s patm e a1 ln( p ) a2 ln v patm s patm ) a3 ln( p ) ln( patm
Figure 2. Free-swelling obtained in CBR tests, showing the dependence of this value to fines content and degree of compaction (M.P., Modified Proctor, N.P., Normal Proctor).
(1)
1332
where a1, a2, a3 are parameters to be determined and patm is the atmospheric pressure. It was also necessary to include in the analysis the hydraulic constitutive equations and specific parameters that are needed to solve the unsaturated flux problem that describes the infiltration of water into the embankment body during rain episodes. These are intrinsic permeability, retention curve and relative permeability. Intrinsic permeability has been determined directly from permeability tests. To parameterize the retention curve, a van Genuchten (1980) model was adopted, using Equations 2 and 3 to relate effective saturation degree Se with suction:
1 Sr Smin Pg Pl 1 Se 1 Smax Smin P
(2)
P P0
0
(3)
Saturated permeability determination in triaxial test equipment. Granulometric curve determination. Modified Proctor compaction test. Traditional oedometer tests of specimens compacted to 95% of the maximum Modified Proctor test energy, using humidity values of optimum, optimum minus two per cent and optimum plus two per cent. Determination of a swelling volumetric deformation versus confining stress curve, by saturating specimens under different confinement stresses in the oedometer equipment. Determination of a swelling volumetric deformation versus confining stress curve, by saturating specimens under different confinement stresses in the oedometer equipment. Shear box tests, Consolidated and Drained (CD). Specimens for these tests were compacted at different energies and humidity conditions, in a similar manner as described before for traditional oedometer tests. Some of the most significant results obtained are detailed in the following lines. The retention curve is shown in Figure 4. A reasonable fit of the wetting path has been achieved by parameters 0.35, P0 0.085 MPa and Sr,mx 1. Figure 5 shows the swelling (expressed in %) induced by wetting in the oedometer equipment specimens compacted at the embankment representative conditions (95% Modified Proctor energy; optimum humidity). Figure 6 shows the volume changes induced by a wetting (two steps)-drying path in the suctioncontrolled oedometer. In this test, unsaturated permeability in each of the steps can be obtained by fitting the evolution of the income/outcome volume of water, using an analytical solution. Figure 7 shows the fitting curve used to obtain unsaturated permeability in one of the steps of the suction-controlled oedometer. Figure 8 shows
where P0 must be determined fitting experimental results; 0 is water superficial stress at the same temperature P0 is measured (usually 0 0.072 N/m at 20C; in the model a constant temperature has been considered, so P P0); is the shape parameter in the retention curve; Smax, Smin are, respectively, the maximum and minimum degree of saturation (Sr) considered in the model. Finally, Pg and Pl represent air pressure and water pressure, respectively. To describe relative permeability, defined as: kr kunsaturated ksaturated (4)
a potential law depending on the degree of saturation has been adopted (Equation 5), where n is a parameter to be determined by fitting experimental data: kr Se n 3.2 Hydromechanical characterization (5)
100
10
Suction (MPa)
A specific laboratory tests campaign has been under-taken, consisting on the following: Retention curve determination. Two different equipments have been used. In the range of low suctions, a Ridley-type tensiometer, developed by CERMES, has been used. Meanwhile, in the high suctions range, tests were undertaken using a Decagon Devices Inc. WP4-model psicrometer. Humidity-controlled oedometer. Suction-controlled oedometer.
0,1
0,01
Drying path Measured values Wetting path
0,001
0 0,1 0,2 0,3 0,4 0,5 0,6 0,7 0,8 0,9 1
Figure 4.
Retention curve (wetting and drying paths).
1333
Confining stress, kPa 0,0 Swelling volumetric deformation, % 5 4 3 2 1 0 -1 50,0 100,0 150,0 200,0 250,0 300,0
Figure 5. Swelling deformation measured after saturation under different confining stresses in the oedometer equipment. It is worth noting that wetting the specimen under 300 kPa confining stress results in a contractive (collapse) deformation.
Figure 8. Relative permeability in terms of degree of saturation, as obtained in suction-controlled oedometer test. Table 1. Parameters used for the hydromechanical model. Definition of parameter Mechanical behaviour in the non-lineal elastic domain Virgin plastic compressibility parameters Symbol a1 a2 a3 r (0) p0* M k p0 0 Srmin Srmax n0 Value 0.012 0.035 0.022 0.7 18 0.0454 0.15 MPa 1.027 8.54 0.085 MPa 0.072 MPa 0.35 0 1 0.315 0.3
Figure 6. Void index as a function of suction. Results obtained in the suction-controlled oedometer test, following a two-steps wetting and a final drying path.
Initial mean yield net stress Slope of critical state line Parameter that describes increase in cohesion with suction Retention curve parameters
Initial porosity Non-associativeness parameter
Figure 7. Temporal evolution of infiltrated water into the specimen on the second wetting step of the suctioncontrolled oedometer test, showing the fitting curve used to estimate unsaturated permeability.
the variation of relative permeability in terms of changes in the degree of saturation. The complete list of parameters used for numerical modelling is presented in Table 1. Figure 9 shows the calculated and measured values
Figure 9. Measured and calculated values for deformations induced by wetting under different confining stress.
1334
of swelling induced by wetting under confining stress used in the determination of parameters a1, a2 and a3 for the non-lineal elastic path. 4 ANALYSIS OF THE EMBANKMENT BEHAVIOUR UNDER WETTING EPISODES
The numerical simulation here presented has been used to reproduce the behaviour of embankments of three different heights, undertaking a sensibility analysis in order to reproduce the effect on the results when changing the saturated (and, indirectly, unsaturated) permeability. The analysis was performed with the finite element code CODE_BRIGHT, developed at the Department of Geotechnical Engineering of UPC. The code solves, simultaneously, in a monolithic manner, the balance equations for heat transfer, water flow and air flow, as well as the mechanical equilibrium equations. It is formulated in terms of two independent stress variables: the net stress (excess of total stress over air pressure), and suction or difference between air and water pressures. The formulation of the code and its numerical implementations are described in Olivella et al. (1996). Calculations have been undertaken initially for a saturated permeability value of 2.5 1010 m/s (corresponding the mean result obtained in triaxial tests), but permeabilities 10 and 50 times greater were later incorporated to undertake the sensibility analysis previously referred. In the aim of simplicity, a permanent superficial 5 cm water sheet has been used to model external boundary conditions (on slopes and top) during rainfall episodes. Horizontal and vertical displacements were fixed at the lower plane of the model, which was considered impervious. The first two meters from the top where considered to be constructed with coarse soil, meaning a non-expansive high-permeability material. As commented previously, volumetric deformation depends not only on the change in the degree of saturation, but also on the confining stress applied to the considered volume of soil. As seen, swelling de-creases when confining stress increases, and deformations may even change their sign to collapse for high stress values. The resulting movement at the top of the embankment is hence the result of integrating deformations along the whole embankment body. Upward movement, then, is greater when low stresses are predominant (embankments of moderate height) and when changes in degree of saturation are significant (high permeability). Figure 10 shows the change in shape of the embankment for some of the studied cases. As detailed later, when the embankment is above 10 m
Figure 10. Calculated changes in the shape of the embankment after the considered rainfall period, for different values of embankment height and saturated permeability: (a) k 2,5 1010 m/s, H 18 m; (b) k 1.25 108 m/s, H 18 m; (c) k 1.25 108 m/s, H 10 m. It is worth noting that, in case (a), water infiltration is so small that the deformation is purely swelling. In case (b), as permeability increases, significant variations of suction may occur inside the embankment, resulting in collapse deformations as a consequence of the higher confining stress. In case (c), although collapse deformations are detected at the base of the embankment, swelling deformations are predominant, resulting in an upward movement on top of the embankment. Table 2. Vertical movements (mm) calculated at top of the embankment (at the axis), depending on embankment height (H) and saturated permeability, k (positive values indicate swelling). H 18 m k 2.5 1010 m/s k 2.5 109 m/s k 1.25 108 m/s 2.8 4.2 13.0 H 10 m 3.0 5.2 17.5 H6m 3.0 11.0 11.0
height and permeability is high enough, collapse deformations may develop in the base of the embankment, and may even result in a downward movement at the top. For smaller embankments, collapse deformations, if any, are not significant, and swelling deformations are predominant. Table 2 shows the movements obtained in the axis at the top, depending on the height and permeability of the embankment. 5 DISCUSSION AND CONCLUSSIONS
As a first consideration, we would like to remark that this work demonstrates how modern
1335
unsaturated soil analysis tools can be easily applied to common engineering problems. On the other hand, the results here presented have let us obtain conclusions about reutilizing materials with expansive potential in embankment construction. First of all, the importance of the (saturated and unsaturated) permeability of the compacted soil on the deformational behaviour of the embankment is revealed. In the analyzed case, the simulation of a long-term rainfall period on an embankment characterized with the permeability obtained in laboratory tests shows that, in that hypothesis, superficial wetting will have a very slow propagation (in terms of change in the degree of saturation) inside the embankment, resulting in moderate expansive phenomena, with acceptable upward movements in top. It is remarkable that, in this case, to obtain an upward movement greater than 1 cm, it is necessary to consider at least a 50 times greater permeability than that obtained in laboratory tests. Exception to this rule are the lower embankments (less than 10 m height), though it must be remarked that the conservative hypothesis of an impervious base has in that case a clear negative effect on the calculated movements. Nevertheless, this result reaffirms the importance of placing a free-draining base in these cases, in order to ensure a good relative permeability configuration and also to minimize capillary water income. Second, it is also important to remark the effect of the confining stress level at the base of the embankment. In the analyzed cases, embankments higher than 10 m, with saturated permeability equal of greater than k 2,5 109 m/s, may reach a maximum swelling deformation at a certain rainfall duration. For longer periods of precipitation, the sign of the movement may invert due to collapse phenomena happening at the most confined zones of the embankment. As an example, Figure 11 shows the predicted movements
for a 18 m height embankment, with a saturated permeability of k 2.5 109 m/s. Finally, the analysis also suggests that, in some cases, the criteria used in common practice to pre-scribe when expansive materials can be reutilized, may result clearly conservative. Related to this, it is evident how numerical modelling, incorporating the recent Unsaturated Soil Mechanics advances, may become remarkably helpful for the engineer when facing this type of materials, getting a clearer idea of the influence on deformations of the different parameters involved. ACKNOWLEDGMENTS The authors would like to acknowledge Aldesa Construcciones, S.A. and Pai Construccions, S.A. for their permission in exposing the results here presented, and specially thank the support given by the Ministerio de Fomento of Spain, through the ID program Proyecto de Investigacin ID Construccin 2004 (ref. 2004/2). REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique, 40, 3: 405430. Alonso, E.E., Lloret, A. & Romero, E. 2000. Efecto de las lluvias en terraplenes. Simposio sobre Geotecnia de las Infraestructuras del Transporte, Barcelona, 2729 September 2000. Madrid: Sociedad Espaola de Mecnica del Suelo e Ingeniera Geotcnica. Gens, A. & Romero, E. 2000. Ensayos de laboratorio. Simposio sobre Geotecnia de las Infraestructuras del Transporte, Barcelona, 2729 September 2000. Madrid: Sociedad Espaola de Mecnica del Suelo e Ingeniera Geotcnica. Olivella, S., Carrera, J. Gens, A., & Alonso, E.E. 1996. Numerical formulation for simulator (CODE_ BRIGHT) for coupled analysis of saline media. Engng. Comput. 13, 7: 87112. Oteo, C. 2004. Los suelos marginales: caractersticas y aprovechamientos. IV Simposio Nacional de Geotecnia Vial. Santander. Madrid: Asociacin Tcnica de Carreteras. Sanchez, M., Gens, A. do Nascimento Guimaraes, L. & Olivella, S. 2005. A double structure generalized plasticity model for expansive materials. Int. J. Numer. Anal. Meth. Geomech., 29: 751787. van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. Jnl. 44: 892898.
Figure 11. Predicted evolution of the movement in top of a 18 m embankment with a saturated permeability of k 2.5 109 m/s. It can be observed how swelling reaches a maximum at a certain rainfall duration, but after that moment, collapse deformations become predominant.
1336
Banking simulation with a constitutive model for unsaturated soil

M. Tanaka, K. Kawai, S. Kanazawa & A. Iizuka
S. Tachibana S. Ohno
Saitama University, Saitama, Japan Kajima Corporation, Tokyo, Japan
T. Takeyama
National Institute for Land and Infrastructure Management, Kanagawa, Japan
ABSTRACT: It is difficult to estimate the current stress state of existing embankments because of their inherent sloping geometric shapes. Therefore an elastic analysis is always carried out in advance and the calculated stress state, assumed as the current stress, is used as an input parameter in the deformation analysis of an embankment. However, the stress state of a banking material can be quite complicated since banking materials are elasto-plastic and unsaturated in most cases, and, their behavior strongly depends on the stress history. In this study, the banking process is simulated with an elasto-plastic constitutive model for unsaturated soil. Specifically, the effects of banking speed and initial suction within the banking material are considered. It is found that the post-banking stress state depends on the distribution of the initial suction due to soil-water retention characteristics. 1 INTRODUCTION soil strongly depends on its stress history. Therefore, we envisage that the stress state is more complicated than that described by the results of Clough and Woodward. In this study, to predict the stress state of an earth structure, banking simulations are conducted with unsaturated soil/water coupled finite element code, DACSAR-UA. 2 ELASTO-PLASTIC CONSTITUTIVE MODEL FOR UNSATURATED SOIL
The prediction of the behavior of functionallysophisticated earth structures, such as an earth dam and a river bank, is not only important for planning but also can help in their operation and maintenance. Recently, the development of a soil/ water coupled simulation technique has enabled prediction of long-term deformation behavior of newly-constructed earth structures. The initial stress state is used as input parameters in these simulations . However, since the current stress state of an earth structure and the applied stress history is unknown, it is difficult to predict the behavior of an existing earth structure. There is no way to identify the stress state within the structure, which is typically associated with a geometric slope. Clough and Woodward (1967) proposed a way to predict the stress state of embankment using elastic analysis. They showed how the weight of the embankment can affect the simulation results and can affect banking simulation. Their study indicates that the stress state of the earth structure does not depend solely on geometrical condition. Additionally, the soil material for banking is compacted soil, namely unsaturated soil that includes air phase within the soil pores. The behavior of unsaturated
Recently, some elasto-plastic constitutive models have been proposed and some of them have been applied to soil/water coupled analysis (Alonso et al. 1987 & 1990, Kohgo et al. 1993a & 1993b, Karube and Kawai 2001, Kawai et al. 2007). In this study, the effects of hysteresis, appearing on Soil-Water Retention Characteristic Curve (SWRCC), are considered when examining the moisture change during embankment construction. Although the only constitutive model proposed by Karube and Kawai considers the hysteresis of SWRCC, it has a problem in that the application of the driest curve makes complicated numerical calculation and can cause computational instability. The constitutive
1337
model for unsaturated soil proposed by Ohno et al. (2007) is used for the simulations in this study. Ohno et al. first indicated that the yield surface consists of effective stress and a parameter expressing unsaturation in some existing constitutive model for unsaturated soil. They used effective degree of saturation as a parameter contributing to the determination of the yield surface that enabled the application of SWRCC model to express hysteresis. Their model is called Se-Hardening model here. In the model, the effective stress of unsaturated soil is defined as: net ps1 where: net pa1 ps Se s, s pa pw , Se Sr Src 1 Src (2) (3) (1)
Se
Se 1
0
Figure 1.
psat
Se 0
apsat
p
Yield surface for unsaturated soil.
Here, a is the maximum at the residual degree of saturation, and n is fitting parameter relating to shape of the yield surface. The original Cam-Clay model was modified for unsaturated soil and the yield function is written as: f ( , , v p ) MD ln p q D v p 0 psat p (9)
In the above equations, is effective stress ensor, net is net stress tensor, is total stress tensor, 1 is second rank unit tensor, ps is suction stress, s is suction, pa and pw is pore air and water pressure respectively, Se is the effective degree of saturation, Sr is the degree of saturation, and Src is the degree of saturation at s . The volumetric change of the unsaturated soil is written as: e e0 ln p psat (4)
Here, M is the ratio q/p on critical state, D is the dilatancy coefficient, and q is the deviator stress, and written as: q 3 (10) s : s , s p1 2 If equation (8) is substituted into equation (9), the following yield function for unsaturated soil is obtained. f ( , Se , v p ) MD ln D p psat exp 1 Se n ln a (11)
Here, e is void ratio, e0 is void ratio at the yield stress, is the compression index, is the expansion index, p is the mean principal effective stress, and is the ratio of the yield stress to psat , the yield stress at saturation. The volumetric strain, v, is expressed as:
v
p ln 1 e0 psat
q v p 0 p
(5)
The elastic volumetric strain, ve, is expressed as:
This equation is reduced to the original Cam-Clay model at saturation. Figure 1 is the conceptual diagram of unsaturated soil. 3 BANKING SIMULATION USING ELASTIC ANALYSIS
ve
p ln 1 e0 psat
(6)
v ve vp, vp is the plastic volumetric strain, and written as:
vp
p ln 1 e0 psat
(7)
Banking simulations with the linear elastic model were conducted to follow the results of Clough and Woodwards study and compared with results of elasto-plastic constitutive model for unsaturated soil. 3.1 Analysis condition
Ohno et al. assumed the effective degree of saturation as a parameter relating to stiffness, and expressed as: exp [(1 Se)n ln a] (8)
Figure 2 shows the virtual embankment used for the simulations. Its height is 30 m and its width at the bottom is 150 m. Both horizontal direction and vertical direction on the bottom are fixed. The
1338
3.75m
75m
Figure 2. Table 1.
Mesh and boundary condition for analysis. Parameters used for elastic analysis. Poisson ratio 0.40 Unit weight of banking materials sat (kN/m3) 21.2
to one solution. This is an objective solution that is useful in geotechnical engineering. However, Young modulus of soil depends on the confining stress and it changes along vertical direction. Additionally, the material used for embankment is compacted soil, namely unsaturated soil that strongly depends on stress history. Therefore, it is obvious that a constitutive model for unsaturated soil is needed for realistic expression of stress state within an embankment. 4 BANKING SIMULATION WITH ELASTO-PLASTIC CONSTITUTIVE MODEL FOR UNSATURATED SOIL
Young modulus E (kN/m2) 9.58 103
30m
Figure 3.
Simulation results of linear elastic model.
vertical direction on the left boundary is allowed. Table 1 shows the parameters used for the simulation. Here, only the weight of the embankment acts as the external force. Two methods were used to apply the weight. In the first case the deadweight is activated on all elements at the same time. In the other case, the deadweight is activated sequentially from the bottom of the embankment to the top. These two methods are called single lift and lift increments, respectively, following Clough and Woodward. 3.2 Analysis results
The banking material is compacted soil, namely unsaturated soil that has air phase in its void. Since the effective stress of unsaturated soil is expressed as the product of suction and soil moisture as in equation (3), inner stress distribution of embankment depends not only on suction but also on soil moisture. When earth structures are constructed, soil material, adjusted to certain water content, is overlaid with a certain thickness and compacted. If the degree of water content is linearly correlated with suction, we can assume that the initial pressure head due to suction distribution within the embankment is constant. In this case, the downward flux occurs at the beginning of the analysis. Therefore, the effect of initial suction distribution is also examined here. 4.1 Analysis condition
The virtual embankment used is the same as in the elastic analysis shown in Figure 2. The flux is not allowed through all boundaries. Table 2 shows the parameters used for the simulation. Additionally, Figure 4 shows the soil-water retention characteristics. The SWRCC model proposed by Kawai et al. (2007) is used. Their model can express the hysteresis
Table 2. Parameters used for unsaturated soil/water coupled simulation. 0.180 0.33 n 1.0 k (m/day) 0.01 0.037 psat (kN/m2) 196 e0 1.20 m 0.8 M 1.333 a 150 t (kN/m3) 14.7
Figure 3 shows analysis results with the linear elastic model. The left and the right halves correspond to single lift and lift increments respectively. The horizontal and vertical stresses are slightly different between the case of single lift and that of lift increments. The biggest difference can be seen along the vertical axis. This is because that the summation of strain of all elements along the vertical direction becomes vertical displacement in the case of single lift while the underlying displacement does not influence the overlying displacement in the case of lift increments as Clough and Woodward noted. With a detailed analysis, the thickness of the layer lifted at one time is reduced and converge
k: Saturated hydraulic conductivity, m: Parameter used for unsaturated hydraulic conductivity equation of Mualem (1976).
1339
in the relationship between suction and the degree of saturation. Two methods were used to determine initial suction distribution. In one simulation, initial total head of all elements are constant. This indicates that suction changes along vertical direction. The plots in Figure 4 show the initial suction and degree of saturation in constant total head simulation. In the other simulation, a constant pressure head is provided. An initial suction of 333.2 kPa and degree of saturation of 0.6 are applied in lifted layer in the constant pressure head simulation. 4.2 Analysis results
Figure 5 shows the results of the constant total head simulation in lift increments. These results correspond to steady state conditions after the embankment had been constructed. Suction and the degree of saturation were influenced slightly by the deformation since the initial suction value provided according to the height to make initial flow steady. The effective stress is bigger than overburden pressure at high position due to suction
1
Srf =1.00 Logistic curve Eq.11) Srf - Src Sr = +S 1+exp( A+Blns) rc

ing Dry
0.8
0.6
0.4
0.2
Src=0.15
0 0 200 400 600 800
Suction (kPa)
Figure 4.
Soil-water retention characteristics.
effect. Therefore, the difference in vertical stress along the vertical direction is smaller than that in the elastic analysis. Moreover, it is found that a certain amount of horizontal stress appears even near the slope. This is because that suction contributes isotropically toward effective stress. Figure 6 shows the results of the constant pressure head simulation in lift increments. Suction distribution profile indicates the system to be present in a steady state. Since all boundaries are impermeable in this simulation, redistribution of suction occurs due to inner water balance after construction of the embankment. Therefore, even at the bottom of embankment, suction is positive and the whole embankment is unsaturated, as shown in Figure 6(d). Consequently, stress along both horizontal and vertical directions is larger than that in constant total head simulation. It is found that the trend of horizontal stress distribution is different between constant total head and pressure head simulations. This is because suction stress, which contributes to the effective stress, is the product of suction and the effective degree of saturation and depends on SWRCC. The SWRCC indicates that stronger suction is necessarily for larger suction stress. Relatively larger suction stress incidentally appears in the middle of the embankment in constant pressure head simulation as shown in Figure 6(c) and (d). This influences displacement. Especially horizontal displacement in constant pressure head simulation differs from that in elastic analysis or in constant total head simulation. However, the displacement is smaller than that in constant total head simulation since suction is bigger on the whole. It is found that the initial suction distribution causes inner flow and can complicate stress distribution. This result indicates that we cannot estimate current stress state of an embankment from overburden pressure and groundwater level.
Wettin -23.1 g: A = , B=4 .6
: A= 7 -34. , B= 5.9
Figure 5.
Simulation results (constant total head).
1340
Figure 6.
Simulation results (constant pressure head).
Figure 7.
Finite element meshes.
DEPENDENCIES OF STRESS AND DISPLACEMENT DISTRIBUTION ON SHAPE OF SLOPE
Here, the banking simulation is performed on 3 new virtual fills with unsaturated soil/water coupled analysis. 5.1 Analysis conditions
Figure 8. Horizontal stress distributions.
The angle of slope in Figure 2 is 22.8-degree. Based on this embankment, fresh banking simulations were performed on embankments of 30-degree, 45-degree and 60-degree. Figures 7 show finite element mesh used for the simulations. All simulations were performed with constant initial pressure head in lift increments. 5.2 Analysis results
Figures 8 show the horizontal stress distributions of the analysis results. On every embankment, dark areas appear in the middle of the slope surface. The origin of this is thought to be suction stress since the amount of horizontal stress on the slope surface is identical. The effect is smaller in steeper embankments. This is because in steep embankments vertical stress is relatively large and the effect of suction stress on horizontal stress is unnoticeable. Large horizontal stress also appears in the center of the bottom in steep embankments. This corresponds to the result of elastic analysis.
Figure 9. Horizontal displacement distributions.
Figure 9 shows the horizontal displacement distribution. There are obvious differences among embankments of differing shapes. In loose embankments, only the slope surface changed. The more abrupt the angle of the
1341
embankment is, the greater the displacement. These results suggest that surface failure occurs in the low angle slope and circular slip occurs in the steep angle slope. 6 CONCLUSIONS
REFERENCES
Alonso, E.E., Gens, A. & Hight, D.W. 1987. Special problem soils. General Report., Proc. 9th Europian RC-SMFE, 10871146. Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils, Gotechnique, 40(3):405430. Clough, R.W. & Woodward, R.J. 1967. Analysis of embankment stresses and deformations, Journal of the Soil Mechanics and Foundations Division, Proc. of the ASCE: 529536. Karube, D. & Kawai, K. 2001. The role of pore water in the mechanical behavior of unsaturated soils, Geotechnical and Geological Engineering, 19(3):211241. Kawai, K., Iizuka, A., Hayakawa, E. & Wang, W. 2007. Non-uniform settlement of compacted earth structures caused by the deformation characteristics of unsaturated soil on wetting, Soils and Foundations, 47(2):195206. Kohgo, Y., Nakano, M. & Miyazaki, T. 1993a. Theoretical aspects of constituitive modeling for unsaturated soils, Soils and Foundations, 33(4):4963. Kohgo, Y., Nakano, M. & Miyazaki, T. 1993b. Verification of the generalized elasto-plastic model for unsaturated soils, Soils and Foundations, 33(4):6473. Mualem, Y. 1976. A new model for predicting the hydraulic conductivity of unsaturated porous media, Water Resources Research, 12(3):514522. Ohno, S., Kawai, K. & Tachibana, S. 2007. Elasto-plastic constitutive model for unsaturated soil applied effective degree of saturation as a parameter expressing stiffness, Journal of JSCE, 63(4):11321141 (in Japanese).
In this study, the banking simulations were performed with the constitutive model for unsaturated soil under various conditions. As it is obvious from the study of Clough and Woodward, it is found that there are some differences due to the construction process even if elastic analysis was performed. Nowadays, the elastic analysis is conducted first and the obtained stress state is assumed to correspond to the initial condition when the behavior of the earth structure including slope is predicted. However, the solutions obtained can differ from the real value due to stress history associated with construction or generation of the earth structure. More sophisticated banking simulation is preferable for realistic solution. Since compacted soil is used for banking, a constitutive model for unsaturated soil is needed. It is found, through unsaturated soil/water coupled analysis, that the initial suction distribution strongly influences the stress state in the embankment. The existing embankments are exposed to climate conditions, such as rainfall and evaporation, and influenced by vegetation uptake. As a result, suction and soil moisture show complicated distribution. This study indicates the importance of field site investigation.
1342
Pavements
Water migration in pavement subgrade

E.J. Nelson & D.D. Overton
ABSTRACT: The development of residential subdivisions in the Southwestern United States is typically conducted in four basic stages: 1) overlot grading to establish the rough grades and elevations of the subdivision; 2) fine grading to establish individual lot elevations and street grades; 3) installation of utilities, streets, curb and gutter, and sidewalks; and 4) construction of residences and final grading of the individual lots. The time period from start to finish of a subdivision can range from several months to years before the site is completely built out. Subsequent to installation of the curb and gutter and prior to completion of the home, lots are typically graded to drain toward the street and the soil on the lot is left low to accommodate installation of landscaping after the home is completed. Additionally, joint utilities such as electrical and telephone are installed just behind the curb and gutter or sidewalks. The backfill of joint trench utilities are often poorly compacted. During this time the site is exposed to precipitation in the form of rain and snow. Runoff and snow melt is directed toward the street and often ponds behind the curb and gutter or sidewalk and on top of the joint trench. This ponding provides a source of water that can travel along the interface between the concrete or pavement and the underlying subgrade. This ultimately results in elevated water contents in the pavement subgrade which can cause heave from soil expansion or frost, loss of subgrade support, and settlement of utility trench backfill. This paper presents actual field data that shows that soil water contents are highest adjacent to the curb resulting in settlement of the trench backfill and heave of the pavement subgrade. Vadose/W modeling was also performed and the results indicate that current grading practices can have a detrimental effect on the ultimate performance of pavements by promoting early saturation of the subgrade materials and utility trench backfill. 1 INTRODUCTION gutter. However, in order to leave room for future landscaping and foundation spoil disposal during building construction, the soil behind sidewalks and curb and gutter is left low. This practice provides a location behind the curb or sidewalk where water can pond. The installation of dry utilities such as gas, phone and electrical are also often located in the future landscape areas between the street and the structure. Compaction of dry utility trench backfill is often not properly compacted or tested. Settlement of the backfill over the dry utility trenches provides another location where site runoff can pond. The dry utility trenches are particularly damaging in that they cross over the water and sewer laterals entering the lots. These water and sewer laterals are typically bedded in sand backfill. This provides a direct conduit for water to enter into the more permeable backfill materials of the wet utility trenches. 2 EXAMPLE SITES
Pavement design manuals discuss the importance of positive drainage at the edge of pavement. In fact one of the design parameters used in design is the selection of a proper drainage coefficient. Geotechnical reports prepared along the Front Range of Colorado often recommend a positive slope of 5 to 10 percent away from the edge of pavements to reduce maintenance and ensure the proper performance of the pavement over its design service life. Unfortunately many developers and site development plans do not implement these recommendations citing space constraints or excessive costs for implementation. The lack of proper drainage is more obvious during the early stages of site development after the streets have been completed but prior to construction of the structures in the lot areas. The impacts of poor grading on the migration of water has been presented in previous studies (Chao et al. 2006 and Overton et al. 2006). Buried utilities, streets and curb and gutter are the first permanent site improvements installed on a development. Quite often the adjacent lots are rough graded to drain toward the street with the intent that runoff will be carried away by the curb and
Backfill settlement within the roadways of several developments in Colorado USA were investigated
1345
to determine the cause of settlements in utility trench backfill below pavements. One of the common factors encountered at each of the sites was the lack of proper drainage behind the curb and gutter or adja-cent to the pavement itself. In two cases runoff from roads and hillsides above the road was directed to the back of curb without any drainage ditches and was expected to spill over the curb into the gutter before entering into the storm water collection system. In other cases drainage ditches were provided on the back of the curb and gutter. However they were located within 0.6 meters of the roadway, and during the winter months snow is plowed into these ditches where, in the spring, it begins to slowly melt and saturate the road embankment and subgrade. Three drainage scenarios are reviewed in this paper. The first scenario is representative of a good grading practice with the overlot grading sloping at 2% away from the curb and gutter. A dry utility trench, 0.6 meters deep, is included behind the curb and gutter and this trench backfill is adequately compacted. This scenario is shown as Figure 1. The second scenario is of poor grading where the overlot grading slopes to the curb and gutter with a 5 percent slope. Additionally, the slope is offset 100 mm below the top of the curb in order to accommodate future landscaping. These grades are typically seen in residential development. The resulting depression behind the curb and gutter provides a low point where runoff from rainfall and snow melt can accumulate. This water typically ponds until it is either evaporated or infiltrates. This scenario is shown in Figure 2. The third scenario is of the same poor grading as scenario two, however, a dry utility trench is assumed to exist behind the curb and gutter. These dry utility trenches are typically 0.6 meters deep. Because of the uncompacted nature of the backfill, the soil in the dry utility trench typically settles. Settlement of 75 to 300 mm has been documented
CURB & GUTTER
SETTLED DRY UTILITY TRENCH

5% SLOPE
SIDEWALK
CURB & GUTTER
ROADWAY
c L
Figure 3. Scenario 3, poor grading practice with dry utility trench.
26 In-Place Water Content (percent) 24 22 20 18 16 14 12 10 0 0.5 1 1.5
Water Content Linear Regression of Water Content
2.5
Distance from Curb (meters)
Figure 4. Comparison of water content data with distance from curb.
for these dry utility trenches. Water runoff from the lot ponds in the trench depression and infiltrates into the soil. This scenario is shown in Figure 3. 3 WATER CONTENT DATA
SIDEWALK
2% SLOPE
3.1 Field data

ROADWAY
COMPACTED DRY UTILITY TRENCH
c L
Figure 1.
Scenario 1, recommended grading practice.
SIDEWALK
5% SLOPE
CURB & GUTTER
ROADWAY
c L
Figure 2.
Scenario 2, poor grading practice.
Soil samples were collected and analyzed during the investigation of various sites. Evaluation of the laboratory data indicated that the water content of the surface subgrade soils was higher next to the curb. Figure 4 shows the distribution of soil water content as it relates to the sample distance from the edge of pavement. In order to eliminate the potential for density variations to affect the water content of the soil, the authors also compared the percent saturation of the soils with their distance from the edge of pavement as shown in Figure 5. It is evident upon review of Figure 5 that the amount of water in the pavements was higher closer to the edge of pavement.
1346
100
Volumetric Water Content (%) 0 5 10 15 20 25 30 35 40

Sandy CL Fill
Percent Saturation (percent)
90
0 1
80
Depth Below Ground Surface (meters)
2 3 4 5 6 7
70
60
Saturation Linear Regression of Saturation
50 0 0.5 1 1.5 2 2.5 3 Distance from Curb (meters)
After 2 years After 5 years
Figure 5. Comparison of percent saturation with distance from curb.
3.2
Modeling results
Figure 6. Water content change at edge of pavement. 2 percent grade away from sidewalk.
3.2.1 Recommended grading profile The first model scenario (Scenario 1) incorporated typical geotechnical engineering recommendations of a 2 percent grade away from the back of the sidewalk. Figures 6 and 7 show the model output for the first scenario at the edge of pavement and the centerline of the roadway. This model showed that the water content of the soil varied only slightly from the water content at which the soil was originally placed. Confirming that proper drainage and grade behind the pavement results in a minimal increase of water content in the subgrade soil. 3.2.2 Common grading practices The model for Scenario 2 used a ground surface that drained toward the pavement at a slope of
6 7 8
Figure 7. Water content change at centerline of roadway. 2 percent grade away from sidewalk.
5 percent and the soil behind the curb and gutter was left approximately 100 mm low in order to accommodate future landscaping. It was assumed, as is the case in many rough graded developments, that water would pond behind the curb until it
1347
Claystone
Modeling was performed using the Vadose/W program from Geo-Slope International. The purpose of this exercise was to use an analytical tool to compare with the field results. The three scenarios described in Section 2 of this paper were analyzed using Vadose/W. All of the scenarios were set up to evaluate the rate of water migration in the top 7.6 meters of soil below a pavement section during the first 5 years of service life. All of the models incorporated 2.4 meters of fill soil over claystone. The top 1.2 meters of the claystone section was assumed to be a weathered claystone typical of the Colorado Front Range. Seasonal precipitation was assumed in all of the models run. Water input from post construction irrigation was not accounted for in these models.
Volumetric Water Content (%) 0 0 1 2 3 4 5

Sandy CL Fill
10
15
20
25
30
35
40
Initial
Weathered Claystone
Claystone
Initial
Weathered Claystone
either percolates into the ground or evaporates. The grades modeled in this scenario are commonly created during fine grading. The resulting low point behind the curb and sidewalk provides a location where runoff from rainfall and snow melt can accumulate. Since the model used local climatic conditions the rate of infiltration and evaporation was included as part of the analysis. Figures 8 and 9 show the results of the model run for Scenario 2. This model output shows that the water content of the subgrade soil increases over time when compared to the subgrade soil below a pavement where proper drainage has been provided as modeled in Scenario 1. The model for Scenario 3 assumed a similar ground surface to Scenario 2. However, this scenario also incorporated an open dry utility trench 0.6 meters deep, located 1.5 meters behind the curb. Dry utilities (telephone, electrical, and gas) are often installed behind the curb and gutter after the roadways have been installed. It is not uncommon for the dry utility trenches to be backfilled with uncompacted fill soil. This practice results in settlement of the backfill soil that provides a location for additional ponding and also transmits water deeper into the subgrade soil. Because of the uncompacted nature of the backfill, the soil in the dry utility trench was assumed to be one order of magnitude more permeable than the surrounding fill soil. It was
0 0 1
Volumetric Water Content (%) 10 15 20 25 30 35
40
Sandy CL Fill
2 3 4 5 6 7 8
Figure 9. Water content change at centerline of roadway. 5 percent grade toward sidewalk.
0 0 1
10
15
20
25
30
35
40
Sandy CL Fill
Volumetric Water Content (%) 0 0

Sandy CL Fill
10
15
20
25
30
35
40
2 3 4 5 6 7 8
Weathered Claystone
Initial
Claystone
Figure 10. Water content change at edge of pavement. 5 percent grade toward sidewalk with uncompacted dry utility trench.
Figure 8. Water content change at edge of pavement. 5 percent grade toward sidewalk.
assumed that the back fill soil would settle approximately 150 mm. Ponded water was assumed in the dry utility trench backfill as part of this scenario. The frequency of ponding was input as a function of the local climatic data. Figures 10 and 11 show the results of this scenario.
1348
Claystone
Initial
Weathered Claystone
Claystone
Initial
Weathered Claystone
0 0 1
Volumetric Water Content (%) 5 10 15 20 25 30 35
40
Sandy CL Fill
2 3 4 5 6 7 8
Figure 11. Water content change at centerline of roadway. 5 percent grade toward sidewalk with uncompacted dry utility trench.
Claystone
Initial
are often the result of poor drainage in rough and finish graded sites. If proper drainage is implemented during construction the degree of settlement and other pavement distress would be lessened. Geotechnical and pavement engineers typically assume that their drainage and grade recommendations will be implemented during design. However, the reality is their recommendations are often not implemented because of space constraints in higher density developments or the developers desire to cut costs. The installation of proper drainage features such as swales, and area drains could significantly improve the performance of the pavement in many areas. Furthermore, pavement engineers may find it appropriate use drainage coefficients for poor drainage conditions instead of assuming that their recommendations for drainage and grade will be implemented. The assumption that poor drainage conditions will exist during and after construction is complete will result in pavements that perform better under final grade conditions. REFERENCES
Chao, K.C., Overton, D.D. & Nelson, J.D. 2006. The Effects of Site Conditions on the Predicted Time Rate of Heave, American Society of Civil Engineers, Unsaturated Soils Conf. 2006. Special Publication 147. Engineering Analytics, Inc. 2008. Response to CTL/ Thompson Report Geotechnical Consultation, Utility Trench Backfill Settlement, Fallbrook Farms, Filing No. 1, Thornton, Colorado. October 20. Engineering Analytics, Inc. 2008. Supplement Data Report in Support of Response to CTL/Thompson Report Geotechnical Consultation Utility Trench Backfill Settlement Fallbrook Farm, Filing No. 1, Thornton, Colorado. November 17. Engineering Analytics, Inc. 2009. Results of Geotechnical Investigation Granby Ranch Filings 2, 2B and 5 Granby, Colorado. September 14. GEO-SLOPE International, Ltd. 2007. GEO-STUDIO VADOSE/W Software Package for Seepage Analysis, Version 7.15. Calgary, Alberta, Canada. Overton, D.D., Chao, K.C. & Nelson, J.D. 2006. Time Rate of Heave Prediction for Expansive Soils. American Society of Civil Engineers, Proceedings of GeoCongress 2006, Geotechnical Engineering in the Information Technology Age.
The model output clearly shows that the presence of a depression, such as settlement in a trench adjacent to the pavement, increases the water content of the soil above that resulting from poor grading. 4 CONCLUSIONS
Both field data and computer modeling results clearly show that the water content of pavement subgrade soils are affected by drainage conditions adjacent to the pavement. The modeling results further indicate that the worse the drainage behind the curb, the greater the increase in water content in the pavement subgrade. Although the recommendations of the pavement engineer with respect to drainage behind the curb are often not implemented during design, the effects of poor drainage can and do have significant effects on the performance and durability of pavements. Damage to pavements from settlement, edge breaks and in some cases lateral displacement
Weathered Claystone
1349
Geoenvironmental applications
Gas migration in sand/bentonite mixtures through preferential paths

D. Arnedo, E.E. Alonso, S. Olivella & E. Romero
Geotechnical Engineering and Geosciences, UPC Barcelona, Spain
ABSTRACT: Sand/bentonite mixtures are considered as buffer material in low and medium radioactive waste disposal schemes. While the bentonite provides low permeability and swelling capacity, the sand is intended to improve mechanical stability and to reduce the amount of bentonite required. One of the requirements of the buffer is to ensure the isolation of the gas-emitting waste. The pressurized gas may migrate towards the host geological media through several mechanisms (dissolution, two-phase flow, preferential paths). This paper is a contribution to this topic and presents experimental results as well as some modelling work. Laboratory experiments were performed on statically compacted 80/20 sand/bentonite mixture recovered from the in situ large scale GMT (Gas Migration Test) buffer. The sample was tested in a specifically designed Perspex test cell which allows the visualization of the boundary of the sample. Two needles inserted inside the sample are used to inject and recover gas. The formation of preferential paths following the compaction layers and the sample-wall interface were detected in the test. The test is analyzed by hydro-mechanical numerical simulations considering randomly distributed properties (porosity, permeability and hardening parameter) and the possibility of opening discontinuities in the permeability law are considered. Also a computational experiment inspired in the conditions prevailing on the in situ large scale GMT test, is presented. 1 INTRODUCTION the conditions prevailing on the in situ large scale GMT test.
Sand/bentonite mixtures are foreseen as buffer material for the engineered barrier system in low and medium radioactive waste disposal schemes in deep geological formations. The generation of gases due to several mechanisms, being the anaerobic corrosion of the steel canisters the most important, becomes a challenge for the long-term safety of the repository. The pressurized gas may migrate towards the host geological media through several mechanisms (dissolution, two-phase flow, preferential paths), releasing radioactive waste into the environment. Preferential paths are believed to play a major role in gas migration once a free gas phase is formed. However, the role of previously existing interfaces (e.g. compaction layers) and the formation of newly created paths are currently poorly understood. This work presents laboratory experiments where the formation of preferential paths is observed. The test presented is simulated numerically in a first effort to reproduce preferential paths in 3D. Finally, the formation and development of marked preferential paths in 2D is presented through a computational experiment inspired in
2 2.1
TEST DESCRIPITION AND RESULTS Tested material
The experimental apparatus is composed by a Perspex cylinder, 70 mm in diameter and 200 mm in height. The gas (air) injection and recovery is performed using two needles inserted into the sample. The tests were performed at the controlled temperature room (22C with 1C variation). The tested material is a sand/bentonite mixture (80/20 in dry mass basis) recovered from the dismantling of the GMT (Gas Migration Test) experiment. The sand/bentonite (80/20) mixture, statically compacted in the laboratory, has been characterized in several studies in order to determine its hydro-mechanical behaviour (Romero et al. 2002, 2003). The material used in this study was extracted from layer 5 and 8 of the GMT engineered barrier. A total of 6 series of tests were performed. The tested material was taken from layer 5 of the
1353
GMT in situ test. Samples have been statically recompacted inside the cell cylinder, in 10 layers, at a prescribed value of dry density. The test cell did not allow to properly saturate the sample before the gas injection stage because the water injection had to be made through the needles and given the sample dimensions saturation times would have been excessively long. For that reason, compaction water content was adjusted to obtain an initial degree of saturation higher than 90% (Table 1). 2.2 Description of test s4
For space reasons, only detailed results from test s4 (series 4) will be presented in this paper together with a modelling contribution. The dry density of the sample was 1.70 Mg/m3. The initial degree of saturation was 0.93. Gas injection was performed by displacing the injection piston at a rate of 1.737 mm3/s. The pressure in the downstream point was maintained at 3 kPa. Figure 1 shows the complete record of the test (injection system pressure, recovery system volume variation and recovery system pressure). In the first stage (until the injection pressure peak) the
Table 1. Initial conditions of test s4 sample. Series 4 2.65 1.70 0.193 0.55 0.93
Sand/Bentonite 80/20 (Layer 5 of GMT Test) Density of solid particles, s (Mg/m3) Dry density, d (Mg/m3) Water content, w Void ratio, e Degree of saturation, Sr
injection was performed at constant velocity of the piston. Once the injection piston reached its maximum displacement it stopped and the volume of the injection system remained constant for the remainder of the test. Two pressure peaks were recorded. First, a small pressure drop of 10 kPa was recorded at a pressure of 179 kPa. This small peak is interpreted as the beginning of inflow of air into the sample. It was also confirmed by visual inspection of the cell wall-sample contact through the Perspex cylinder. The second peak was recorded at an injection pressure of 1876 kPa. It seems that this peak took place simultaneously with the achievement of the maximum displacement of the piston. Pressure dropped 136 kPa immediately after the peak. The recovery system was maintained at a constant pressure of 3 kPa and its volume variation recorded. Air recovery started just before the injection pressure peak and increased sharply after the injection pressure drop, and remained oscillating around 4500 to 5000 mm3/h. Once the recovery piston reached its maximum displacement, the volume of the recovery system remained constant for the remainder of the test, and the outgoing mass of air was derived from the recovery system pressure build-up. The volume of injection system was determined by fitting the initial pressure record (before the first peak) to the pressure curve for a constant mass system which is reducing its volume at the same rate of 1.737 mm3/s. A volume close to 106 mm3 was obtained i.e. without deviation with respect to the expected estimation. The difference between the constant mass system pressure and the recorded injection pressure during the tests was used to estimate the injected mass of air into the sample, using the following relationship: n(t ) Pcms (t ) Pinj (t ) V (t ) RT (1)
Figure 1. Results for the gas injection test in a sand/ bentonite mixture. Pressures on the left axis and flow rate on the right axis (measurements).
where n corresponds to the mols of air flowing into the sample, Pcms is the pressure for the constant mass system, Pinj is the recorded injection system pressure, V is the injection system volume record, R is the ideal gas constant and T is the absolute temperature. Figure 2 shows the cumulative mass of air flowing into and out of the sample against time (molecular mass of air is 28.95 gr/mol). Their derivatives give the estimated mass flow rates, plotted in Figure 3. Figure 4 shows pictures of the test cell after testing. Some preferential paths mainly located in the contact between the sample and the cell wall. The opening of some compaction layer interfaces is also observed. Those zones were just detected after the small pressure peak recorded at 179 kPa,
1354
which is interpreted as the beginning of inflow of air into sample following a preferential horizontal pathway through some compaction layer interface. 3 MODELLING TEST S4
Figure 2. Estimated cumulative air mass inflow and outflow during the test (measurements).
Figure 3. Estimated injection and recovery flow rates (measurements).
In this section results from a modelling work performed to simulate the gas injection tests using a coupled approach are summarized (for details see Arnedo et al. 2007). The coupled solution of the mechanical and two phase flow equations is performed using finite elements with CODE_ BRIGHT (Olivella et al. 1996). An approach for gas and water flow through discontinuities including mechanical interactions is used. This approach has been described in detail in Alonso et al. (2006) and in Olivella & Alonso (2008). Model calculations are compared with the injection/recovery pressures and the injection/recovery flow rates. To properly simulate the test, the sample has been represented in 3D. In addition, both the injection and recovery systems have to be included in the model. This is achieved by the use of linear elements simulating the needles and the injection/recovery pistons (no mechanical properties and connected to the sample at the injection and recovery points). These elements are idealized with a porosity of n 1, high conductivity (permeability of 102 m2) and a flat retention curve with a very low air entry value (very fast saturation/ desaturation). The total volume of each system corresponds to the volume values obtained from the test. 3.1 Boundary conditions Constant volume conditions are applied. The flow boundary conditions are applied at the end node of the linear elements representing the injection and recovery system, which are connected at the appropriate point in the sample. Figure 5 shows a scheme of the injection and recovery systems together with the steps considered in the simulation for each system, which corresponds to the two stages of each system during test s4 described previously. The recovery system is initially saturated. The injection system comprises two linear elements. The displacement of the injection piston is simulated by the injection of water through the boundary node of the first element, at the same rate of volume variation applied during the test. The second element of the injection system has a very small volume, and it is almost impervious to water (krw 1020 Sr2). This second element is necessary in order to avoid the injection of water into the sample.
Figure 4. Images of the samples after gas injection. (a), (b): preferential path along cell-sample contact and opening of compaction interface. Horizontal discontinuities indicate the boundaries of compaction layers.
1355
Table 2. model. E (MPa) s M
Parameters of the mechanical constitutive 50 0.35 0.0001 0.001 1.07 (0) r (MPa1) K p0*(mean value) (MPa) 0.049 7.0 70 0.03 0.26*
Table 3. model.
Parameters for the element permeability 5.0 1018 0.355 107 1.6 102
Figure 5. Scheme of the injection and recovery systems and steps of the analysis (in days). Finite element mesh for numerical simulation including a representation of initial porosities.
Matrix intrinsic permeability, k (m2) Reference porosity Initial fracture aperture, b0 (m) Threshold strain, 0
3.2
Constitutive laws
The mechanical model adopted for the sand/ bentonite mixture is an elasto-viscoplastic model based on BBM (Barcelona Basic Model, Alonso et al. 1990) in which the behaviour of the soil depends on net stresses and suction. Table 2 shows the mechanical parameters used in the present calculations and determined from different hydromechanical tests reported by Romero et al. (2002, 2003). In the permeability model, the effect of the presence of a planar discontinuity embedded in each element is considered. The equivalent element permeability is computed as follows, assuming laminar flow inside the planar fracture (cubic law): kelement k fracture b s b b3 kmatrix kmatrix (2) s s 12 s
The variation of capillary pressure caused by changes in the aperture is also included. Considering Kelvins law to describe the capillary pressure necessary to desaturate a fracture, the air entry value of the elements is obtained as: P k fo 2 2 bo Po 3 kf b bo b (4)
For the retention curve of the material the van Genuchten (1980) model is adopted, while a power law is considered for both liquid and gas relative permeability (see Table 3). Some heterogeneity of the hydraulic and mechanical properties has been considered. A random porosity field with the same mean porosity as the tested material (mean 0.355) is generated. This means that the density of the materials is different depending on the element. Then, for each element, a volumetric deformation associated to the degree of compaction is computed from the mean void ratio, emean:
where b is the fracture aperture and s is a characteristic dimension. Intrinsic permeability of the matrix material is dependent on porosity through Kozenys law. Changes in aperture are computed from the element deformation: b b0 b b s s(o) (3)
v
e emean 1 emean
emean
mean 1 mean
(5)
The hardening parameter of the mechanical constitutive model, p0*, is obtained as follows for each element: 1 emean * * p0 p0 mean exp v (6)
where 0 is a threshold strain necessary to start aperture variations. The calculated normal strain or volumetric strain is changed into a discrete aperture through equation (3). In this work the computed permeability changes are isotropic. Each element may develop a different permeability as a function of aperture but it is considered locally isotropic.
where (-) is the plastic compressibility parameter of the soil. In the case presented, the random porosity field has uniform distribution, with no spatial correlation, and a variance in the order of 105. The small variance selected is consistent with some local determination of density reported
1356
in Delahaye & Alonso (2002). This procedure was successful in developing clear preferential paths in 2D analysis for a buffer material representing the GMT in situ test which are presented in the next section (see also Alonso et al. 2006). Results in 3D indicate that the preferential paths are smoother as compared with 2D. 3.3 Results
Figure 6 shows the computed injection and recovery system pressures against the measured pressures during test s4. The injection system pressure build-up is very well captured, although the maximum pressure achieved is slightly smaller. After the pressure peak, once the injection piston is stopped, the pressure falls, more rapidly in the model than in the experiment. However, the injection pressure ends at the same value that was recorded in the test. Recovery system pressure rises faster indicating that gas flow is higher in the model than in the test. Figure 7 shows the computed inflow of air versus the inflow of air during the test. The general trends are well captured, especially the shape of the curve. In this case the model is not able to reproduce the fast decay of the flow immediately after
Figure 8. Computed versus estimated outflow of air.
Figure 9. Computed gas phase flux vectors (m/s). (a) 156.5 hours (pressure peak); (b) 204.5 hours.
Figure 6. sures.
Computed/measured injection/recovery pres-
the injection pressure peak. This drop is possibly related to the fast closure of a gas pathway in the vicinity of the injection needle in the test, just after the pressure drop. Once the movement of the injection piston is stopped, the computed inflow rate of air remains one order of magnitude higher than the values obtained during the test. The same is observed in the computed outflow rate, shown in Figure 8. Plastic deformations are computed in the vicinity of the injection needle. Figure 9a shows the advective gas flux vectors at 156.5 hours (pressure peak). Gas flows mainly through the zone near the injection point. No clear preferential flow zones are visible, and the calculated air flow remains essentially spherical. Figure 9b shows the fluxes at 204.5 hours, during the injection pressure decay (at constant volume of the injection system). A dipole type of flow is established within the specimen. Injected gas is now well connected with the recovery point. 4 HETEROGENEOUS ANALISYS OF GMT MAIN TRANSMISSION ZONE
Figure 7. Computed versus estimated inflow of air.
A large scale gas migration test was conducted by NAGRA (the Swiss radioactive waste management agency) at the Grimsel Test Site. The engineered barrier (the cylindrical concrete silo and the compacted sand-bentonite mixture barrier) were built
1357
in place. Then the access drift was backfilled with sand. In order to saturate the compacted sandbentonite barrier the system was pressurized at 0.55 MPa with water coming from a pervious ring located at the backfill-barrier contact. Gas was injected inside the silo during several stages and it was allowed to escape through a mortar vent at the top of the silo. The barrier was highly instrumented to detect stress and fluid pressure variations. Figure 10 shows a detail of the engineered barrier. A main zone for gas migration through the sand-bentonite barrier (layers 9 and 10) was identified during the experiment. Further data and interpretation of results is reported in the modelling analysis of GMT (Olivella & Alonso 2005). The simulations presented here are focused in the main transmission zone and include the effects of the heterogeneity of hydraulic and mechanical properties of the sand-bentonite mixture. 4.1 Boundary conditions
4.2
Constitutive laws
The same hydro-mechanical model described in the previous section is considered. The parameters of the mechanical constitutive model are the same as in the previous section (table 2). In the embedded fracture permeability model the threshold strain, 0, has been reduced to 103. In order to simulate the sand-bentonite heterogeneity the same procedure described in the previous section is used. 4.3 Results
A water pressure of 0.55 MPa was applied on the boundary (1) in order to simulate the pressurization and saturation stage of GMT. A prescribed mass flow of gas (5.2 106 kg/s, corresponding to the RGI-3b gas injection phase of GMT) was injected at the mortar vent boundary (2) from day 315 until the end of the analysis (day 4500). Gas was allowed to escape through the boundary (1) when the gas pressure exceeded the prescribed water pressure of 0.55 MPa.
Figure 10 shows also the initial porosity field considered for case 1. The mean value is 0.3 and the variance is in the order of 105 with random uniform distribution and no spatial correlation. Due to the small variations in porosity considered, initial permeability of all elements, which follows Kozenys law, therefore there is no initially marked heterogeneity. This situation will change as the elements deform in response to changes in net stresses (effect of pressurization, saturation and gas injection). Figure 12 shows the advective fluxes of gas (m/s) At day 335, after 20 days of gas injection, a marked preferential path has developed near the injection boundary. This path becomes progressively closed and a new main path is developed from day 400 to the end of the analysis. Figure 13 shows gas fluxes at day 4500, once steady state has been reached. Typical gas pressure peak is calculated and permeability increases more than 2 orders of magnitude due to the deformation undergone by the element. Figures 13 and 14 show the development of a preferential path obtained in a different simulation (flux vectors). The initial porosity field is a different realization of the same random field of the case presented above (uniform distribution, no spatial correlation). In this second case the main path progresses further than in the first case, reaching the top of the domain.
Figure 10. Detail of the GMT engineered barrier (dimensions in millimetres). Scheme of the boundary conditions and finite element mesh and initial porosity field. Case 1.
Figure 12. Advective gas fluxes (m/s). Case 1. Day 4500. End of the analysis.
Figure 11.
Advective gas fluxes (m/s). Case 1. Day 335.
Figure 13.
1358
REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Alonso E.E., Olivella, S. & Arnedo, D. 2006, Mechanisms of gas transport in clay barriers. Journal of Iberian Geology 32 (2): 175196. Arnedo, D., Alonso, E.E., Olivella, S. & Romero, E. 2008. Gas injection tests on sand/bentonite mixtures in the laboratory. Experimental results and numerical modelling. Physics and Chemistry of the Earth 33: S237S249. Delahaye, C. & Alonso, E.E. 2002. Soil heterogeneity and preferential paths for gas migration. Engineering Geology 64: 251271. Olivella, S. & Alonso, E.E. 2005. Modelling the hydromechanical behaviour of GMT in situ test including interface elements. Project Report 0505. NAGRA. Olivella, S. & Alonso, E.E. 2008. Gas flow through clay barriers. Gotechnique 58(3): 157176. Olivella, S., Gens, A. Carrera, J. & Alonso, E.E. 1996. Numerical formulation for simulator (CODE_ BRIGHT) for coupled analysis in saline media. Engineering computations 13(7): 87112. Romero, E., Alonso, E.E. & Knobelsdorf, J. 2002. GMT/ IR 01-06: Laboratory tests on compacted sandbentonite buffer material for the GMT Emplacement Project. NAGRA. Romero, E., Garca, I. & Alonso, E.E. 2003. GMT/ IR 02-02: Laboratory gas tests on compacted sandbentonite buffer material used in the GMT in-situ emplacement. NAGRA. van Genuchten, M. Th. 1980. Closed-form equation for predicting the hydraulic conductivity of unsaturated soils, Soil Sci. Soc. Am .J .44: 892898.
Figure 14.
SUMMARY
A series of dipole tests in a cylindrical sample confined volumetrically, to investigate the mechanisms of gas flow in sand-bentonite mixtures, and to examine preferential path development have been performed and modelled. The development of preferential paths along compaction interfaces and between the cell wall contact was observed after the recorded small drop in injection pressure, which was interpreted as the beginning of inflow of air into the sample. The subsequent pressure increase, before the major pressure peak, was related to the difficulty for the injected gas to find the recovery needle. Some air remained trapped within the sample-cell system at the end of the test. The 3D model includes embedded joint elements within the continuous elements to simulate crack opening and a small initial heterogeneity. Development of clearly marked preferential paths in 2D is modelled using GMT configuration. A weak spatial variability of soil properties is considered as well. Discharge peaks are a natural consequence of the formulation. The results are encouraging and may lead to more realistic computational procedures for the complex problem of gas migration through buffers.
1359
Geotechnical issues in the design and construction of PMOBs

A.R. Cabral & M. Ltourneau J.S. McCartney & J. Parks E. Yanful & Q. Song
Universit de Sherbrooke, Dpartement de Gnie Civil, Sherbrooke, Qubec, Canada University of Colorado at Boulder, Boulder, USA University of Western Ontario, Department of Civil and Environmental Engineering, Canada
ABSTRACT: Passive Methane Oxidation Biocovers (PMOBs) are large-scale biofilters that reduce methane emissions from landfill sites through incorporation of a substrate that provides favorable conditions for methanotrophic bacteria. This paper addresses some key geotechnical issues that must be considered in the development of rational designs for PMOBs. In particular, the geotechnical properties of the unsaturated cover soils that should be selected to lead to liquid and gas flow conditions which are optimal for methane oxidation are discussed. Modeling to assess the impacts of water infiltration on preferential gas flow through PMOBs on landfill slopes is presented. The construction process for a field-scale PMOB installed at the Saint-Nicphore landfill (Quebec, Canada) is discussed, with focus on how the geotechnical issues were addressed. 1 1.1 INTRODUCTION Methane emissions from landfills several geotechnical considerations, which are the focus of this paper. 1.2 Passive methane oxidation biocovers
Fugitive methane (CH4) emissions from landfills account for approximately 24% of the anthropogenic emissions in the US (IPCC 2007), as much as 30% in Europe (Huber-Humer et al. 2008) and 19% in Canada (Environment Canada 2008). Several studies have identified that CH4 oxidation in final covers is a technically viable and economical solution to reduce fugitive emissions, without the need for added maintenance or infrastructure (Humer and Lecher 1999; Chanton and Liptay 2000; Gebert et al. 2003 Gebert and Grngrft 2006; Dever et al. 2007; Jugnia et al. 2008a; Jugnia et al. 2008b; Rannaud et al. 2009). Landfill emissions are difficult to control once biogas collection is discontinued, which usually happens 20 to 30 years after placement of the final cover. Only a long term solution, such as a passive CH4 oxidation biocover (PMOB) can prevent these fugitive emissions from reaching the atmosphere. The effectiveness of a PMOB, both immediately after installation (i.e., when it helps to reduce the portion of the total emissions not captured by active collection systems) or in the long term (when the collection system is shut down), depends on proper design. PMOB design has
PMOBs are large-scale biofilters that can be used with regulatory-accepted hydraulic barriers for landfills. Their goal is to reduce CH4 emissions. The primary component of a PMOB is a substrate that provides favorable conditions for methanotrophic bacteria, which oxidize CH4 gas into carbon dioxide (CO2) and water. PMOBs include a Gas Distribution Layer (GDL) that intercepts gas flow from the waste or escaping through any cracks or fissures in an underlying hydraulic barrier (Abu Hejleh and Znidarcic 1995; Beven and Germann 1982). The substrate layer may also be integrated with a vegetation layer to provide additional nutrients to the substrate and enhance aeration of the substrate through the roots. A schematic of a PMOB system integrated into a regulatory-accepted cover system is shown in Figure 1. The attractiveness of this design is that it uses the same layers prescribed by government regulations for landfill covers to achieve the additional goal of CH4 oxidation. The main design goal of PMOBs is to oxidize 100% of the mass of CH4 that reaches the base of the substrate layer. The PMOB should be designed to maintain this efficiency in the long-term, even
1361
2.2
Impact of flow processes in layered soils
Figure 1.
Typical PMOB components.
when there are changes in atmospheric conditions and CH4 generation patterns. 2 2.1 PMOB DESIGN ISSUES Cover soil degree of saturation
It also turns out that most natural soils contain sufficient methanotrophic bacteria to oxidize CH4 emissions (Scheutz et al. 2009). However, not all soils are adequate be considered as a substrate layer. In order to function properly, the substrate soil layer in a PMOB should ideally have a degree of saturation, Sr, which is within a range that is optimal both for bacterial growth and effective gas flow. Effective gas flow means that inbound oxygen (O2) and outbound biogas must be allowed to flow at rates that provide sufficient residence time for oxidation by methanotrophic activity. As the PMOB component is near the ground surface, it is unlikely that Sr will be constant with depth and will fluctuate with time on an hourly, daily and seasonal basis. If the Sr is too low, then gas flow will be too fast through the substrate layer, and the efficiency of CH4 oxidation may decrease. Conversely, if the Sr is too high, then O2 will not be able to enter the substrate, and CH4 will not be able to escape effectively through the PMOB. Biogas will flow through the most permeable path, such as cracks or along pipes and wells. Rannaud et al. (2009) presented a preliminary discussion about the importance of hydraulic aspects, in particular the evolution of the Sr, on the behavior of PMOBs. The influence of the Sr on O2 penetration into the substrate materiala key issue as far as oxidation efficiency is concernedis discussed by Rannaud et al. (2009). Air can become occluded for Sr values greater than approximately 85% (Nagaraj 2006), causing a drastic reduction in gas (O2) flow (Aachib et al. 2004; Cabral et al. 2004). If such occlusion occurs, biogas will tend to migrate to other areas of the cover. If emissions concentrate in a particular spot in the PMOB (referred to as a hot spot), the overall CH4 oxidation efficiency of the PMOB can be negatively affected.
The features of the GDL and substrate layers that make them suitable to function as components of PMOBs may result in flow processes at their interface that may alter the performance of the PMOB. Specifically, the contrast in hydraulic properties between the substrate and the coarser GDL soils will form a capillary barrier. The substrate is, in this case, the Moisture Retaining Layer (MRL) and the GDL acts as Capillary Break Layer (CBL). The capillary break may alter the Sr along the interface between the substrate and gas distribution layer, as shown in the results from a column infiltration study with a CBL at 0 m in Figure 2 (McCartney et al. 2007). Proper design must consider the contrast in hydraulic properties for the materials. As nearly all landfill covers are constructed on a slope, the slope of the PMOB system will lead to a non-uniform distribution in water content along the substrate-GDL interface along the length of the slope. Specifically, lateral diversion of water in the MRL will lead to increased water contents near the toe of the slope (Parent and Cabral 2005; 2006). If air occlusion occurs downslope within the MRL, the biogas will have the tendency to flow upslope within the GDL then migrate upwards through zones of the substrate where the water content is lower. These hydraulic and geometric features of PMOBs will certainly lead to zones where the substrate is submitted to higher CH4 loadings, potentially decreasing the oxidation efficiency of CH4 in the final cover. Minimization of non-uniform distributions in water content along the length of the slope is paramount to the long term effectiveness of a PMOB. Alternatively, the zones expected to receive greater gas fluxes must be designed accordingly. Material selection and characterization is the first step in the design of a capillary barrier. In the case of a
0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0
0 hrs 50 hrs 176 hrs 275 hrs 375 hrs 475 hrs 550 hrs 775 hrs 950 hrs 1200 hrs 10 20 30 40 Volumetric moisture content, % 50
Figure 2. Impact of capillary break on water content profiles in a MRL during infiltration (McCartney et al. 2007).
1362
Elevation from base, m
PMOB, the steps proposed by Parent and Cabral (2005) can be followed, but with a different aim: to build a system that minimizes air occlusion inside the substrate. 3 PROPOSED PMOB LAYOUT
regulatory-prescribed landfill. For simplicity of construction, the GDL was selected to be 0.15-m thick; the substrate was selected to be 0.3-m thick and the vegetation support layer was selected to be 0.15-m thick. 4 COVER MATERIALS
A practical PMOB layout was developed as part of this study, whereby the PMOB does not constitute an additional layer but rather is incorporated into the protection layer that normally is placed on top of the impermeable hydraulic barrier layer in regulatory-accepted cover systems. In this layout, the goal of the PMOB is to recuperate whatever fugitive emissions are not captured by the landfill gas collection system, and that eventually migrates through cracks or fissures within the impermeable layer (see Fig. 3). The fugitive emissions that pass through the hydraulic barrier and reach the GDL are redistributed more uniformly at the base of a substrate layer, thereby avoiding concentrated CH4 loadings onto the substrate. The vegetation support layer can be incorporated (mixed) with the substrate layer or be constructed as an independent layer, depending on the characteristics of the materials available at a particular landfill. As part of the research teams evaluation into the impact of geotechnical issues hydraulic issues affecting the oxidation efficiency of PMOB systems, a test PMOB was designed and constructed at the Saint-Nicphore landfill in Quebec, Canada. Specifically, a PMOB system was designed to evaluate the water and gas flow processes involved in PMOBs in order to provide validation data for numerical modeling efforts. The test PMOB system was designed following two primary constraints imposed by an industrial partner: (i) The cover system had to fulfill the requirements of local Quebec landfill regulations, which requires, from bottom up: 0.30 m of sand, 0.45 m of low permeability material, a 0.45 m protection layer and a 0.15 m for vegetation growth, and (ii) Materials available on site had to be used. The PMOB was designed to be located in the top 0.6 m of the
Coarse recycled aggregate (0.15 m) Sampling profile Substrate (0.3 m) Geotextile (separation)
Ideally, the substrate layer should be composed of a material (or mixture of materials) that maintains a very high air-filled porosity (i.e., low Sr). This allows deeper penetration of molecular O2, a key element to attain high CH4 oxidation efficiency (Cabral et al. 2009). A fine sand found in great quantities within the property at Saint- Nicphore was selected. Intuitively, such soils should be avoided, because they retain water when unsaturated. However, as mentioned above, one important constraint is that locally available materials had to be used. For the vegetation support layer, top soil also available was selected. The topsoil was found to have essentially the same granulametric curve as the substrate sand, but with higher organic content. Further, the vegetation layer should be relatively loose to permit free growth of plants. A locally available 510 mm gravel in Quebec that was previously characterized by Cabral et al. (2007) was used in this study as the GDL. The properties of the substrate sand that are relevant to PMOB performance are shown in Table 1. Gas flow column studies were performed on substrate sand and substrate sand mixed with compost. The addition of compost led to an increase in organic matter. During column flow testing, the pH of the substrate mixed with compost increased, which indicates greater amounts of oxidation. Hanging column tests were performed on the vegetation layer and substrate sand using the approach described by McCartney et al. (2008). The Soil-Water Retention Curve (SWRC) data for these soils is shown in Figure 5(a). The topsoil was placed in loose conditions (porosity of 0.58) to promote an amenable density for plant growth, while
Table 1.
Material properties. Organic matter (%) pH (before oxidation test) 6.68 6.21 pH (after oxidation test) 6.66 6.78
Soil
Cracks (preferential path for biogas) Drainage to low point
Gs 2.72 2.29
Figure 3. Schematic of PMOB components with preferential gas flow through cracks and fissures in the hydraulic barrier.
Substrate 1.14 sand Subst 21.38 mixed with compost
1363
100 90 80 70 60 50 40 30 20 10 0 100
Substrate Sand Vegetation layer topsoil 10 1 0.1 0.01 Particle size (mm) 0.001
model was fitted to the SWRC data (van Genuchten 1980), and the van Genuchten-Mualem model was used to predict the Hydraulic Conductivity Functions (HCFs) shown in Figure 5(b). Comparison of the predicted HCFs indicates that the gravel will be less conductive than the substrate sand for suctions greater than 0.04 kPa, because the gravel drains rapidly to dry conditions. This implies that the substrate layer will certainly be affected by the capillary break effect. 5 MODELING OF WATER FLOW IN A PMOB
Figure 4. Grain size distribution of topsoil and substrate sand.
Percent finer than (%)
60 50 40 30 20 10
Substrate sand (fitted) Substrate sand (data) Vegetation layer (fitted) Vegetation layer (data) GDL gravel (Cabral et al. 2007)
0 1.E+00 1.E+01 1.E+02 (a) 1.E-03 1.E-02 1.E-01 Suction (kPa) 1.E+00
1.E-02 1.E-04 1.E-06 1.E-08 1.E-10
Substrate sand GDL sand Vegetation layer
1.E-12 1.E+00 1.E+01 1.E+02 (b) 1.E-03 1.E-02 1.E-01 Suction (kPa)
Figure 5. (b) HCF. Hydraulic properties of soils: (a) SWRC;
the substrate sand was lightly compacted (porosity of 0.45). Despite the difference in porosities of these soils, the SWRCs for the topsoil and substrate soils have the same air entry suction and slope. The properties of the GDL from Cabral et al. (2007) are also shown in this figure. The van Genuchten
VADOSE/W, a commercial finite element program (Geo-slope 2004), was employed to model water and heat flow through the PMOB in response to atmospheric boundary conditions. Krahn (2004), Adu-Wusu et al. (2007), and Song and Yanful (2008) provide detailed description and required input parameters for VADOSE/W. Coupled heat transport was considered in simulations to evaluate the impact of vapour transport, a flow process that may have a significant impact on PMOB performance prediction. This process may be important to consider for suction values greater than those measured for the hydraulic properties in Figure 5, which will occur in the case of evaporation. The predicted water balance components in the PMOB are shown in Figure 6(a) using weather data from the area of the Saint-Nicphore landfill for the period of September 2006 to August 2007. The rainy seasons in the modeling period at the site were concentrated between September and November 2006, and from May to June 2007. Figure 6(a) indicates that surface evaporation caused most portion of water loss from the cover, especially in the period of May to August 2007 due to high temperature and low relative humidity (data not presented). Drainage from the bottom of GDL accounted for 32% of total precipitation and mostly occurred before the end of 2006. The distribution in degree of saturation along the interface between the MRL (Sand 1) and GDL (Gravel) for different times during the simulation period is shown in Figure 6(b). This figure shows a drying process after a period of extreme rainfall that saturated the cover. The substrate was essentially saturated for a period of 3 months during the winter, but dried to a degree of saturation ranging from 0.3 to 1. The interface at the toe of the slope was found to remain fully saturated, while the crest of the slope experienced the greatest changes in degree of saturation due to weather fluctuations.
1364
50
Cumulative value (mm)
40 30 20 10 0 -10 -20 -30 -40
Precipitation Runoff Soil water storage Water balance closure Evapotranspiration Drainage
irrigation during such dry periods (as evidenced by field measurements of water content). During March to April 2007, the simulated Sr along the interface between MRL and GDL ranged from 0.4 to 0.8 for most parts of the PMOB. These water content conditions in the substrate layer may provide relatively better environment for the activity of methanotrophic bacteria compared to other months in the modeling period. However, the distribution in Sr along the slope during these periods will certainly lead to higher gas fluxes at the crest.
Date
(a) 1.0
0.9 0.8 0.7 0.6 0.5 0.4 0.3 0.2 0.1 0.0 0 Degree of saturation 11/15/06 12/15/06 1/15/07 2/15/07 3/15/07 4/15/07 5/15/07
(b)
10 20 30 40 50 Horizontal distance from toe of slope (m)
Figure 6. (a) Water balance components based on the configuration of PMOB constructed at the site and climatic data of Sept. 2006 to Aug. 2007; (b) Degree of saturation along the interface between the substrate layer GDL.
The modeled Sr at the interface between MRL and GDL [Fig. 6(a)] shows the following features: From November 2006 to February 2007 the saturation in the substrate sand was over 0.8, which may not be beneficial to gas migration. The high saturation in the substrate during this period could be attributed to large precipitation in the previous months (Sept. to November 2006) and low evaporation in this stage. After May 2007 to the end of the simulation, the modeled saturation values at the interface between MRL and GDL were less than 0.2, which may be attributed to large surface evaporation and transpiration of vegetation as well as small precipitation in this period. The low Sr in the substrate layer may lead to significant gas flow. However, water content at this low saturation may not be optimal for the activity of methanotrophic bacteria. To achieve maximum efficiency for methanotrophic bacteria in this stage, the water content in the substrate layer may have to be adjusted through proper
The modeling results show that evaporation played an important role in the prediction of water content in the substrate layer. Estimation of evaporation in VADOSE/W was on the basis of input data including air temperature, relative humidity, wind speed, and radiation energy. In the current study, the net radiation was estimated by the program itself based on the latitude of the site. If measured net radiation at the site were available, the accuracy of the modeled evaporation would have been greatly enhanced (Zornberg and McCartney 2005). Simulation results may be significantly impacted by the predicted HCFs in Figure 5(b), as errors in the HCF predicted using the van Genuchten-Mualem model have been observed to be greater than 4 orders of magnitude (McCartney and Parks 2009). The test PMOB will also provide opportunity to calibrate input parameters in the modeling. Abdolhazdeh et al. (2010) evaluated the accuracy of PMOB performance predictions made using laboratory-obtained SWRCs and found that more accurate predictions could be obtained using the SWRC and HCF back-calculated from field measurements.
CONSTRUCTION OF TEST PMOB SYSTEM
The test PMOB was constructed on top of an existing silty cover layer currently serving as a low permeability layer, similar to the layout shown in Figure 3. The test PMOB was constructed adjacent to a control cell with dimensions that were the same as a regulatory accepted landfill cover, where a silty soil serves as protection layer, which is capped by 0.15 m of top soil. To evaluate the reliability of the modeling results which indicated successively worse Sr conditions for PMOB performance along the length of the slope, the test PMOB was constructed as a long, thin strip. Because the degree of saturation results indicated favorable performance for the local site weather, the same layer dimensions as those used in the modeling effort in the
1365
previous section were used in the design. Although 510 mm gravel was used for the GDL, any coarse material (or industrial residue) could have been adopted; in fact, when the time comes to construct the actual final cover, any available coarse material will be considered. A geotextile was used to prevent contamination of the gravel layer with fines from the substrate sand. Cracks in the underlying hydraulic barrier were simulated by inserting 10-mm stainless steel tubes across the existing silty clay interim cover, until the bottom reached the waste mass. Three such tubes were inserted every 10-m along the main axis of the 50-m long plot). These cracks were simulated also in the control cell. By creating these simulated cracks, emissions from the two plots (PMOB and control cell) can be compared in a more straightforward manner. Emissions monitoring will be performed using flux chambers. To prevent biogas from entering the PMOB, a P-trap was installed in the drain located at the toe. The cells were constructed in the fall of 2009. The first step was to confirm the thickness of the existing interim cover [Fig. 7(a)] and adjust the thickness in areas where this was needed [Fig. 7(b)]. Figure 7(c) shows the excavation of the silty clay in the region where the PMOB will be constructed. The final result is depicted in the photograph in Figure 7(d). Sand cone tests were then performed to obtain the in situ unit weight of the silty clay.
Figure 9.
Placement of substrate sand over geotextile.
The gas distribution layer placed atop the waste is shown in Figure 8(a), while the separation geotextile is depicted in Figure 8(b). After placement of the substrate layer [Fig. 9(a)] samples were taken and brought to the laboratory to perform column oxidation tests. The final PMOB, with the top soil already in place is shown in Figure 9(b). A single temperature and water content profile was installed to evaluate the hydraulic response of the PMOB during the winter and spring. Additional instrumentation will be added in the spring of 2009. 7 CONCLUSIONS
The results of this study indicate that short PMOBs may be more suitable than full-coverage biocovers due to non-uniform distributions in the Sr along the slope over time. A proposed PMOB and field testing program have been implemented to evaluate the role of the Sr and slope length. ACKNOWLEDGEMENTS Funding for this study was provided by Cascades Inc. and NSERC (Canada) under the UniversityIndustry Partnership grant number CRD 379885-08. The views are those of the authors alone. REFERENCES
Aachib, M., Mbonimpa, M. & Aubertin, M. (2004). Measurement and prediction of the oxygen diffusion coefficient in unsaturated media, with applications to soil covers. Water, Air, and Soil Pollution. 156(1): 163193. Abdolahzadeh, A.M., Cabral, A.R., Lafond, J. & Allaire, S. (2010). Hydraulic aspects of the design of a passive methane oxidation biocover. GeoFlorida 2010. Accepted. Abu-Hejleh, A.N. & Znidarcic, D. (1995). Desiccation theory for soft cohesive soils. Journal of Geotechnical Engineering, ASCE, New-York. 121(6): 493502. Adu-Wusu, C., Yanful, E.K., Lanteigne, L. & OKane, M. (2007). Prediction of the water balance of two soil cover systems. Geotechnical and Geological Eng. 25: 215237.
Figure 7.
Preparation of experimental PMOB.
Figure 8. Placement of GDL and of separation geotextile.
1366
Beven, K. & Germann, P. (1982). Macropores and water flow in soils. Water Resources Res. 18(5): 13111325. Cabral, A.R., Moreira, J.F. & Jugnia, L.B. (2009). Biocover performance of landfill methane oxidation: Experimental results. J. Environ. Eng. ASCE: Tentatively Accepted. Cabral, A.R., Tremblay, P. & Lefebvre, G. (2004). Determination of the diffusion coefficient of oxygen for a cover system composed of pulp and paper residues. Geotechnical Testing J. 27(2): 184197. Cabral, A.R., Parent, S.-. & El Ghabi, B. (2007). Hydraulic aspects of the design of a passive methane oxidation barrier. 2nd BOKU Waste Conf., Vienna. 223230. Chanton, J.P. & Liptay, K. (2000). Seasonal variation in methane oxidation in a landfill cover soil as determined by an in-situ stable isotope technique. Global Biogeochem. Cycles. 14: 5160. Dever, S.A., Swarbrick, G.E. & Stuetz, R.M. (2007). Passive drainage and biofiltration of landfill gas: Australian field trial. Waste Management. 27(2): 277286. Environment Canada (2008). National Inventory report: Greenhouse gas sources and sinks in Canada, 1990 2006. E. Canada. Ottawa, Environment Canada. Gebert, J., Grngrft, A. & Miehlich, G. (2003). Kinetics of microbial landfill methane oxidation in biofilters. Waste Management. 23(7): 609619. Gebert, J. & A. Grngrft (2006). Performance of a passively vented field-scale biofilter for the microbial oxidation of landfill methane. Waste Management. 26(4): 399407. Geo-slope (2004). VADOSE/W 2004 users guide. Geo-slope International Limited, Calgary, Alberta, Canada. Huber-Humer, M., Gebert, J. & Hilger, H. (2008). Biotic systems to mitigate landfill methane emissions. Waste Management Research. 26(1): 3346. Humer, M.H. & P. Lechner (1999). Alternative approach to the elimination of greenhouse gases from old landfills. Waste Management Research. (17): 443452. IPCC (2007). Climate change 2007: Mitigation and Control. Working Group III to the 4th Assess Report of the IPCC. B. Metz and al. Cambridge, UK, NY, USA. Jugnia, L.B., Cabral, A.R. & Greer, C.W. (2008a). Biotic methane oxidation within an instrumented experimental landfill cover. Ecological Engineering. 33(2): 102109.
Jugnia, L.B., At-Benichou, S., Fortin, N., Cabral, A.R. & Greer, C.W. (2008b). Activity, dynamics and diversity of methanotrophs within an experimental landfill cover. Soil Science of America Journal. (accepted). Krahn, J. (2004). Vadose zone modeling with VADOSE/W: An engineering methodology. Geo-slope Int. Calgary. McCartney, J.S. & Parks, J. (2009). Uncertainty in predicted hydraulic conductivity functions. 17th ICSMGE. Alexandria, Egypt. McCartney, J.S., Villar, L. & Zornberg, J.G. (2007). Estimation of the hydraulic conductivity function of an unsaturated clay using an infiltration column test. Proc. 6th Brazilian Conf. on Unsat. Soils (NSAT). Salvador. Nov. 13. McCartney, J.S., Villar, L. & Zornberg, J.G. (2008). Nonwoven geotextiles as hydraulic barriers for capillary rise. GeoAmericas. Cancun, Mexico. March 35. Nagaraj, T.S., Lutenegger, A.J., Pandian, N.S. & Manoj, M. (2006). Rapid estimation of compaction parameters for field control. Geotechnical Testing Journal. 29(6). Parent, S.-. & Cabral, A.R. (2005). Material selection for the design of inclined covers with capillary barrier effect. Geo-Frontiers 2005. Austin, TX. Parent, S.. & A.R. Cabral (2006). Design of inclined covers with capillary barrier effect. Geotechnical and Geological Engineering Journal. 24: 689710. Rannaud, D., Cabral, A. & Allaire, S. (2009). Modeling methane migration and oxidation in landfill cover materials with TOUGH2-LGM. WASP. 198(1): 253267. Scheutz, C., Kjeldsen, P., Bogner, J.E., De Visscher, A., Gebert, J., Hilger, H.A., Huber-Humer, M. & pokas, K. (2009). Microbial methane oxidation processes and technologies for mitigation of landfill gas emissions. Waste Management and Research 27(5): 409455. Song, Q. & Yanful, E.K. (2008). Monitoring and modeling of sand-bentonite cover for ARD mitigation. WASP. 190: 6585. van Genuchten, M. (1980). A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. SSSAJ. 44: 892898. Zornberg, J.G. & McCartney, J.S. (2005). Evaluation of evapotranspiration from alternative landfill covers at the Rocky Mountain Arsenal. Experus 2005, 2729 June.
1367
Coupled solute mass transport and evaporation in saline soil

A.S. Dunmola & P.H. Simms
Department of Civil and Environmental Engineering, Carleton University, ON, Canada
ABSTRACT: The presence of salts is detrimental to plant growth in soils and also impedes consolidation in disposed mine tailings. The current study conducted small-scale column drying tests on clean soil prepared to homogenous initial pore-water concentrations using a model wax column technique. This study characterized the coupling of 1D solute mass transport in unsaturated soils to surface evaporation. Rapid salt accumulation in the top 1 cm of treated soil was observed. This coincided with the profile gravimetric water content for treated soil being higher compared to control soil. However, throughout the experiment, the relative evaporation for treated soil was only a fraction of values for control soil. This is attributed to the surface salt crust acting as a capillary barrier to the upward flux of water through soil to the atmosphere. This study demonstrates the significance of salt crust formation in predicting evaporation from saline soils and mine tailings disposal facilities. 1 INTRODUCTION opposed to simulated radiant energy. Only wind was simulated and constituted the main driver for evaporation. The accumulation of salts at the soil surface may be affected by the 1D mass transport of solute dissolved in pore-water. The mass transport of dissolved solutes in unsaturated soil and mine tailings (where advection is insignificant due to very low pore-water velocity) is known to occur mainly by diffusion (Shimojima et al. 1996; West et al. 2003). Characterizing the profile pattern of this solute mass transport is important for the accurate prediction of the rate of evaporation from saline soils and mine tailings. The current study aims at characterizing this 1D mass transport and how it couples to evaporation rate, using clean soil. How this 1D mass transport affects surface salt accumulation and evaporation from soil is investigated in this paper. For this study, a model wax column technique, commonly used in ionic (fertilizer) mobility studies in soil was employed for studying 1D mass transport. Some practical implications of observations from this study for agriculture and mine waste disposal is advanced.
The inter-boundary flux of water from soil into the atmosphere through the process of evaporation has both agricultural and geotechnical implications (Fujimaki et al. 2006; Wilson et al. 1991). In saline soils, evaporation leads to the accumulation of salts at the surface, which is detrimental to crop growth. On the other hand, salt build-up at the soil surface lowers the rate of evaporation, with as high as 90% reduction in potential evaporation being reported for soil (Chen, 1992). Also, evaporation has geotechnical implications for mine waste management. The ability of an old layer of surface-disposed thickened mine tailings to support fresh layer is affected by the rate of strength gain by former as a result of evaporative losses. The presence of salts in surface-disposed thickened mine tailings slows the rates of evaporation and consolidation of such facilities. In fact, field observations have shown that salt can completely shut down evaporation from thickened tailings deposits after 3 weeks of deposition (Simms et al. 2007). Salinity is known to reduce the rate of evaporation from soils through three mechanisms: effect of albedo, depression of vapor gradient and physical obstruction of salt crust to flow of water (Fujiyasu and Fahey, 2000). The effect of albedo in suppressing evaporation in thickened mine tailings has been demonstrated to be minimal on a field scale, indicating one or both of the other two factors may be relatively more significant (Simms et al. 2007). Hence, for this paper, the effect of albedo on evaporation was excluded by conducting all the column experiments under ambient laboratory light as
2 2.1
EXPERIMENTAL METHODOLOGY Preparation and sampling of soil columns for drying test using petroleum-jelly wax column technique
The Petroleum Jelly Wax-Column Technique (PJWCT) was introduced by Khasawneh and Solileau (1969) as a simple but effective method
1369
of preparing soil columns that can be sectioned into samples as thin as 5 mm. The technique has since been used for many ionic (fertilizer) mobility/ diffusion studies in soil (Khasawneh et al. 1974; Akinremi and Cho, 1991; Hao et al. 2000; Kumaragamage et al. 2004; Olatuyi et al. 2009). The technique involves using a molten mixture of petroleum jelly and paraffin wax (1.2 parts to 3parts by mass) to make a mould with a cylindrical cavity that can be used for packing and destructively sampling soil. The dimension of the wax columns used in the current study is shown in Figure 1. The cylindrical bore of each wax column (Figure 1) was packed with soil (artificial silt) with geotechnical properties as shown in Table 1, and prepared at about 100% degree of saturation. Prior to packing, the soil was prepared to an homogenous pore-water salt concentration using a solution of reagent-grade NaCl (Lot # 8 J9286; Purity 99.0%; BioShop Canada Inc., Burlington, ON). At the beginning of experiment, the total weight of each packed soil column was recorded. One soil column was destructively sampled to test for soil parameters at each sampling event. Destructive sampling involved sectioning soil columns
5.0 cm
into 1 cm thick slices using an adjustable hacksaw (Mastercraft Canada, Toronto ON) and a Jobmate plastic mitre box (Trileaf Distribution Canada, Toronto ON) used as a cutting guide. Minimal disturbance to soil columns and cross-contamination was ensured when cutting the soil column into sections. 2.2 Experimental conditions and analyses
Rectangular wax column (made from molten 1.2: 3 mixture of petrol. Jelly and paraffin wax) 15.0 cm Cylindrical bore for packing soil.
9.5 cm
Figure 1. Schematic of petroleum jelly-wax column used for packing soil. Table 1. Geotechnical properties of artificial silts (from Fisseha et al. 2007). Property Specific Gravity D10, D50, D60 (microns) Cu (D60 /D10) Liquid limit (%) Plastic limit (%) Saturated hydraulic conductivity (m/s) Value 2.48 1, 31, 41 41 19 13 1.7 106
The soil columns were initially prepared at 30% Gravimetric Water Content (G.W.C) using NaCl solution (concentration, C 0.20; mass of salt/mass of solution). A total of 9 soil columns were prepared, including one control soil column (prepared at 30% G.W.C with distilled water only). An additional wax column was filled with distilled water for Potential Evaporation (PE) rates determinations during the experiment. The soil columns were left to desiccate under room temperature and ambient lighting, with wind simulated by means of oscillating fan at one end of drying platform. Initial calibrations with the fan showed a PE rate of between 8 and 12 mm/day. There was no significant effect of variability in PE rates due to location of soil columns on the drying platform. The ambient temperature and Relative Humidity (RH) was monitored during the experiment using a USB502 RH/Temperature Data Logger (Measurement Computing, Norton, MA). One soil column each was destructively sampled on days 2, 3, 4, 5, 7, 9, 11 and 14, as previously described. The PE rate as well as Actual Evaporation (AE) rates from both treated and control soil columns were determined by mass difference at the beginning of each sampling event. Profile soil samples retrieved from sectioned soil column were kept in sealed Ziploc plastic bags and thoroughly mixed by hand prior to conducting Electrical Conductivity (EC), Na concentration, gravimetric and total suction analyses. EC analysis involved extracting the soil sample (1 part) with 5 parts (by mass) of de-ionized water using an orbital shaker (175 rpm for 30 minutes) followed by centrifuging at 3000 rpm (1000 X g) for 2.5 minutes. The EC of resulting supernatant was determined using a previously calibrated Traceable Conductivity Meter (VWR International, Friendswood, TX). The Na concentration of the supernatant was inferred from the EC data by means of calibration curve previously obtained using standard NaCl solutions (Fisher Scientific, Ottawa ON). Also, the G.W.C of profile samples were determined by difference in mass after placement in oven at 105C for 24 hours. In addition, profile total suction for each sampled soil column was obtained using a WP4-T Dewpoint PotentiaMeter (Decagon Devices Inc., Pullman, WA).
1370
2.3
Measurement and prediction of relative evaporation from desiccating soil columns
The relative evaporation, RE (ratio of AE/PE) from each sampled soil column was calculated using measured AE and PE rates. The RE is known to be a function of the total suction at the soil surface as given in Wilson et al. (1997): W vV w RT ha AE e 1 ha PE
(1)
0.8 0.6 0.4 0.2 0.0 0 2 4 6 8 10 12
8 6 4 2 0 14
Grad [w ] w p w s y p ys
(2)
Actual Evaporation (mm/day)
where w G.W.C (%); wp G.W.C at a depth within the soil; ws G.W.C at the soil surface; yp depth from soil surface (cm); ys depth at the soil surface (0 cm). Grad[w] gives a measure of the potential for a point at depth in soil to supply water needed for an upward flux required to satisfy the evaporative demand at the soil surface. Grad[w] was therefore calculated and compared for both treated and control soil column at the end of the experiment (day 14). 3 3.1 RESULTS Measured evaporation rates/fronts from saline and control soil columns
Time (Days)
RE (Treated) RE (Control) PE
Figure 2. Potential evaporation rates during experiment and Relative evaporation for treated and control soil columns.
4 3 2 1 0 1 2 3 4 5 6 7 8 9 10 11 12 13 14
The PE determined throughout the experiment ranged from 5 to 9 mm/day (Figure 2). The RE for both control and treated soil columns was initially high, then rapidly declined in valued and eventually became somewhat constant (Figure 2).
Time (Days)
Measured AE (TRT) Predicted AE (TRT)
Figure 3. Actual evaporation rates measured and predicted for treated soil columns during the experiment.
1371
Pot. Evaporation (mm/day)
Relative Evaporation
where total suction at surface (kPa); Wv the molecular weight of water (18.016 kg/mol): Vw the specific volume of water (1/998 m3/kg); R universal gas constant (8.314 J/mol K); T temperature of air above soil column (K) and ha relative humidity of air above soil column (%). Thus, the RE for each sampled treated soil column was also predicted from the measured total suction at the surface of soil column using equation (1). The predicted RE was then used to calculate the expected AE from corresponding treated soil column using the measured PE, and compared to the AE measured throughout the drying test. A gradient in G.W.C is defined as the change in G.W.C between 2 points in the soil per unit thickness of soil given as:
However, the RE values for the control soil column was consistently a multiple (2.3 to 5.8 times) of those for treated soil columns (Figure 2). Also, the AE predicted from surface total suction data for the treated soil columns was consistently slightly higher than the AE measured (Figure 3). This difference between measured and predicted AE rates became pronounced after day 2, following the formation of a salt crust on the soil surface. There was a sharp decline in Gravimetric Water Content (G.W.C) at all depths within the first 2 days of the drying experiment for the treated soil column (Figure 4). This was followed by a somewhat rapid decline in G.W.C for the first 3 cms of the treated soil column. For the rest of the profile, the decline in G.W.C after day 2 of the experiment was either very gradual or not significant (Figure 4). For treated soil column, there was 53.9% and 37.7% decrease in G.W.C at the top and bottom 1 cm of column, respectively, by the end of experiment. For control soil column however (data not shown), there was 90% and 83% decrease in G.W.C at top and bottom 1 cm, respectively, at the end of drying experiment.
1.0 10
soils, the G.W.C generally decreases towards the soil surface, the trend being more pronounced in the former (Figure 6). 3.2 Measured profile total suction and 1D solute transport for treated soil column
Figure 4. Profile gravimetric water contents for treated soil columns during the experiment.
Grad [w]
0.0 0 2 0.4 0.8 1.2 1.6 2.0
4 6 8 10 12 14
Treated Soil Control soil
Figure 5. Profile gradients in gravimetric water contents for treated and control soil columns at the end of experiment.
0 0 2 4 8 12 16 20 24
4 6 8 10 12 14
Treated Soil Control soil
The profile total suction for treated soil columns throughout the experiment is shown in Figure 7. The general trend observed for the treated columns was very high suctions at the top 2 centimeters of soil, and a gradual decrease in total suction with depth. The total suction values reached the as-deposited values at 4, 6, 7 and 10 cm depths on days 2, 4, 7 and 11, respectively. In contrast, for the control soil, a high total suction was observed at the end of the experiment for the top 1 cm, but the total suction reduced to almost the as-deposited values with depth (Figure 7). The total suction for the control soil column is essentially matric, given that there was no salt supplementation, as well as considering the profile G.W.C being generally low at the end of experiment (Figure 6). For treated soil, beyond day 4, total suction at the soil surface (1 cm) increased only slightly, but below the soil surface, significant increases was observed over time, the only exception being for day 14 when total suction slightly decreased in comparison to day 11 (Figure 7). The profile salt concentration shows the highest accumulation of salt to be at the top 1 cm of soil column (Figure 8). This surface build-up of salt continued until day 9, after which a slight decline in Na concentration was observed (Figure 8). Also, a slight accumulation of salt was observed at 2 cm depth up until after day 3, when slight decrease was observed. The surface salt build-up generally coincides with slight decrease in salt concentrations below the as-deposited values at remaining depths within the soil column (Figure 8). A mass balance of the salt (data not presented) showed redistribution within the entire profile.
Total Suction (kPa)
0 0 2 4 6 8 10 12 14
Day 0 Day 7 Day 14 CTR Day 11 Day 2 Day 14 Day 4
Also, as shown in Figure 5, at the end of experiment, the gradient in G.W.C for treated soil showed a larger range compared to the control soil. For the treated soil, Grad[w] is lower at depths and increases towards the soil surface (Figure 5). Also, at the end of experiment, the values of G.W.C were generally higher for the treated soil compared to control soil (Figure 6). For both treated and control
Figure 7. Profile total suction for: control soil column on day 14 (Day 14CTR); treated soil columns during the experiment.
1372
Depth (cm)
Figure 6. Profile gravimetric water contents for treated and control soil columns at the end of drying experiment.
Depth (cm)
Depth (cm)
5,000 10,000 15,000 20,000 25,000 30,000 35,000 40,000 45,000
120 100 80 60 40 20 0 2 4 6 8 10 12 14
Day0 1cm 2cm 3cm 6cm 8cm 10cm
Time(Days)
Figure 8. Profile Na concentration for treated soil columns during the drying experiment.
4 4.1
DISCUSSION Effects of pore-water salinity on measured evaporation rates/fronts from soil columns
The relative evaporation from both treated and control soil columns follow the general trend for unsaturated soils (Wilson et al. 1994). Throughout the experiment, the RE for the treated soil was only a fraction of the value for the control soil (Figure 2). This is because the rate of evaporation from soil is known to be controlled by the gradient in vapor pressure (or R.H) between soil and the overlying atmosphere (Wilson et al. 1997), which is also a function of total suction at the soil surface (Edlefsen and Anderson, 1943). This significant reduction in the rate of evaporation due to porewater salinity has been reported for soil (Chen, 1992; Fujimaki et al. 2006). There was a more uniform drying with profile for the control soil column, but a wider range of G.W.C with depth for the treated soil column. The top 3 cm of the treated soil showed more rapid decrease in water content compared to rest of the profile. Though the gradients in G.W.C were still high at the end of the experiment for treated soil, its RE was low compared to control soil. This is a result of the formation of salt crust at the soil surface on day 2 acting as a capillary barrier to the flux of water through the soil column into the atmosphere. The effect of this salt crust formation was also observed with the AE predicted from surface total suction data for treated soil column being slightly higher than the corresponding measured AE, especially so after the formation of crust by day 2. 4.2 1D solute mass transport within soil and effect on evaporation rates
a net upward 1D mass transport of salt within the soil column, coinciding with the G.W.C at the top 1 cm been the lowest throughout the experiment. There seems to be a preferential transport of dissolved solute as opposed to upward flux of water as depicted by a higher G.W.C gradient at the end of the drying experiment. This comparably lower upward water flux may be attributed to the formation of salt crust at the soil surface by day 2 that lowered evaporation rate from the treated soil column. Osmotic suction from the dissolved solute constituted a significant contributor to the profile evolution of total suction over time, as depicted by relatively low total suction determined for the control soil at end of experiment (Figure 7). This significantly higher profile total suction further explains the RE from treated soil being only a fraction of the value for the control soil column. Even though the total suction of treated soil is higher than for control soil, the matric suction of former is expected to be lower, as evident from its higher profile G.W.C at the end of experiment. This lower matric suction evolution for the treated soil will negatively impact on its consolidation property, resulting in lower shear strength and bearing capacity when compared to control soil (Rasaam and Williams, 1999). Therefore pore-water salinity lowers the rate of evaporation from soil and has negative geotechnical implications for evolution of profile shear strength in engineered structures. 4.3 Practical implications of coupled solute mass and unsaturated water transport in saline soil The coupled transport of solute and water observed in the current study has both agricultural and geotechnical/geo-environmental implications. In arid/semi-arid climates with very high evaporative demand, the co-transport of solutes and water within the soil results in the surface accumulation of salts. This surface-accumulated salt could result in the retardation of leaf and root growth, as well as overall reduction in crop yield. The preferential 1D transport of solute as observed in current study may partly explain the reduced water availability at the crop rooting zone as reported in Katerji et al. (2003). Therefore, modeling crop response to water stress should consider the reduction in rootingdepth water availability resulting from pore-water salinity such as implemented in CERES-Maize model (Castrignano et al. 1998). The accumulation of salts at the soil surface also has geotechnical and geo-environmental implications. Surface deposition planning of dewatered mine tailings is based on ability to predict when sufficient desiccation and consolidation of an old layer would permit the deposition of a fresh layer.
The general trend observed from study was an accumulation of salt at the top 1 cm of the soil column. This surface accumulation resulted from
Na+ Conc.(partsperthousand)
1373
For saline mine tailings stack, the profile evolution of shear strength resulting from desiccation may be complicated by the preferential transport of salts to the surface that keeps the deposited layer relatively wet at depths while dry at the surface few centimeters. In addition, the reduced water availability at the top few centimeters of a surfacedeposited sulphidic mine tailings may restrict sulphide oxidation to the top of the layer. Hence, the potential for acid mine drainage for saline dewatered mine tailings deposit may be lower compared to non-saline tailings. Therefore, the challenge for deposition planners is optimizing operations in a way that geotechnical benefit of increased shear strength due to desiccation is balanced out with the need to restrict oxygen diffusion into saline tailings deposits. This is especially true for arid and semiarid climates where management of hypersaline mine tailings deposits may be a challenge. 5 CONCLUSION
The current study demonstrated that pore-water salinity could shut down the relative evaporation from soil to just a fraction of the value from nonsaline equivalent soils. This results from the accumulation of salts at the top 1centimeter of the soil that forms a barrier to evaporation. The agricultural implication is reduced water availability at the crops rooting depth that decreases yield. This also has negative geotechnical consequence of slowing down the profile evolution of shear strength in saline tailing deposits. The salt accumulation may however be beneficial by restricting oxygen diffusion to top few centimeters of deposited layer, thus reducing the potential for acid mine drainage. Therefore, modeling crop response to water deficit should consider the preferential transport of salts to the crops rooting depth. In addition, saline mine waste deposition planning should have the objective of optimizing desiccation for strength gain without jeopardizing the geo-environmental benefit of reduced oxygen diffusion afforded by salt crust formation. ACKNOWLEDGEMENT Scholarship funding from the Natural Sciences and Engineering Research Council of Canada (NSERC) to first author is gratefully acknowledged. REFERENCES
Akinremi, O.O. and Cho, C.M. 1991. Phosphate transport in Ca-saturated systems: II. Experimental results in a model system. Soil Sci. Soc. Am. J. 55: 12821287.
Castrignano, A., Katerji, N., Mastrorilli, M. 1998. Modeling crop response to soil salinity: review and proposal of a new approach. Ecol. Model. 111: 107120. Chen, X.Y. 1992. Evaporation from salt-encrusted sediment surface: Field and laboratory studies. Aust. J. Soil Research. 30: 429442. Fisseha, B., Glancy, T. and Simms, P. 2007. The influence of intial degree of saturation on the shrinkage curve of silts. Proc., 60th Canadian Geotechnical Conference, Ottawa, ON. October 2124, 2007. 10101015. Fujimaki, H., Shimao, T., Inoue, M. and Nakane, K. 2006. Effects of salt crust on evaporation from a bare soil. Vadoze Zone J. 5: 12461256. Fujiyasu, Y. and Fahey, M. 2000. Experimental study of evaporation from saline tailings. J. Geotech. Geolo. Eng. 126 (10): 1827. Hao, X., Cho, L.M. and Racz, G.J. 2000. Chemical retardation of phosphate diffusion in simulated acid soil amended with lignosulfonate. Can. J. Soil Sci. 80: 289299. Katerji, N., Van Hoorn, J.W., Hamdy, A., Mastrorilli, M. 2003. Salinity effect on crop development and yield, analysis of salt tolerance according to several classification methods. Agric. Water. Managm. 62 (1): 3766. Khasawneh, F.E. and Soileau, J.M. 1969. Soil preparation and sampling techniques for studying ion movement. Soil Sci. Soc. Am. Proc. 33: 476477. Khasawneh, F.E., Sample, E.C. and Hashimoto, I. 1974. Reactions of ammonium ortho- and polyphosphate fertilizers in soil. I. Mobility of phosphorus. Soil Sci. Soc. Am. J. 38: 446451. Kumaragamage, D., Akinremi, O.O., Cho, C.M. and Goh, T.B. 2004. Phosphorus diffusion from monocalcium phosphate co-applied with salts in a calcareous soil. Can. J. Soil Sci. 84: 447458. Olatuyi, S.O., Akinremi, O.O., Flaten, D.N. and Crow, G.H. 2009. Accompanying cations and anions affect the diffusive transport of phosphate in a model calcareous soil system. Can. J. Soil Sci. 89 (2): 179188. Rasaam, D.W. and Williams, D.J. 1999. A relationship describing the shear strength of unsaturated soil. Can. Geotech. J. 36: 363368. Shimojima, E., Yoshioka, R. and Tamagawa, I. 1996. Salinization owing to evaporation from bare-soil surfaces and its influence on the evaporation. J. Hydrol. 178: 109136. Simms, P, Grabinsky, M. and Zhan, G. 2007. Modeling evaporation of paste tailings from the Bulyhanhulu mine. Can. Geotech. J. 44: 14171432. West, A.C.F., van Geel, P.J., Raven, K.G., Nguyen, T.S., Belfadhel, M.B. and Flavelle, P. 2003. Groundwater flow and solute transport in a laboratory-scale analogue of a decommissioned in-pit tailings management facility. Can. Geotech. J. 40: 326341. Wilson, G.W., Fredlund, G.W. and Barbour, S.L. 1991. The evaluation of evaporative fluxes from soil surfaces for problems in geotechnical engineering. Proceedings, Can. Geotech. Conf. pt 2 Pp 68/168/9. Wilson, G.W., Fredlund, G.W. and Barbour, S.L. 1994. Coupled soil-atmosphere modeling for soil evaporation. Can. Geotech. J. 31(2): 151161. Wilson, G.W., Fredlund, G.W. and Barbour, S.L. 1997. The effect of soil suction on evaporative fluxes from soil surfaces. Can. Geotech. J. 34: 145157.
1374
Non-isothermal infiltration in plastic bentonite

B. Franois
Universit de Lige, Departement ArGEnCo, Chemin des Chevreuils, Belgium FRS-FNRS, Fonds National de la Recherche Scientifique, Bruxelles, Belgium
L. Laloui
Laboratory of Soil Mechanics (LMS), Ecole Polytechnique Fdrale de Lausanne (EPFL), Lausanne, Switzerland
ABSTRACT: In deep geological repository involving a multi-barrier system, the engineered materials, generally made of bentonite, constitute an important mechanical, thermal and hydraulic barrier to prevent the interaction between wastes and biosphere. To better understand the THM (thermo-hydro-mechanical) processes in the clayey confining system, experimental tests must be simulated by the means of advance numerical tools. A unified thermo-mechanical constitutive model for unsaturated soils, ACMEG-TS, has been developed and implemented in the finite element code LAGAMINE. Material is considered as tri-phase non-isothermal thermo-plastic medium. The results of numerical simulations of non-isothermal infiltration tests through FEBEX bentonite are compared with the available sensors measurements and interpreted in the light of elasto-thermo-plasticity of unsaturated soils. Through those simulations, the advancements in the understanding of the involved processes brought by the ACMEG-TS model are discussed. 1 INTRODUCTION and the comparison between obtained results and in-situ measurements is discussed. Finally, we interpret the obtained results in terms of mechanical irreversibilities, coupling between mechanical, thermal and water retention responses and diffusion processes.
In the coming years, safe and definitive solutions will likely be completed for managing the large quantities of high-level radioactive wastes that stem mainly from nuclear electricity production. Deep geological repositories constitute one of the most promising options to isolate such wastes from the human environment. In this context, the highly coupled thermo-hydro-mechanical (THM) phenomena that occur in the engineered and geological barriers must be captured adequately by means of numerical analysis (Laloui et al., 2008). To validate and calibrate the mathematical models, numerical simulations of in-situ experiments reproducing analogue studies have to be performed. As a preliminary test of a large scale in-situ test (FEBEX), Lloret et al. (2004) designed and carried out a non-isothermal infiltration test through the bentonite inside cylinders that aim at simulating the water that saturates the barrier in a repository. This experimental test encompasses the main THM processes that must be considered in numerical predictions of the behaviour of an engineered barrier for nuclear waste disposal. The paper presents, first, the mathematical formulation of the used constitutive and numerical tools and the calibration of the material parameters of the FEBEX bentonite. Then, the features of the performed numerical analysis are presented
THEORETICAL FORMULATION
2.1 Equilibrium and balance equations The balance equations are obtained by considering that the material is composed of solid matrix and voids filled with liquid and gas phases. The liquid phase is assumed to contain two species: liquid water and dissolved air. Similarly, the gas phase is composed of dry air and water vapour. To uniquely describe the state of the material, four primary state variables are needed: gas pressure pg, water pressure pw, temperature T, and solid displacement vector u. The solid phase component is assumed to be deformable while the water phase is slightly compressible. The solid, liquid, and water phases are assumed in thermal equilibrium, and a unique temperature is defined at each node. The mass balances are described for the components (also called species) present in the mixture rather than for the phases. Therefore, the conservation of mass of each chemical species (water
1375
and air) is assumed. The equilibrium and balance equations, as well as the water and heat flows, are expressed in the moving current configuration through a Lagrangian actualised formulation. The Lagamine finite element code including the threephase field equations described above has been used (Charlier et al., 2001). The governing equations (solid, liquid, gas and energy balance equations) are presented in Collin et al. (2002), Collin (2003) and Collin et al. (2006). 2.2 Constitutive relations
fl fg
kw grad pw gw grad y w
(6) (7)
kg
g
grad p g grad y

g g
The conservation equations which govern the THM equilibrium of the system need to be expressed in terms of the primary state variables (u, pw, pg, and T), after introduction of the constitutive relationships detailed below. The liquid water bulk density depends on the pore water pressure pw and temperature T through the water bulk modulus w and the volumetric water thermal expansion coefficient w : p pw 0 w w w 0 1 w T T0 w (1)
where kw and kg are the tensor of intrinsic water and gas permeability, respectively, g the gravity acceleration, y the vertical, upward directed coordinate, and w and g the dynamic viscosity of the water and gas. The water permeability, assumed isotropic (kw kwI, I, being the identity matrix), depends on the degree of saturation Sr according to the following equation: kw kw,satSrCKW 1 (8)
where CKWI is a material parameter. kw,sat is the saturated permeability which depends on the soil porosity n through the Kozeni-Karman relation: kw,sat kw 0,sat
1 nEXPM
n EXPN
1 n0 EXPM
n0 EXPN
(9)
where w0, pw0, and T0 are the initial values of water bulk density, pore water pressure, and temperature, respectively. The bulk density of the water vapour is determined through the following equation: pw pg Mv p M v ,0 v (2) v exp R T R T w where pg is the gas pressure, Mv is the vapour molar mass ( 0.018 kg/mol), R is the constant of perfect gases (8.3143 J/(mol.K), w is the bulk density of liquid water, T is temperature expressed in Kelvin. The saturated vapour pressure pv,0 is given by: pv,0 a exp b / T (3)
where kw0,sat is the saturated water permeability corresponding to the reference porosity n0 and EXPN and EXPM are material parameters of the Kozeny-Karman relation. The dynamic viscosity of the water and gas are assumed linearly dependent on the temperature:
g g 0 1 g ,T T T0
w w 0 1 w,T T T0
(10) (11)
where w0 and g0 are the dynamic viscosity of water and gas, respectively, at initial temperature T0. w,T and g,T are material parameters. The velocity of water vapour diffusion is related to the air bulk density gradient: (12) i v n 1 Sr Dg grad a g is the tortuosity of the material. D is the air diffusion coefficient depending on temperature and gas pressure: D 5.893 10 6 T 2.3 pg (13)
with a 112659MPa and b 5192.74K for temperature range between 273 K and 373 K. The gas phase is assumed to be an ideal gas in agreement with the Dalton law: pg pa pv Consequently, the bulk density of dry air is: pg pv M a pg M a v M a pM a a a RT RT RT Mv (4)
(5)
where Ma is the dry air molar mass ( 28.8 103 kg/ mol) and v the bulk density of water vapor. The velocity of liquid and gas phases are governed by the generalized Darcys law:
where D, T, and pg are expressed in m2/s, Kelvin, and Pascal, respectively. The heat transport is governed by conduction and convection: fT grad T c p,w w fw c p,a ia a fg c p,v i v v fg
(14)
1376
where cp,w, cp,a and cp,v are the specific heat of liquid water, dry air and water vapor. is the thermal conductivity of the mixture as deduced from the thermal conductivity of each phase: s 1 n w nSr g n 1 Sr (15)
where s, w, and a are the thermal conductivity of solid, liquid water, and gas phase, respectively. The behaviour of the solid matrix is assumed to be governed by the generalized effective stress tensor through combinations of mechanical stresses and fluid pressures (Bishop, 1961; Laloui & Nuth, 2009): pg I Sr pg pw I
(16)
In this Lagrangian approach, the Cauchy strain tensor is used: 1 (17) L LT 2 u where L is the displacement (u) gradient X defined in the global axis (X) in the moving current configuration (Charlier et al., 2001). This strain tensor is related to the generalized effective stress tensor through the mechanical constitutive model: d C : d where C is the mechanical constitutive tensor. 2.3 Mechanical constitutive model The mechanical model, called ACMEG-TS (Franois & Laloui, 2008), is based on an elastoplastic framework, the total strain increment d being decomposed into non-linear, thermo-elastic, d e, and thermo-plastic, d p, components. The elastic part of the deformation is expressed as follows: d E 1d (1 / 3)sIdT (19) (18)
Figure 1. Effect of (a) temperature and (b) suction on the shape of coupled mechanical yield limits.
where p is the mean effective stress, q the deviatoric stress and pc is the preconsolidation pressure. b, d and M are material parameters. pc depends on temperature T and suction s in addition to the p volumetric plastic strain, v (Salager et al., 2008): pc 0 exp vp 1 T log T / T0 ; if s se pc p exp p 1 log T / T T 0 c0 v if s se 1 ; log s / s s e (21)

where E is the elastic tensor and s the thermal expansion coefficient of the solid skeleton. The plastic mechanism of the material is induced by two coupled hardening processes: an isotropic and a deviatoric one. Using the concept of multi-mechanism plasticity, both mechanisms may induce volumetric plastic strain (Hujeux, 1979). The yield functions of the two mechanical, thermo-plastic mechanisms have the following expressions (Figure 1): fiso p pc riso d p fdev q Mp 1 b Log rdev 0 pc (20)
where pc0 is the initial preconsolidation pressure at ambient temperature T0 and for suction lower than the air-entry value se. is the plastic compressibility modulus and T and s are material parameters. riso and rdev are the degree of mobilization of the isotropic and the deviatoric mechanisms and are hyperbolic functions of the plastic strain induced by the isotropic and the deviatoric mechanisms, respectively (Laloui & Franois, 2009). The details on the flow rules and the consistency conditions of both plastic mechanisms are detailed in Franois and Laloui (2008). 2.4 Water retention constitutive model
As long as the soil is drying, suction increases, and the degree of saturation Sr tends to decrease mainly when the air-entry suction se is reached. Under re-wetting, a hysteretic phenomenon occurs, also represented by a yielding process (Figure 2).
1377
After dismantling, the water content, as well as the dry density, at different locations was measured. 3.2 Material characteristics
Figure 2. Schematic representation of water retention curve modelling: the hydraulic hysteresis. The amplitude of the hysteresis is governed by the shys parameter.
A wetting-drying cycle activates two successive yield limits in the (Sr s) plane (fdry and fwet, along the drying and wetting paths, respectively): fdry s sd 0; fwet sd shys s 0 (22)
where sd is the drying yield limit and shys a material parameter considering the size of the water retention hysteresis. The drying limit sd increases with the decrease of degree of saturation: sd sd 0 exp h Sr (23)
where sd 0 is the initial drying limit and h the slope of the desaturation curve in the (Sr ln s) plane (Figure 2). sd 0 is a function of temperature T and volumetric strain v (Franois & Laloui, 2008): sd sd 0 exp h Sr 1 T log T T0 e log 1 v
(24)
where T and e are material parameters describing the evolution of air-entry suction with respect to temperature and volumetric strain, respectively. 3 3.1 THE INFILTRATION TEST Presentation
The cylinders are filled with FEBEX bentonite blocks (Lloret et al., 2004). The THM properties of the FEBEX bentonite have been extensively investigated over the last decade (Villar, 1999; Lloret et al., 2003, Romero et al., 2005, among others). A series of numerical simulations of oedometric tests on thermal, hydraulic and mechanical paths have been performed to calibrate the material parameters of the ACMEG-TS model (Franois & Laloui, 2008) (Figures 3 to 6; compression being positive) leading to the determination of the mecahnical and water retention parameters reported in Table 1. The tests performed by Lloret et al. (2004) start at an initial applied stress of 0.1 MPa and a compaction suction of about 127 MPa. Numerical simulations of oedometric compression tests at different suctions and ambient temperature are compared with experimental results in Figure 3 while Figure 4 reproduces the effect of oedometric compression tests for natural hydroscopic suction (127 MPa). Figure 5 reproduces the retention curve observed on wetting at three different temperatures and compares the results with numerical simulations. Figure 6 illustrates water retention hysteresis in addition to the effect of density. The parameters governing the thermal and hydraulic diffusion in FEBEX bentonite are directly established from a literature synthesis. Villar (2002) reported the saturated hydraulic conductivity as a function of the dry density. Good agreements between Kozeni-Karman relation (Equation 9) and the evolution of permeability with dry density can be obtained for kw0,sat 41014m/s
Non-isothermal infiltration tests through the bentonite, inside cylinders 60 cm in length and 7 cm in diameter, aim at simulating the water that saturates the barrier in a repository. A temperature of 100C is imposed at the bottom of the cylinder while granitic water is injected at ambient temperature (2030C) under a pressure of 1.2 MPa over the upper lid of the cell (Lloret et al., 2004; Villar et al., 2005). The available measurements during the tests are temperatures at 50, 40, 30, 20 and 10 cm from the heaters and the volume of water intake. The duration of the tests has been 6, 12 and 24 months.
Figure 3. Numerical simulations (white points) of oedometric compressions on FEBEX bentonite at different suctions. Comparison with experiments (bold points).
1378
Table 1.
Set of the ACMEG-TS mechanical and water retention parameters for FEBEX bentonite. [MPa], [-], [C1] [-], [-], [-], [MPa] [-], [-], [-], [-], [MPa], [-], [-], [-], [-] 9.78, 0.4, 6.67 104 14.3, 16.1, 2.1, 4 1, 1.5, 1.2, 1 4, 8.64, 0.7, 10.8, 0.6
Elastic parameters Eref, v, s Isotropic plastic parameters , s, T, pc Deviatoric plastic parameters b, d, M, Water retention parameters se0, h, T, e, shys
Figure 4. Numerical simulations (lines) of oedometric compressions on FEBEX bentonite at two different temperatures under a suction of 127 MPa. Comparison with experiments (dashed lines). Figure 6. Numerical simulations of retention curves of FEBEX bentonite at two different dry densities. Comparison with experiments.
corresponding to n0 0.37 (i.e. d 1.7g/cm3) and EXPN EXPM 6.5 (Figure 7a). In addition, the water permeability depends on the degree of saturation of the bentonite. Pintado et al. (2002) determined that CKW1 2.9. The thermal diffusion of each phase has been calibrated to reproduce the experimental evolution of the thermal diffusion of the bentonite with respect to its degree of saturation (s; w; a) (0.7; 2.1; 0) [W/ mC)] (Figure 7b). The heat capacity of the solid matrix is cs 1091 J/kgC (Gens et al., 1998). Table 2 reports the material parameters of FEBEX bentonite in relation to the thermal and hydraulic diffusion processes. 3.3
Figure 5. Numerical simulations of retention curves of FEBEX bentonite at different temperatures. Comparison with experiments.
Features of analysis
The problem is treated under axisymmetric conditions around the longitudinal axis of the cylinder. The modelled domain, 3.5 cm in width
1379
Figure 7. Effect of the degree of saturation on the thermal diffusivity of FEBEX bentonite. Comparison between laboratory measurements (Gens et al., 1998) and simulation.
and 60 cm long, is meshed with isoparametric elements. Gravity is not considered (Figure 8). The temperature is imposed at 100C in the bottom of the cylinder which is assumed impervious. At the top of the cylinder, the pore water pressure is fixed to 1.2 MPa and the temperature is maintained as a constant. The thermo-hydraulic initial conditions have been imposed as follows: T0 23C; s0 127 MPa; Sr0 0.54. Several disturbing effects related to the experimental procedures needed to be considered in the numerical simulations. The construction gaps as well as the deformability of the cell have been considered in the computations by imposing a progressive horizontal displacement of the lateral cylinder wall up to 1 mm after 24 months (Figure 8) in agreement with the measured overall decrease of the dry density from an initial value of 1.65 g/ cm3 to 1.57 g/cm3 after 12-month and to 1.55 g/cm3 after 24-month tests (Lloret et al., 2004). As the cylinder is not insulated, the laboratory temperature has a non-negligible effect on the temperature distribution measured inside the cylinder. To consider this, radiation elements were introduced as thermal boundary conditions along the sample. A temperature of 23C was fixed as the laboratory temperature and the emissivity of the cylinder (Teflon) was ascribed to 0.4. The thermal flux qrad through radiation from the cylinder to the
Table 2. Material parameters of the diffusion processes in FEBEX bentonite. The parameters of the water retention curve are defined in Table 1. Thermal parameters Solid thermal conductivity Water thermal conductivity Air thermal conductivity Solid heat capacity Water heat capacity Gas heat capacity Liquid thermal expansion coefficient Solid thermal expansion coefficient Liquid dynamic viscosity thermal coefficient Hydraulic parameters Intrinsic waterpermeability Fluid dynamic viscosity Kozeni-Karman coefficient 1 Kozeni-Karman coefficient 2 Relative permeability coefficient Volumetric parameters Initial porosity Solid specific mass Water specific mass Air specific mass Liquid compressibility s w a cp,s cp,w cp,a w s w,T kw0,sat w EXPM EXPN CKW1 n0 s w a 1/w [W/(mC)] [W/(mC)] [W/(mC)] [J/(kgC)] [J/(kgC)] [J/(kgC)] [C1] [C1] [C1] [m2] [Pa s] [] [] [] [] [kg/m3] [kg/m3] [kg/m3] [Pa1] 0.7 2.1 0 1091 4200 1000 4 104 6.77 104 0.01 4 1021 0.001 6.5 6.5 2.9 0.4 2700 1000 1.18 3.33 1010
1380
Figure 8. Finite element mesh used in the simulation of the infiltration test in FEBEX bentonite. The y-axis is the axis of symmetry of revolution.
laboratory is computed according to the Kirchhoff equation:

4 4 qrad SB Tcyl Tlab
(25)
where Tcyl and Tlab are the temperatures of the cylinder and of the laboratory (Tlab 23C). SB is the Stephan-Boltzmann constant (SB 5.67 108 W/(m2K4)). Finally, the daily and seasonal temperature variation of the injected water, from 20C to 30C, has not been considered. An averaged temperature of 23C was fixed at the top surface. 3.4 Results Figure 9 compares the results of simulations with the experimental measurements in terms of temperature, water content and dry density distributions in the axis of the cylinder. First of all, the radiation elements, combined with accurate thermal diffusion and heat capacity of each phase of the bentonite, permit the reproduction of the sharp temperature decrease in the vicinity of the heater. At a location of 10 cm within the heater, simulations predict a temperature drop of 46.5C (in comparison with 49C, experimentally measured). Therefore, the thermal gradient is mainly concentrated in the bottom of the cylinder. The elevated temperature at the base of the cylinder produces an evaporation of water which diminishes the water content. In addition, that drying process induces shrinkage of the bentonite which
Figure 9. Results of the simulation of the infiltration test compared with experiment: a) Temperature distribution; b) Water content distribution; c) Dry density distribution.
is marked by an increase of the dry density at the bottom. On the contrary, at the top, the bentonite is wetted due to the water intake. That hydration process induces a swelling of the bentonite, except at the very top where numerical simulation predict a collapse phenomenon. The cooler region is also resaturated through the vapour that arises from the drying of the bottom part, migrates and condensates in the top part. In terms of water flux, the numerical simulation captures the main registered evolution of cumulative water intake (Figure 10).
1381
Figure 10. Evolution of cumulative volume of water intake. Comparison with experiment.
The good agreement validates the selected coefficient of permeability and its dependence with temperature, porosity and degree of saturation. 4 CONCLUSIONS
The numerical simulation of this infiltration test provides clear evidence of the ability of the used numerical tools to reproduce, from both the qualitative and quantitative viewpoints, the THM processes that occur in an engineered clay barrier. The THM behaviour of the FEBEX bentonite is well-documented in the literature. The results of accurate laboratory experiments were used to calibrate the model parameters through the numerical simulations of a series of oedometric tests. On the contrary, the parameters of the thermal and hydraulic diffusion models were extracted from the literature. The couplings between the thermohydraulic evolution and the mechanical response occurring in the cylinder have been underlined. It shows a combination of thermal and wetting expansion together with irreversible strains of the bentonite induced by hardening effects. This highly non-linear behaviour is a characteristic of the processes occurring in the engineered barrier of nuclear waste disposal. REFERENCES
Bishop, A.W. 1959. The principle of effective stress. Tecnisk Ukeblad, 39: 859863. Charlier, R., Radu, J.P. & Collin, F. 2001. Numerical modelling of coupled transient phenomena. Revue Franaise de Gnie Civil, 5(6): 719741. Collin, F. 2003. Couplages thermo-hydro-mcaniques dans les sols et les roches tendres partiellement saturs. PhD Thesis, University of Lige.
Collin, F., Chambon, R. & Charlier, R. 2006. A finite element method for poro mechanical modelling of geotechnical problems using local second gradient models. International Journal of Numerical Methods in Engineering, 65: 17491772. Collin, F., Li, X., Radu, J.P. & Charlier, R. 2002. Thermo-hydro-mechanical coupling in clay barriers. Engineering Geology, 64: 179193. Franois, B. & Laloui, L. 2008. ACMEG-TS: A constitutive model for unsaturated soils under non-isothermal conditions. Int. J. Num. Anal. Meth. Geomech., 32: 19551988. Gens, A., Garcia-Molina, A.J., Olivella, S., Alonso, E.E. & Huertas, F. 1998. Analysis of a full scale insitu test simulating repository conditions. Int. J. Num. Anal. Meth. Geomech., 22: 515548. Hujeux, J.C. 1979. Calcul numrique de problmes de consolidation lastoplastique. PhD Thesis, Ecole Centrale, Paris. Laloui, L. & Franois, B. 2009. ACMEG-T: A soil thermo-plasticity model. J. Eng. Mech., 135(9), pp. 932944. Laloui L. & Nuth M. 2009. On the use of the generalised effective stress in the constitutive modelling of unsaturated soils. Computer and Geotechnics, 36 (12): 2023. Laloui, L., Franois, B., Nuth, M., Peron, H. & Koliji, A. 2008. A thermo-hydro-mechanical stress-strain framework for modelling the performance of clay barriers in deep geological repositories for radioactive waste. Keynote lecture, Unsaturated Soils: Advances in GeoEngineering: 6380. Lloret, A., Romero, E. & Villar, M. 2004. FEBEX II Project: Final report on thermo-hydro-mechanical laboratory tests. Publicacin tcnica 10/2004, ENRESA. Lloret, A., Villar, M., Sanchez, M., Gens, A., Pintado, X. & Alonso, E.E. 2003. Mechanical behaviour of heavily compacted bentonite under high suction changes. Gotechnique, 53: 2740. Pintado, X., Ledesma, A. & Lloret, A. 2002. Backanalysis of thermohydraulic bentonite properties from laboratory tests. Engineering Geology, 64: 91115. Romero, E., Villar, M. & Lloret, A. 2005. Thermo-hydromechanical behaviour of two heavily overconsolidated clays. Engineering Geology, 81: 255268. Salager, S., Franois, B., El Youssoufi, M.S., Laloui, L. & Saix, C. 2008. Experimental investigations on temperature and suction effects on compressibility and pre-consolidation pressure of a sandy silt. Soils and Foundations, 48(4): 453466. Villar, M. 1999. Investigation of the behaviour of bentonite by means of suction-controlled oedometer tests. Engineering Geology, 54: 6773. Villar, M. 2002. Thermo-hydro-mechanical characterisation of a bentonite from Cabo de Gata: A study applied to the use of bentonite as sealing material in high level radioactive waste repositories. Publicacin tcnica 04/2002, ENRESA. Villar, M., Garcia-Sineriz, J.L., Barcena, I. & Lloret, A. 2005. State of the bentonite barrier after five years operation of an in-situ test simulating a high level radioactive waste repository. Engineering Geology, 80: 175198.
1382
Numerical modelling of the resaturation of swelling clay with gas injection

P. Gerard & J.-P. Radu
Universit de Lige, Dpartement ArGEnCo, Lige, Belgique
J. Talandier & R. de La Vaissire R. Charlier & F. Collin
Agence nationale pour la gestion des dchets radioactifs (ANDRA), Chatenay-Malabry, France Universit de Lige, Dpartement ArGEnCo, Lige, Belgique
ABSTRACT: During long-term storage of radioactive waste, steel containers are corroded and this process leads to hydrogen production. An in-situ experiment has been designed by Andra, in order to analyze the influence of gas migration on the dynamics of resaturation of the bentonite plug. The competition between liquid water coming from argillite and a gas injection at both ends of the plug is studied. Predictive numerical modelling of the experiment is performed in order to support the design, with a special emphasis on the coupling between the water and gas transfers and the mechanical behaviour. 1 INTRODUCTION can alter the good confinement between the canisters and the access gallery. The paper deals with an in-situ experiment that is currently performed by Andra in its underground laboratory at Bure. The objective of the PGZ experiment is the analysis of the dynamic of the bentonite plug resaturation, studying the competition between the liquid water coming from the host formation and a gas injection at both ends of the plug. In order to support the design of the experiment, predictive numerical simulations of the gas migration in both the host formation and the swelling clay plug are achieved, with a special emphasis on coupling between the gas transfers and the mechanical strains and stresses. The influence of damage or plasticity on the fluid transfer parameters, which can explained the preferential pathways, are not taken into account in the modelling of the experiment. In this paper, the concept of the experiment and the boundary value problem of the simulations are described in section 2. The hydro-mechanical model is then defined in section 3. In section 4 the results of the modelling are presented and analyzed, before the conclusions (section 5). 2 BOUNDARY VALUE PROBLEM
In the frame of nuclear waste disposal, the numerical modelling of the underground facilities construction is used to simulate the different physical processes occurring during the live-time of the nuclear waste. The final objective of the design is the following: the host formation and the engineered barrier systems must be as impervious as possible in order to preserve the biosphere from the radionuclides migration. The thermo-hydromechanical modelling needs numerical codes able to tackle this highly coupled problem. In most cases, these numerical computations are achieved under the assumption of constant gas pressure. However, after the introduction of the waste canisters, different potential gas sources exist in the storage gallery (Volckaert et al. 1994, Ortiz et al. 2002). The main origin of the gas is the corrosion of canister steel component. In this latter case, hydrogen is mainly produced and it is thus important to study how the gas will diffuse in the host formation and in the bentonite plugs. Indeed gas overpressure can maybe have a negative impact on the design. Experimental tests showed that gas entry and breakthrough are often accompanied by the development of preferential pathways, which propagate through the samples (Horseman et al. 1999, Delahaye & Alonso 2002, Olivella & Alonso 2008). Moreover the gas migration into the bentonite can delay the resaturation of the plug, which reduces the swelling pressure of the bentonite and
The general concept of the PGZ experiment is presented on Figure 1. A 25 m long horizontal borehole is drilled in the callovo-oxfordian argillite,
1383
Figure 1.
Schematic view of PGZ experiment.
between two existing galleries in the underground research laboratory at Bure. The diameter of the drilling is 101 mm. A plug of MX-80 bentonite is set therein. The plug is 50 cm long and the contact with the borehole is assumed perfect. The gas injection is achieved through two porous stones (with a radius of 85 mm) located at both ends of the plug, imposing a gas pressure higher than the initial water pressure. An axisymetrical modelling of the experiment is performed with the finite element code Lagamine. Lateral boundaries are defined by the symmetry axis of the experiment (Figure 2). The packers and the porous stones are not simulated. The initial water pressure in argillite is 4.5 MPa. The stress state is assumed isotropic and the initial total stress in argillite is equal to 12.3 MPa. Different phases are considered for the modelling. First the borehole drilling is modelled for 1 hour, decreasing the relative humidity at the borehole wall down to 70% and the radial stresses down to the atmospheric pressure. Then a 2 days open-drift period is imposed, during which water pressures remains constant at the wall. During the second step, the MX-80 bentonite plug is set, with an initial water saturation of 70%. A 5 days resaturation phase of the plug is then modelled, during which the displacements are fixed at the plug ends and at the borehole wall (due to the porous stones and the packers). During these two first steps, the gas flow problem is not solved and gas pressure remains constant. Finally, during the third step, a nitrogen gas pressure of 7 MPa is imposed at the end of the plug (gas pressure is increased in 4 days and then maintained constant). The other hydraulic and mechanical boundary conditions remain the same as the ones imposed during the second step. 3 BIPHASIC FLOW MODEL AND MECHANICAL BEHAVIOURS
Figure 2. Boundary conditions during the gas injection phase and cross sections and elements definition for the results.
phase. The advection of each phase is modelled by a Darcys flow for unsaturated cases. The diffusion of the components is taken into account in each phase by the Ficks law. The quantity of dissolved nitrogen is proportional to the quantity of gaseous nitrogen, controlled by Henris law. Details about the exact formulation of the biphasic flow model and the value of the water and nitrogen properties used in the modelling are available in (Gerard et al. 2008). The retention curves of the different materials are given by the van Genuchtens relationship: Sr,w Sres Smax Sres
m
Sr,w
1 pc 1 m 1 Pr 1 if pc 0
and (1)
with Sr,w the water relative saturation, Smax the maximal saturation, Sres the residual saturation, pc the capillary pressure and Pr and m two parameters of the van Genuchtens law. Cubic relations of the effective saturations are adopted for the water and gas relative permeability functions: Sr,w Sres kr,w Smax Sres
3
To reproduce water and nitrogen flows in porous media, a biphasic flow model is used. The developed model manages explicitly liquid water and dissolved nitrogen in the liquid phase and a mixture of vapour water and gaseous nitrogen in the gaseous
1 Sr,w kr, g A. Smax Sres
(2)
A coefficient A is introduced in the gas relative permeability function in order to distinguish the
1384
intrinsic permeability according to the considered fluid (water or gas). However such experimental data are available neither for the MX-80 bentonite nor for the callovo-oxfordian argillite. The values of this parameter for the modelling are based on experimental results on Boom clay and FEBEX swelling clay (Volckaert et al. 1994, Villar 1998), that show a high gas intrinsic permeability for the swelling clay. All the hydraulic parameters of argillite and bentonite are presented in Table 1. For the considered materials and stress levels, the solid grain deformability is no more negligible and the general Biot framework is used to model the hydro mechanical coupled terms (Biot 1941). The proposed mechanical model for the bentonite is a simplified version of the BBM model (Alonso et al. 1990), expressed in terms of the net stress and the suction and considering only the nonlinear elastic behaviour described by the following relationship:
e em es d v d v d v
Table 2. Mechanical parameters for the different materials. Argillite E0 0 c b Young modulus (MPa) Poisson coefficient (-) Cohesion (MPa) Friction angle () Biot coefficient (-) 4000 0.30 3 20 0.6 Bentonite 150 0.2 1.0
with ij the effective stress, b the Biot coefficient, ij the Kronecker symbol. The mechanical parameters for the different materials are presented in Table 2. 4 NUMERICAL RESULTS
m ds d K Ks
(3)
m is the mean net stress, s is the suction, where K is the bulk modulus and Ks is the bulk modulus related to suction changes. Details about the evolution of Ks with suction and the value of these coefficients for bentonite can be found in (Gerard et al. 2008). The mechanical law for the argillaceous rocks is an associated elastoplastic perfectly plastic model, defined by a Drucker-Prager yield surface. This constitutive mechanical model is written in terms of the Bishops definition of effective stress (Nuth & Laloui 2008):
ij ij b Sr,w pw 1 Sr,w pg ij

(4)
Table 1. Hydraulic parameters for the different materials. Argillite

sat kint
Bentonite 9.6 1020 0.35 0.3789 18 1 0.4 107 0.0494
m Pr Smax Sres A
Intrinsic permeability (m2) Porosity (-) van Genuchten coefficient (-) van Genuchten parameter (MPa) Maximal saturation (-) Residual saturation Gas permeability parameter (-) Tortuosity (-)
4.8 10 0.18 0.3289 15 1 0.01 100 0.25
20
A first modelling without gas injection shows a time resaturation at the centre of the bentonite plug about 30 days (Figure 3Node 2). For the gas injection modelling, nitrogen pressure is imposed at both ends of the plug the 7th day. The swelling clay is not totally saturated at the beginning of the injection, what makes easier the nitrogen migration into the gaseous phase. The high gas permeability of the bentonite ensures a quasi instantaneous homogeneous gas pressure of 7 MPa into the plug, as observed on the gas pressures profiles along a section through the plug (Figure 4). The water relative saturation becomes also quickly homogeneous in the plug (Figure 3). Moreover nitrogen injection avoids the long-term total resaturation of the swelling clay, related to the imposed gas pressure higher than the initial water pressure. A low coupling between the gas migration and the water transfers is observed, evidenced by the water overpressures obtained at the centre of the plug end when gas is injected (Figure 5). Nevertheless, these water overpressures disappear after 30 days and no water pressure higher than the initial water pressure is observed, which decreases the risk of hydro-fracturation. The gas pressure profiles along a section through the bentonite and argillite (at the centre of the plug) are presented on Figure 6. The vertical dotted line indicates the transition between the swelling clay and the argillaceous rock. Gas pressures are homogeneous in the plug. Then nitrogen migrates in argillite and argillite is desaturated on 25 cm long after 1 month. The slope change in the gas profiles corresponds to the transition between the saturated domain (where the diffusion of dissolved nitrogen is the predominant transfer mechanism) and the unsaturated domain (where the advection of gaseous nitrogen prevails), as explained in (Gerard et al. 2008).
1385
1 0.95 0.9 0.85
7 6 5
Pg (MPa)
t=11 days t=1 month t=6 months t=1 year t=3 years
4 3 2
Sr,w(-)
0.8
0.75 0.7 0.65 0.6 0 5 10
N2 injection
Node 1 - Without gas injection Node 1 - With gas injection Node 2 - Without gas injection Node 2 - With gas injection 15 20 25 30
1 0
100
200
300
400
500
r (mm)
Time (days)
Figure 3. Time evolution of water relative saturation in the bentonite plug with and without gas injection.
8 7 6
Figure 6.
Gas pressure profiles along section C1.
Pg (MPa)
5 4 3 2 1 0 0 50 100 150
t=11 days t=1 month t=6 months t=1 year t=3 years
200
250
Figure 7.
1.00 0.98 0.96 0.94
Gas pressures distribution after 1 year.
Y (mm)
Figure 4.
10
Gas pressures profiles along section C2.
0
-10
10
20
30
40
50
60
Srw (-)
0.92 0.90 0.88
Pw (MPa)
-20
Without gas injection With gas injection
0.86 0.84 0.82 0.80
7 days t=11 days t=1 month t=6 months t=1 year t=3 years
-30
100
200
300
400
500
r (mm)
-40
-50
Figure 8. Relative water saturation profiles along section C1.

Time (days)
Figure 5. Time evolution of water pressure with and without gas injection at the center of the plug end.
The nitrogen migration is also illustrated on Figure 7, where the gas pressures distribution is drawn after 1 year. The nitrogen moves in argillite, developing an elliptic gaseous front around the plug. The desaturation observed in the bentonite plug and in argillite is low (Figure 8). The influence
of the gas injection on the coupling between the mechanical behaviour and the fluids transfers is thus reduced, according to the Bishops effective stress definition (equation 4). For instance, the axial swelling behaviour in bentonite is weakly influenced by the gas injection and a swelling pressure about 4.5 MPa is recovered (Figure 9), which should ensure a sufficient confinement of radioactive waste for practical applications of long-term storage.
1386
7 6 5
Radial swelling pressure (MPa)
t=1 month t=6 months t=1 year t=3 years
1 0 -1 -2 -3 -4 -5 -6 -7 0 10 20 30 40 50
*Y (MPa)
4 3 2 1 0
50
100
150
200
250
-8
Y (mm)
Time (days)
Figure 9.
Axial net stress profiles along section C2.

12
Van Eekelen criterion Beginning of drilling phase Beginning of resaturation phase Element A Element B Element C
Figure 12. Time evolution of radial swelling pressure in element D without gas injection.
10
II [MPa]
9,5 days 6
0 -30 -10 10 30 50 70
I [MPa]
Figure 10. Stress paths at different points of argillite without gas injection.
12
Van Eekelen criterion Beginning of drilling phase Beginning of resaturation phase Beginning of gas injection phase Element A Element B Element C
10
0 -30 -10 10 30 50 70
I[MPa]
Figure 11. Stress paths at different points of argillite with gas injection.
The Figures 10 and 11 show the stress paths without and with gas injection at different integration points of argillite (defined on Figure 2). The different modelling steps are evidenced on the stress paths. For the problem without gas injection (Figure 10), a small zone of argillite around the borehole becomes plastic during the resaturation phase (elements A and B). At the contact with the plug end (element A), argillite remains plastic during the resaturation phase, mainly due to the stress concentration at the corner between the swelling
clay and the argillaceous rock. At the middle of the bentonite plug (element B), argillite remains plastic until the development of swelling pressures in the bentonite (from 9,5 daysFigure 12). The radial swelling of the plug affects the stresses in argillite. Radial stresses in argillite increase and orthoradial stresses decrease, which implies a reduction of the deviatoric stress and an elastic behaviour. In element C located 10 cm behind the borehole wall, the behaviour remains elastic during all the resaturation phase, which shows the small damage zone developed around the cavity. With a gas injection phase, the stress paths at the three elements remain mostly similar (Figure 11). It shows the low coupling of the gas transfers on the mechanical behaviour. The main difference is observed at the borehole wall, at the middle of plug (element B). At the beginning of the gas injection, the behaviour becomes immediately elastic, whereas a plastic zone subsists without gas injection. It can be explained by the bentonite behaviour during the first injection days. Gas pressure is increased to 7 MPa at the plug end in 4 days. At the same time, the net stresses at the middle of the plug remain negligible, because the water relative saturation is still too low to develop swelling pressures in the bentonite (Figure 13element D). During the first injection days, it implies an increase of the total stress imposed by the bentonite to argillite, which is mainly controlled by the gas pressures increase (Figure 13element D). At the same time, the water pressures in argillite (element B) dont rise so fast as the gas pressure in the bentonite. Moreover, the contribution of the gas pressure into the Bishops effective stress is negligible, because the desaturation is low (Figure 13element B). As a consequence, in spite of the absence of swelling pressure in bentonite during the first 4 injection days, a radial confinement of argillite is observed, which involves a decrease of the deviatoric stresses. The argillite behaviour becomes elastic faster than without gas injection.
II[MPa]
1387
12 10
Bentonite - Element D - Total stress Bentonite - Element D - Radial swelling pressure Argillite - Element B b(Sr,m pw + 1 Sr,w pg) Argillite - Element B bSr,w pw
Pressure - Stress (MPa)
8 6 4 2 0
0-2
-4 -6
10
20
30
40
50
Time (days)
Figure 13. Time evolution of total stress in bentonite and hydraulic contribution to Bishops effective stress in argillite with gas injection.
Then radial swelling pressures appear, which induces a new decrease of the deviatoric stress. 5 CONCLUSIONS
An in-situ experiment in its underground laboratory at Bure is proposed by Andra in order to study the influence of gas production on the dynamic of resaturation of the bentonite plugs. Predictive numerical modelling is performed to support the design of the experiment. The modelled in-situ test consists of a borehole drilled, inside which a plug of MX-80 bentonite is set. The bentonite is naturally resaturated by water coming from the argillaceous rock. At the same time a gas pressure, higher than the initial water pressure in the host rock, is imposed at both ends of the plug. The developed model takes into account mechanical equilibrium, water and gas transfers in partially saturated conditions. It manages explicitly liquid and vapour water, gaseous and dissolved nitrogen. Elastoplastic and non linear elastic model are used, respectively for the argillite and for the swelling clay. The numerical results show the instantaneous gas migration in the plug, explained by the high gas permeability of the bentonite. Gas pressures in the swelling clay are thus homogenous and equal to the imposed gas pressure. Then nitrogen migrates in argillite and a 25 cm long zone is partially desaturated. Water overpressures are observed in the bentonite, due to the fast migration of the gas. Nevertheless, these water overpressures are not higher than the initial water pressures in the host rock and disappear quickly, which reduce the risk of hydrofracturing. Numerical results highlight the low coupling between the mechanical behaviour and the gas transfers. It can be explained by the Bishops effective stress definition and the low water relative saturation observed in argillite, which reduces the influence of the gas pressures on the mechanical stresses and strains. The analysis of
the argillite stress path at the borehole wall (at the middle of the plug) shows that the gas injection implies a decrease of the deviatoric stress and an elastic behaviour, whereas argillite remains longer plastic without gas injection. The gas pressures distribution is mainly influenced by the highly gas permeability of bentonite. It is based on experimental data on FEBEX bentonite (Villar 1998). Experimental data on the fluid transfers in the MX-80 bentonite and in the callovooxfordian rocks in the quasi-saturated domain are needed in order to define with precision retention curves and gas and water relative permeability functions, especially in the quasi-saturated domain. Other coupling effects could be taken into account, as the development of an excavated damage zone around the borehole and its influence on the argillite transfer parameters (for instance the permeability). Such coupling could evidence the development of preferential pathways for the gas migration. The PGZ experiment is presently performed in the underground laboratory at Bure. The comparison between experimental data and the predictive numerical results will follow in order to improve our models. REFERENCES
Alonso, E.E., Gens, A., Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40:405430. Biot, M.A. 1941. A general theory of three-dimensional consolidation. Journal of Applied physics 12:155164. Delahaye, C.H. & Alonso, E.E. 2002. Soil Heterogeneity and Preferential Paths for Gas Migration. Engineering Geology 64: 251271. Gerard, P., Charlier, R., Barnichon, J.-D., Su, K., Shao, J.-F., Duveau, G., Giot, R., Chavant, C. & Collin, F. 2008. Numerical modelling of coupled mechanics and gas transfer around radioactive waste in long-term storage. Journal of theoretical and applied mechanics 38(12): 2544. Horseman, S.T., Harrington J.F. & Sellin, P. 1999. Gas Migration in Clay Barriers. Engineering Geology 54: 139149. Nuth, M. & Laloui, L. 2008. Effective stress concept in unsaturated soils: clarification and validation of a unified framework. International Journal of Analytical and Numerical Methods in Geomechanics 32:771801. Olivella S. & Alonso E.E. 2008. Gas flow through clay barriers. Gotechnique 58(3): 157176. Ortiz, L., Volckaert, G. & Mallants, D. 2002. Gas Penetration and Migration in Boom Clay, a Potential Host Rock Formation for Nuclear Waste Storage. Engineering Geology, 64: 287296. Villar, M.V. 1998. Ensayos para el proyecto FEBEX. CIEMAT-report 70-IMA-L-5-51, prepared for ENRESA. Volkaert, G, Ortiz, L., De Canniere, P., Put, M., Horseman, S.T., Harrington, J.F., Fioravante, V. & Impey, M. 1994. Modelling and experiments on gas migration in repository host rocks. MEGAS Project, Final Report, Phase 1.
1388
THMC analysis of saturation and heating processes of an expansive clay barrier in radioactive waste isolation
L. do N. Guimares A. Gens
Federal University of Pernambuco, Recife, Brazil Technical University of Catalonia, Barcelona, Spain
M. Snchez S. Olivella
Texas A&M University, College Station, USA Technical University of Catalonia, Barcelona, Spain
ABSTRACT: A fully coupled formulation combining reactive transport and an existing thermohydro-mechanical (THM) code is briefly described. Special attention has been given to phenomena likely to be encountered in clay barriers used as part of containment systems of nuclear waste. The types of processes considered in the chemical formulation include hydrolysis, complex formation, oxidation/reduction reactions, acid/base reactions, precipitation/dissolution of minerals and cation exchange. Both kinetically controlled and equilibrium-controlled reactions have been incorporated. The formulation has been implemented in the numerical code CODE_BRIGHT. An application is presented concerning the performance of a large scale in situ heating test simulating high-level nuclear waste repository conditions. 1 INTRODUCTION In this paper a fully coupled THMC formulation is briefly described that allows the performance of numerical analyses considering all four aspects of behaviour: thermal, hydraulic, mechanical and chemical and their interactions. The formulation takes into account the possibility that the material is partly saturated. The formulation is applied to the analysis of a real large scale in situ heating test simulating nuclear repository conditions. 2 2.1 DESCRIPTION OF THE FORMULATION THM formulation
Compacted swelling clays are often envisaged as the main component of engineered barriers for radioactive waste disposal. These barriers are subjected to thermal loading due to the heat emitted by the waste and to hydration from water coming from the adjacent rock. As a consequence of these thermo-hydraulic phenomena, mechanical and chemical changes arise that, in turn, may affect all other aspects of behaviour. A correct understanding and prediction of these barriers would require, therefore, the performance of fully coupled thermo-hydro-mechanical and chemical (THMC) numerical analyses. Although there have been a number of coupled THM formulations proposed and the corresponding numerical codes have been available for some time (e.g. Gawin et al., 1995; Olivella et al., 1994, 1995), coupled THMC formulations have lagged behind. Only recently, a number of proposals have been put forward in this field (e.g. Guimares et al., 2007; Thomas et al., 2001). The research is now reaching a state that, through the incorporation into efficient computer codes, it is possible to perform coupled THMC numerical analyses to tackle engineering problems of a high level of complexity.
Reactive transport equations are solved together with the THM equations. The THM formulation presented here is based on the one proposed in Olivella et al. (1994) applied to clays. The formulation is defined by the governing balance equations, constitutive equations and equilibrium restrictions. Balance equations apply to mass of water (unknown: liquid pressure, Pl), mass of air (unknown: gas pressure, Pg), internal energy (unknown: temperature, T) and linear momentum (unknown: displacements, u). Constitutive equations and equilibrium restrictions relate, directly or indirectly, the variables
1389
that appear in the balance equations with the main unknowns. As an example, the balance equation for the mass of water is expressed as: w w w (Sl l wlw Sg g wg ) ( jw l jg ) f t (1)
species concentrations is established that allows a reduction in the degrees of freedom of the reactive transport problem (Guimares et al., 2007). 3 3.1 EXAMPLE OF APPLICATION The FEBEX in situ heating test
where is the porosity, Sl and Sg the liquid and gas degree of saturation, l and g, the liquid and gas densities in Kg/m3 of phase, wlw the mass fraction of water in the liquid (close to one in dilute solutions) and wgw the mass fraction of water vapour in the gas phase. The term Sl l wlw is the water content in the liquid phase per unit volume and Sg g wgw the water content in the gas phase. jlw and jgw are the mass flux of water in the liquid and gas phases respectively. The term f w may represent, in the context of this paper, the water production or loss associated with precipitation or dissolution of minerals. An example of constitutive equation is the retention curves that relates the degree of saturation Sl to the unknowns Pl, Pg and T. An example of equilibrium restriction is the determination of the water vapour mass fraction using the psychrometric law. Full details of the THM formulation are presented in Olivella et al. (1994). 2.2 THMC formulation
The basic formulation has been extended to include chemical phenomena in the analysis, allowing the modelling of coupled THMC problems (Guimares et al., 2007). Now a number of additional chemical species are considered in the liquid phase. The mass continuity of each chemical species is expressed by the reactive transport equation: (Sl l ci ) ji Ri t (2)
A large scale in situ heating test is being performed at the Grimsel Test Site, an underground research laboratory located in the Swiss Alps. The rock in the site is mainly granite. The test tries to reproduce the main features of the current Spanish concept for deep underground disposal of high level radioactive waste (Febex, 2000). The test layout (Figure 1) consists in placing two heaters at the end of a drift of 2.28 m diameter and 70.4 m length especially bored for this purpose. The diameter and length of the heaters (0.9 m and 4.54 m respectively) corresponds to the actual dimensions of the canister envisaged for radioactive waste storage. The heaters are placed in the axis of the drift at a 1 m distance from each other. The space between the drift and the heaters is filled by blocks of compacted bentonite with a smectite content in the range of 88%96%. The test is heavily instrumented with measurements of temperatures, relative humidity (equivalent to total suction), pore pressures, displacements, and stresses. The test schedule involves switching on the heaters an applying increasing power until the temperature reaches a value of 100C at some point in the bentonite. From that moment on, the power of the heaters is constantly adjusted in order to keep the maximum temperature in the bentonite barrier at the 100C mark. The test was run in this way for five years until one of the heaters was switched off and dismantled. The experiment continues with the other heater without a planned finishing date. 3.2 Features of analysis and chemical model
4.33 m
4.54 m
1m
2.70 m
17.40 m
Figure 1.
Layout of the Febex in situ heating test.
1390
I H
where, ci is the concentration of species i in mol/Kg of solution and Ri is the total production rate of species i due to chemical reactions, in mol/m3/s. ji is the total flux of species i expresses mol/m2/s. As in Olivella et al. (1994), this flux is considered as the sum of advective and non-advective fluxes. Advective flux is, in turn, the sum of the movement of the liquid phase with respect to the solid phase (governed by generalized Darcys law) and of the solid phase with respect to the reference configuration. The non-advective flux is the consequence of molecular diffusion and mechanical dispersion. In addition to the usual homogeneous reactions in liquid phase (aqueous complex formation, acid-base and oxidation-reduction), heterogeneous reactions (dissolution/precipitation of minerals and cation exchange) are also taken into account. In the chemical equilibrium problem, a relationship between the
Only a brief overview of the analysis characteristics will be given. More information is reported in
TEST AREA
CONCRETE PLUG BENTONITE HEATER 1 LINER HEATER 2
2.28 m
0.90 m
CONTROL SYSTEM A B1 C D1 L E1 F1 M1 N G M2F2 E2 D2 B2 INSTRUMENTED SECTIONS 4.54 m 2.98 m
Febex (2000) and Gens et al. (2002). The numerical computations have been carried out using the fully coupled THMC version of CODE_BRIGHT. Radial symmetry has been assumed resulting in a 1-D axisymmetric analysis that is quite appropriate for examining the behaviour of the bentonite barrier and immediate adjacent rock. After a short initial period of applying increasing power to the heaters, a constant 100o C temperature has been applied to the contact between the heater and the buffer, in correspondence with the test protocol. Although the full heating test has lasted for five years, the analysis has been run up to 100 years to check and predict long term THMC conditions. The rock thermal, hydraulic and mechanical boundary conditions have been based on the results of the comprehensive site investigation carried out at the site. The initial conditions of the bentonite were as follows: dry density 1,7 g/cm3 and water content 14.4%. This results in an initial degree of saturation of 0.65 and an initial suction of 115 MPa. Both the bentonite (solid phase and interstitial water) and the hydration water were subjected to a full chemical characterization. The thermo-hydromechanical parameters were determined in an independent laboratory testing programme. The selected primary species (Guimares et al., 2 2007) are H2O, Ca2, K, Na, Mg2, Cl, SO4 , HCO3 , H, SiO2(aq), NaX. They are all in the liquid phase except adsorbed sodium, NaX, where X stands for the clay mineral. The secondary species, in 2 equilibrium with the primary ones, are: OH, CO3 , CO2(aq), CaCO3(aq), CaHCO3 , CaSO4(aq), CaCl, MgSO4(aq), MgHCO3 , MgCl, NaHCO3(aq), NaSO4 , NaCl(aq), KSO4 and HSiO3. A total cation exchange capacity CEC 86.2 meq/100 g of solid has been used. The exchangeable cations (CaX2, NaX, KX and MgX2) are also secondary species. Finally, the phases in equilibrium with the solution of the liquid phase are calcite (CaCO3), gypsum (CaSO4.2H2O), anhydrite (CaSO4), chalcedony (SiO2) and carbonic gas (CO2) at 103.5 atm. As a result of these assumptions the number of unknowns per node was 13 : 1 temperature, 1 liquid pressure, 1 displacement and 10 primary chemical species (excluding water). 3.3 Results of the THMC analysis
Figure 2 shows the variation of temperature at those points. Close to the heater the temperature is maintained at 100o C. At the other two points there is, after the initial transient period, a gentle temperature rise over time. Figure 3 shows the evolution of degree of saturation for the first five years. It can be observed that there is a strong drying close to the heater followed by a milder hydration afterwards. The part of the barrier close to the rock becomes quickly saturated, but the central part is still unsaturated after five years. In fact, the long term calculations (Figure 4) indicate that more than 35 years would be required to achieve
100
80
temperature (C)
60
40
buffer near heater central buffer

20
external buffer
0 0 400 800 1200 1600 2000
time (day)
Figure 2. Variation of temperature with time (5 years) at three points of the bentonite barrier.
1.00
0.80
liquid saturation
buffer near heater central buffer external buffer

0.60
Only a small selection of results will be shown here. They have been chosen to demonstrate the capabilities of the formulation and to point out some of the significant THMC features of the problem. THM results are given for three points of the buffer: one near the heater, one in the central part of the barrier and one closer to the rock (external buffer).
0.40 0 400 800
time (day)
1200
1600
2000
Figure 3. Variation of degree of saturation with time (5 years) at three points of the bentonite barrier.
1391
1.00
1E+0
Aqueous Concentration (mol/Kg of liquid)
Ca++ at 0 years Ca++ at 100 years Na+ at 0 years

1E-1
Na+ at 100 years K+ at 0 years K+ at 100 years
0.80
liquid saturation
1E-2
buffer near heater central buffer external buffer

0.60
1E-3
1E-4
0.40 10000
time (day)
20000
30000
40000
1E-5 0 2
Distance to axis (m)
10
Figure 4. Variation of degree of saturation with time (100 years) at three points of the bentonite barrier.
0.4
Figure 6. Distribution of the aqueous concentration of cations Ca2, Na and K in the bentonite barrier and rock at various times.
Exchangeble cations (meq/100g of dry bentonite)
40
0 days
Ionic strength (molality)
0.3
45 days 1 year 5 years
CaX2
32
MgX2
24
0.2
100 years
NaX
16
0.1
0 years 5 year 100 years
0.0 0 2
KX
0 0.4 0.6
Figure 5. Distribution of ionic strength in the bentonite barrier and rock at various times.
0.8
1.0
1.2
Figure 7. Distribution of exchangeable cations concentration in the bentonite barrier at various times.
complete barrier saturation. More information on the THM results is given in Gens et al. (2002). Figure 5 presents the ionic strength distribution in the buffer and in the rock at various times up to 100 years. Naturally the ionic strength in the granitic water is much lower than that of the bentonite water. It can be noted that the diffusion process is slow, after 100 years the solutes have only penetrated about 8 m in the rock. The ionic strength near the heater increases because of evaporation but the value reached is not very high. The computed concentrations for some of the individual cations can be examined in Figures 6
and 7. The tendency is towards a dilution of the aqueous species in the barrier, especially, at shorter times, near the rock (Figure 6). Concentrations in the rock increase progressively due to diffusion. The exchangeable cations in the buffer clay do not vary much (Figure 7). There are some initial changes due to the impact of heating but the final cation content in the solid phase after 100 years is rather similar to the initial one. It should be stressed, however, that there are some uncertainty over the details of cation exchange processes are elevated temperature.
1392
0.05
chalcedony (% of dry bentonite)
0.04
0.03
0 yeas 1 year
0.02
5 years 50 years 100 years
computer code (CODE_BRIGHT) in order to perform numerical analysis. An example of application has been briefly described involving the simulation of a quite complex case: a large scale in situ heating test that simulates the conditions of a repository for high level nuclear waste. It can be concluded that, using the computational tool presented, the performance of coupled THMC analysis or real engineering problems is already a feasible proposition. ACKNOWLEDGEMENTS This work has been supported by ENRESA and the European Commision. The authors are also grateful for the financial support given by CAPES (Coordenao de Aperfeioamento de Pessoal de Nvel Superior, Brazil). REFERENCES
FEBEX, 2000. Full-scale engineered barriers experiment for a deep geological repository for high level radioactive waste in crystalline host rock. Final Report. In Technical Publication 1/2000. Madrid: ENRESA. Gawin, D., Baggio, P. & Schhrefler, B.A., 1995. Coupled heat, water and gas flow in deformable porous media. In Int. J. Num. Meth. Fluids, 20, 967987. Gens, A., Guimares L. do N., Garca-Molina, A., Alonso, E.E. 2002. Factors controlling rock-clay buffer interaction in a radioactive waste repository. In Engineering Geology, 64, 297308. Guimares, L. do N., Gens, A. & Olivella, S. 2007. Coupled thermo-hydro-mechanical and chemical analysis of expansive clay subjected to heating and hydration. Transport in porous media, 66, 341372 Olivella, S., Carrera, J., Gens, A., & Alonso, E. E., 1994. Nonisothermal multiphase flow of brine and gas through saline media. In Transport in Porous Media, 15, 271293. Olivella, S., Gens, A., Carrera, J. & Alonso, E. E., 1995. Numerical formulation for a simulator (CODE_ BRIGHT) for the coupled analysis of saline media. Engineering Computations, 13, 87112. Thomas, H.R., Cleall, P.J. & Hashm, A.A., 2001. Thermal/hydraulic/chemical/mechanical (THCM) behaviour of partly saturated soil. Desai et al. (ed.) Computer Methods and Advances on Geomechanics, Rotterdam: Balkema, vol. 1, 743748.
0.01
0.00 0.4 0.6 0.8 1.0 1.2
Figure 8. Distribution of chalcedony concentration in the bentonite barrier at various times.
To show the type of results concerning the behaviour of precipitated minerals, Figure 8 is presented showing the distribution of chalcedony concentration at various times. It can be noted that the mineral is dissolved at the two ends of the barrier. Near the rock the mineral dissolves because of the entry of very dilute granitic water. In contrast, near the heater there is initially some precipitation due to water evaporation induced by heating. However, in the long term, the mineral disappears from the inner zone of the barrier due to the higher solubility of the chalcedony leading to an increase in the concentration of SiO2(aq). This originates a molecular diffusion transport that leads to the development of a dissolution front (clearly visible in the 100 years results) that travels towards the cooler outer part of the barrier. 4 CONCLUSIONS
The paper has presented a very general and fully coupled THMC formulation that incorporates a significant number of THM processes and homogeneous and heterogeneous chemical reactions. The formulation has been incorporated into a
1393
Laboratory studies of the addition of bentonite to landfill cover soil

R.L.S. Izzo, C.F. Mahler & K. Huse
COPPE/UFRJ, Rio de Janeiro, Brazil
ABSTRACT: The material used to cover landfills is an important factor both during their lifetime and upon their closure. Different procedures have been studied to reduce the permeability of landfill covers in efforts to enhance recovery of the gases produced by waste decomposition and to reduce the amount of leachate produced by rainwater infiltration. Among these, the method of mixing small quantities of bentonite in the cover soil has been shown to be a technically and economically feasible solution. Not only does coverage with this mixture reduce permeability, it also prevents the formation of cracks from shrinkage during the drying process. To ascertain the ideal percentage of a special sodium bentonite developed for use in soil, we designed a device incorporating a tensiometer to monitor the suction, moisture and temperature of soil samples and to provide characteristic curves for the samples tested. This paper presents details of this device and the differences in behavior of the landfill cover in function of the percentage of bentonite added to the cover material. The results indicate the efficacy of using bentonite mixtures and of the device developed to study them. 1 INSTRUCTION bentonite employed as an additive to improve the performance of the cover soil. We evaluated the formation of cracks during the drying process by measuring the volumetric variation of the cover soil sample. We also monitored the moisture variation (indirectly through the weight variation of the sample), temperature (at the top and bottom of the sample) and suction (at the base of the sample). 2.1 Tensiometer
We previously developed an automatic tensiometer to measure the soil suction in the laboratory and various field situations (Mahler et al., 2002, 2004; and Diene and Mahler, 2007). We tested the device in the laboratory in mini-lysimeters at different stages of its development. The results showed that the device works very well under different conditions. Its application to study the behavior of landfill cover systems is an interesting alternative because this allows monitoring the effects of different cover soil formulations. One of the important aspects of landfill cover systems is to reduce permeability, which in turn decreases the amount of leachate formed from rainwater infiltration and enhances the capture of landfill gas for energy production. We subsequently developed a more elaborate device to allow study of the crack formation process caused by shrinkage of the cover layer from drying, incorporating this tensiometer. This combined device measures the suction, moisture and temperature of the cover sample, to provide insight into the effectiveness of various cover soil formulations that include bentonite to increase permeability. 2 MATERIALS AND METHODS
In this section we present the tensiometer and the associated instruments developed to study the process soil shrinkage, along with the soil and the
A tensiometer is a traditional device to measure soil suction (soil moisture tension). It basically consists of a porous ceramic stone, a pressure transducer and a metal, glass or plastic cover to contain these elements (Diene and Mahler, 2007). A pressure transducer has a diaphragm that distends according to the pressure of the liquid or gas acting on its closed end in a watertight compartment. On the other side of this compartment, at atmospheric pressure, an extensometer measures the diaphragms deflection through the change in electrical resistance of the diaphragm, which varies with this movement. The porous element acts as an interface of the soil, water and measurement system. A metal, glass or plastic cover contains the various elements, so that the porous element, transducer, water and soil can interact together. If the soil is not saturated, the transducer will measure negative pressures (suction), while if it is saturated the pressures will be positive (leakage). The water present in the voids
1395
in the soil must have hydraulic continuity with the water in the porous element, i.e., the porous element must be in contact with the soil to produce the pressure readings. The tensiometer can run either on electricity from the power grid or batteries in the field. It generates analog output signals that are stored in a data logger. Traditional tensiometers are prone to cavitation, which generally occurs at around 80 kPa. This can be prevented by using devices constructed as previously described (Mahler et al., 2002; Mahler et al., 2004; Mahler et al., 2005; Diene and Mahler, 2007) and summarized in the items below: a. Use of a high-quality porous ceramic stone, with high bubbling pressure; b. Use of deaerated water during saturation of the ceramic stone and assembly of the tensiometer; c. Use of plastic or very smooth steel in assembling the tensiometer to avoid the formation of air pressure between the water and the devices wall during saturation of the system; d. Careful assembly, saturation and calibration; e. Special care in selecting the technical specifications and characteristics of the transducer. With the second prototype of this tensiometer, we managed to measure suctions of up to 350 kPa (Mahler et al., 2004), and subsequently with a third prototype, pressures up to 1450 kPa (Diene and Mahler, 2007). As mentioned, the tensiometer developed for this project consists of three elements: a pressure transducer, a porous ceramic stone and a plastic body (Figures 1 and 2).
2.2 Sample preparation 2.2.1 Introduction To prepare the soil samples, the clumps were first broken up and the soil was passed through a number 10 sieve (2-mm mesh). The samples were then homogenized with water and placed in a humid chamber for at least 24 hours. Afterward they were compressed in a tripartite cylindrical mold (of the type used to measure the California bearing ratio CBR) wrapped in a plastic film (Figure 3) to keep the soil from sticking to the mold. Two holes were drilled in the mold for placement of thermometers at the top and the base. If there are no further authors place the cursor one space behind the word ABSTRACT: and type your abstract of not more than 150 words. The top of the first line of the abstract will be 73 mm (2.87) from the top of the type area. The complete abstract will fall in the abstract frame, the settings of which should also not be changed (Width: Exactly 187 mm (7.36); Height: Automatic; Vertical 73 mm (2.87) from margin; Lock anchor). 2.2.2 Equipment for shrinkage testing The device developed allowed not only measurement of the soil shrinkage itself (by measuring the variation of volume), but also the suction, temperature and moisture (through weight variation). Figure 4 shows a schematic of the device and Figure 5 shows the elements and the assembled device. The data on volume and weight variation are recorded by hand while the data on temperature and suction are recorded automatically. 2.3 Soil and bentonite used 2.3.1 The soil studied The soil used in this study was obtained from the waste treatment station in the city of Nova Iguau in the state of Rio de Janeiro. This soil is utilized as cover for the municipal landfill and consists of sand, silt and clay components. Figure 6 shows the grain size distribution.
Figure 1.
View of the pressure transducer used.
Figure 2. Two views of the plastic body and porous element assembled.
Figure 3. Assembly of the tripartite cylindrical mold CBR size.
1396
SAMPLE
MEMBRANE
BURETTE
ACRYLIC RECIPIENT
TENSIOMETER METAL BASE
TERMOMETER
Figure 4. Schematic of the equipment developed for the shrinkage test. Figure 6. Grain-size distribution curve of pure soil, soil mixed with 5% bentonite, pure bentonite and soil mixed with 10% bentonite.
Figure 5. device.
(a) plastic body, (b) metal base, (c) assembled
The specific gravity (SG) of this soil was determined to be 2.67. The Atterberg limits were also obtained: the liquid limit was 44.5% water content and the plastic limit was 21.12%. A soil compression test was also performed according to the ABNT NBR 7182 standard from the Brazilian Association of Technical Standards. The compression energy was normal on the Proctor scale, with optimum moisture of 24.7% and apparent specific dry weight of 1.52 g/cm3. The hydraulic conductivity, measured through variable load testing, was on the order of 108 m/s. According to Ferreira (2006), the field capacity of this soil is a moisture content of approximately 31% (see Figure 7). 2.3.2 Bentonite The additive used to reduce the soil permeability was a commercial sodium bentonite containing additives, produced by the company Bentonit Unio Nordeste S.A. This bentonite has a hydraulic conductivity of about 1 1014 m/s (Ferrari, 2005). The grain-size curve of this bentonite is shown in Figure 6. The real grain density (Gs) of this bentonite was determined to be 2.78. The liquid limit was 540% and the plastic limit was 60% (Ferrari, 2005). 2.3.3 Soil/bentonite mixture as a cover system Among the problems faced by sanitary landfills are infiltration of rainwater and escape of methane gas
Figure 7. Characteristic curves for the soil from Nova Iguau (Ferreira, 2006).
from waste decomposition. Various cover systems have been developed to impede this infiltration and gas escape, among them textile or plastic tarps and various types of soil mixtures. The impermeability of soil cover systems can become impaired due to erosion, drying (causing shrinkage and cracking) and the presence of animals and vegetation. One of bentonites properties is its great ability to expand. Adding bentonite to natural soil reduces the latters permeability, thus making it more resistant to the formation of cracks and better at preventing rainwater infiltration. Determination of the ideal amount of bentonite to add to a particular soil for landfill cover is done by measuring the hydraulic conductivity of the soil without bentonite and with various percentages of bentonite added. In this study, after initial testing of permeability with various soil/bentonite ratios, we chose two bentonite percentages. The results are shown in Figure 8.
1397
Figure 8. Hydraulic conductivity versus percentage of bentonite added.
Figure 9. Comparison of the suction for the soil with different bentonite percentagesCBR size.
It can be seen that only with the addition of at least 4% bentonite by dry weight to the soil did the hydraulic conductivity reach the order of 109 m/s (represented by the points below the broken line in Figure 8). While this value is the lower limit considered adequate for landfill cover material by the Rio de Janeiro state environmental agency, it is not considered sufficient by the counterpart authorities in the United States and Germany, both of which suggest a hydraulic conductivity on the order of 1010 m/s. From analyzing the data obtained, we decided to focus the studies on pure soil and mixtures containing 5% and 10% bentonite. Thus, we performed characterization tests of the soil with addition of 5% bentonite by dry weight. The liquid limit was 75.5% and the plastic limit was 28.26%. There was only a slight change in the grain-size curve of the soil mixed with 5% bentonite in relation to the pure soil, with a slight increase in the fraction of clay and silt (Figure 6). The big gain was in the geotechnical characteristics, with an increase in the plasticity indexes. The optimum moisture for soil with 5% bentonite, determined by the compression test, was 23.4%, corresponding to an apparent specific dry weight of 1.55 g/cm3, and the hydraulic conductivity was on the order of 109 m/s. 3 RESULTS
Figure 10. Comparison of the moisture versus volumetric variation for the CBR-size test bodies.
Besides suction, we also tested the pure soil and soil with 5% and 10% bentonite for moisture variation in relation to volumetric variation (Figure10). It is important to mention that the samples were prepared above the optimum compression. Also, the initial moisture levels of each sample were different because it was not possible to harmonize them. These factors evidently caused some discrepancies in the samples behavior in the drying tests performed. Consideration must also go to the possible variability of the material in each sample and small differences during their compression due to the human factor. 4 CONCLUSIONS
We ran tests for the pure soil and with addition of 5% and 10% bentonite by dry weight. The suction results, obtained with a tensiometer coupled to the drying test equipment, are shown in Figure 9. There was an increase in suction with increasing bentonite content in the soil. This occurs due to the increase in fine grains in the soil with the addition of the bentonite.
As expected, we found that bentonite improves the original soil characteristics, by significantly reducing the permeability starting at addition of 4%. Besides this, the addition of bentonite improved the soils plastic behavior, contributing to its expandability and reducing the formation of cracks.
1398
The greater the percentage of bentonite added, the greater was the increase in suction measured (Figure 9). On the other hand, the addition of bentonite led to an increase in plasticity, resulting in slightly greater contraction and lower permeability. The device developed performed very well, supplying important data to evaluation the formation of cracks in compacted soil layers used to cover sanitary landfills. AKNOWLEDGEMENTS We thank the Brazilian research support agencies CNPq, FAPERJ and DAAD for funding, and Bentonit Unio Nordeste Ltda and Nova Gerar/ Haztec for donating materials used in the study. REFERENCES
DIENE, A.A.; MAHLER, C.F., Um instrumento para Medida de potencial Matricial nos solos sem Ocorrncia de Cavitao. (An instrument to measure the potential matrix of soils without the presence of cavitation). R. Bras. Ci. Solo, v. 31, pp. 792802, 2007. Brazilian Journal of Soil Science (in Portuguese) FERRARI, A.A.P., Viabilidade da Utilizao de Silte Compactado como Material de Impermeabilizao em Aterros de Resduos (Viability of the use of compacted silt as material for the impermeabilization of
landfills). M.Sc. Dissertation, USP, So Paulo, SP, Brazil, 2005. (in Portuguese) FERREIRA, M.A.S., Aplicao de Modelos de Avaliao Qualitativa e Quantitativa dos Percolados em um Aterro Sanitrio (Application qualitative and quantitative models to evaluate leachate in a landfill). M.Sc. Dissertation, COPPE/UFRJ, Rio de Janeiro, RJ, Brazil, 2206. (in Portuguese) HUSE, K., Estudo da Influncia da Adio de Bentonita em um Solo Areno-Siltoso para Uso como Cobertura de Aterros (Study of the influence of the addition of bentonite to a silty-sandy soil for use in a landfill cover system). M.Sc. Dissertation, COPPE/UFRJ, Rio de Janeiro, RJ, Brazil, 2007. (in Portuguese) MAHLER, C.F.; PACHECO, A.C.C.S.; GONALVES, H.S., Development of an Automatic Tensiometer in the Laboratory Using a Mini-Lysimeter. In: UNSAT, 2002, Recife. 4th International Congress on Unsaturated Soils. Rotterdam: A.A. Balkema Publishers, 3, 2002. MAHLER, C.F.; PACHECO, A.C.C.S.; SOUZA, H.G., Desenvolvimento de um Tensimetro para Medio de Suces Elevadas (Development of a tensiometer for measurement of high suctions). In: 5 Simpsio Brasileiro de Solos No-Saturados, 2004, Anais. 2004 (In: Annals of the 5th Brazilian Symposium on Nonsaturated Soils). So Carlos. Anais. CD-ROM. (in Portuguese) MAHLER, C.F.; DIENE, A.A.; GONALVES, H.S., Um novo instrumento para Medida de Suco nos Solos (A new instrument for suction measurement of soils). Solos e Rochas - Revista Brasileira de Geotecnia, So Paulo, Brazil, v. 28: n. 3, pp. 309318, 2005. (in Portuguese)
1399
Thermal-hydro-mechanical behavior of unsaturated bentonite blocks: Experimental approaches

J.O. Lee, W.J. Cho, S.K. Kwon & J.W. Choi
Korea Atomic Energy Research Institute, Yuseong-gu, Daejeon, Republic of Korea
ABSTRACT: Bentonite blocks are used for the buffer of a high-level waste repository and the understanding of their thermal-hydro-mechanical behavior is essential to predict and assess the barrier performance of the bentonite buffer. Material property measurements and a mock-up test KENTEX were carried out to investigate the thermal-hydro-mechanical behavior of the bentonite buffer for a Korean reference disposal system. For Korean bentonite, the thermal conductivity increased with an increasing dry density and water content. The hydraulic conductivity decreased with increasing the dry density, while it increased a little with an increasing temperature. The suction was sensitive to the water content, and the swelling pressure increased with an increasing dry density. In the mock-up test, preliminary and qualitative THM results were obtained: the temperature rapidly reached a steady state, while the water content underwent a slow change in its distribution. The total pressure had a gradual increase corresponding to the build-up of the swelling pressure caused by the evolution of the saturation front from the hydration surface. 1 INTRODUCTION 2 MATERIALS AND METHODS
A reference disposal system for high-level wastes in Korea (KRS) was developed by the Korea Atomic Energy Research Institute in 2002. The design concept of the KRS is based on a multi barrier system which is composed of engineered barriers and surrounding plutonic rock, and it will be constructed in bedrock several hundred meters below the ground surface (Kang et al., 2002). The engineered barrier system of the KRS consists of a HLW-encapsulting disposal container, a buffer between the container and the wall of a borehole, and a backfill in the inside space of the emplacement room, to isolate the HLW from the surrounding rock masses. Korean bentonite blocks are being considered for use as a buffer. The bentonite blocks, when emplaced in a bore hole of the emplacement room, remain unsaturated, and they are exposed to the radioactive decay heat from the high-level wastes, the intrusion of ground water from the surrounding rock, and the swelling pressure of the bentonite buffer. These conditions may affect the thermal-hydro-mechanical (T-H-M) behavior of the bentonite blocks and eventually the barrier performance of a HLW repository. In this connection, this study carries out laboratory tests to measure the material parameters of bentonite blocks and a mock-up test KENTEX to investigate the coupled T-H-M behavior of the bentonite buffer for a Korean reference disposal system.
2.1 Bentonite The raw bentonite used in this study was taken from the Jinmyeong mine of Kyeongju, Korea (hereafter, referred to as Korean bentonite). It was dried below 110 C, pulverized, and then passed through No. 200 of ASTM (American Society for Testing and Materials) standard sieves. The Korean bentonite contains montmorillonite (70%), feldspar (29%), and small amounts of quartz (1%). The Atterberg limit of the bentonite is as follows: a liquid limit of 244.5%, a plasticity of 46.1%, and a plasticity index of 198.4%. The compacted bentonites for the material property measurements are prepared by adjusting the water content of the bentonite and then compacting it to a desired dry density. The adjustment of the water content is done by adding de-mineralized water to the dried bentonite by pre-weighed amounts using an ultra sonic humidifier. After the water content adjustment, the bentonite is compacted in a cell, designed for each material property measurement, to a desired dry density using a hydraulic press, and then it is left in a desiccator for more than 3 days for its water content equilibrium. The bentonite blocks used the KENTEX test were prepared as described in the following section 2.3.
1401
2.2
Material parameter measurements
2.2.1 Thermal conductivity Thermal conductivity is the property of a material that indicates its ability to conduct heat. It is defined as a heat flux (energy per unit area per unit time) divided by a temperature gradient (temperature difference per unit length). A quick thermal conductivity meter (Kyoto Electronics, QTM-500) was used to measure the thermal conductivity of compacted bentonite. The dimensions of the compacted bentonite for the measurement are 0.15 0.06 0.02 m. All the measurements were done at 25 C, examining the dependence of the thermal conductivity upon the dry density and water content of compacted bentonite. 2.2.2 Hydraulic conductivity Hydraulic conductivity is defined as the ratio of Darcys velocity to the applied hydraulic gradient by Darcys law. An experimental apparatus shown in Fig. 1 was employed to measure the hydraulic conductivity of compacted bentonite. The dimensions of the compacted bentonite were a diameter of 0.05 m and a height of 0.025 m. The water was supplied from the bottom to the top of the chamber at a hydraulic pressure of 0.92 MPa depending on the dry density of the compacted bentonite. The measurements were conducted to investigate the effect of the dry density of the compacted bentonite and the temperature on the hydraulic conductivity. 2.2.3 Suction Suction is a property that drives the groundwater movement in the unsaturated compacted bentonite by capillary force. The measurement of the suction was conducted using an indirect technique by means of a capacitive humidity sensor as shown in Fig. 2. The compacted bentonite has the dimensions of 0.05 m in the inner
Figure 2. Experimental apparatus for the measurements of suction.
diameter and 0.05 m in the length, and it has a hole drilled in it to allow a capacitive hygrometer sensor (Vaisala HMT #334) to be installed inside. The suction is determined from the measured humidity (RH) using Kelvins law (Fredlund and Rahardjo, 1993): S 10 6 RT RH ln Vw 100 (1)
where R is the universal gas constant (8.3143 J/K mol), T the absolute temperature (K), and Vw the molar volume of water (1.8 105 m3/mol). The measurements were carried out in experimental conditions: a volumetric water content of 3%, 7%, 12%, 17%, and 22%, a dry density of 1.5 Mg/m3 and 1.6 Mg/m3, and a temperature of 25 C, 45 C, and 70 C. 2.2.4 Swelling Pressure Bentonite, upon its contact with water, has a force of repulsion among its particles and subsequently increases in volume. Swelling pressure is the pressure by which the bentonite acts on a system with constant volume. It was measured using the experimental apparatus shown in Fig. 3. The apparatus includes four sections: main test section, pressurized water-feeding section, vacuuming section, and signal and data processing section. The main test section consists of a stainless steel cylindrical cell of which the dimensions are 0.05 m in height and 0.05 m in diameter, porous metal filters to avoid the loss of bentonite particles, and pressure sensors mounted in a vertical direction. The pressurized water-feeding section consists of a stainless steel reservoir and a highly pressurized helium gas bomb which is used for feeding the water in the reservoir to the compacted bentonite sample. The stainless steel reservoir was designed to tolerate a pressure of 3 MPa. In the vacuuming section,
Figure 1. Experimental apparatus for the measurements of hydraulic conductivity.
1402
C 6
2 IBM Compatible 1 2 3 5 4
A : vacuuming section processing section 1 : cylindrical cell feeding tank
B : main test section C : signal and data D : pressurized-water feeding section 2 : pressure sensor 4 : helium gas bomb 3 : pressurized-water 5 : vacuum pump
Figure 3. Experimental apparatus for the measurements of swelling pressure.
Figure 4. View of the KENTEX facility and a cross section of its confining cylinder.
3 a vacuum pump was connected to the outlet of the main test section to extract the air entrapped in the compacted bentonite. The signal and data processing section was designed to discern swelling pressures from the pressure sensor and then to display their values on digital indicator. The measurement is performed at 25 C until the swelling pressure reaches a steady-state. 2.3 Mock-up test (KENTEX)
MATERIAL PARAMETER CHARACTERISTICS
KENTEX (KAERI Engineering-scale T-H-M Experiment for an Engineered Barrier System) is a large scale of experiment to investigate coupled T-H-M behavior in an engineered barrier system. It was designed as a third of the engineered barrier system adopted for the Korean reference repository concept (Kang et al., 2002). The KENTEX consists of a confining cylinder with a cylindrical heater and bentonite blocks, a hydration tank, and a data logging and instrumentation system (Fig. 4). The vertical steel confining cylinder simulates a disposition hole excavated in the floors of the emplacement rooms in the repository, and its inner diameter and height are 0.75 m and 1.36 m, respectively. There are 24 nozzles in the side wall of the confining cylinder to supply groundwater to the bentonite blocks. The cylindrical heater with a diameter of 0.41 m and a length of 0.68 m simulates a disposal canister generating decay heat and includes three 1 kW heating elements. The bentonite blocks with a dry density of 1.5 Mg/m3 are made of Korean bentonite as mentioned in the previous section 2.1. The temperature at the interface of a cylindrical heater and the bentonite buffer is maintained at 90 C. Sixty eight sensors are installed to measure the temperature, the humidity, and the total pressure. The groundwater is injected through the nozzles on the side wall of the confining cylinder from the hydration tank with a constant pressure of 0.5 MPa.
The thermal conductivities of the blocks of compacted bentonite with the dry densities of 1.2 to 1.8 Mg/m3 were measured within the water content range of 10 wt% to 20 wt%. As shown in Fig. 5, the thermal conductivity of the compacted bentonite increased with increasing the dry density in a constant water content, and increased with increasing the water content in a constant dry density. However, the thermal conductivities were nearly constant at a temperature range of 20 to 100 C. The hydraulic conductivities were measured for the compacted bentonite with the dry densities of 1.4 to 1.8 Mg/m3. They were found to decrease with an increasing dry density of the bentonite. The relation between the logarithm of the hydraulic conductivity, K (m/s) and the dry density of the bentonite, d (Mg/m3), can be fitted to a straight line expressed as follows (Fig. 6 (a)): log K 4.07d 6.13 r2 0.92 (2)
Fig. 6(b) shows that the hydraulic conductivities of the compacted bentonite at a temperature of 80 C were up to about 3 times greater than those at 20 C, and such a change of the hydraulic conductivities in the bentonite at elevated temperatures is attributable to the changes in the viscosity and the density of the water (Cho et al., 1999). The suction values of compacted bentonite ranged from 9.1 MPa to 386.8 MPa under the given experimental conditions. Fig. 7(a) represents the suction variation as a function of water content for the compacted bentonite with the dry densities of 1.5 Mg/m3 and 1.6 Mg/m3. As shown in the figure, the suction values decreased with increasing the water content for each dry density. This is probably due to an increase in the ratio of water to air in the voids filled with water and air in unsaturated compacted bentonite. It was also shown in Fig. 7(a) that the suction increased a little with
1403
Figure 8. Dependence of the swelling pressure on the dry density and temperature.
Figure 5. Thermal conductivities of the compacted bentonite.

10
-9
10
-9
(a)
Hydraulic Conductivity (cm/s)
10
-10
Dry density = 1.8 g/cm3 open symbols: Cho et al.[5] solid symbols: this study
(b)
10
-10
10
-11
10
-11
log K = -4.0 7d - 6.13 ( N=21, r2 = 0.92 )
10
-12
due to surface tension decreases and thus the suction decreases as the temperature increases. The swelling pressure developed rapidly over the early period, and then reached a nearly constant value after 15 days. The swelling pressures were measured for 30 days in which they were regarded as a steady state. They were found to increase with an increasing dry density in the range of 1.4 to 1.8 Mg/m3. The relation between the swelling pressure, Ps and the dry density of the bentonite, d can be fitted to a straight line expressed as follows (Fig. 8(a)): Ps 5.48 102 exp (d/2.36 101) 16.469 r2 0.99 (3)
10
-12
K ( cm/s )
1.4 1.6 1.8
10
-13
20
40
60
80
100
o
120
140
160
Bentonite Dry Density (g/cm3)
Temperature ( C )
Figure 6. Hydraulic conductivities versus (a) dry density at a temperature of 20 C, and (b) temperature at the dry density of 1.8 Mg/m3.
There was a little increase of the swelling pressure with increasing temperature from 20 to 80 C, as shown in Fig. 8(b). 4 COUPLED T-H-M BEHAVIOR OF UNSATURATED BENTONITE BLOCKS
Figure 7. Suction versus water content for the dry densities and temperatures.
the dry density of compacted bentonite, which is attributed to an increase of capillary effect by a smaller void size resulting from the compaction of the bentonite to a larger dry density. The effect of temperature on changes in the suction as a function of the water content is shown in Fig. 7(b). As seen in the figure, the suction variation was negligible with an increase in the temperature for the water content up to 17%. At the water content of 22%, there was a little variation of the suction for 70 C. That is, the suction was lower than those for lower temperatures, which implies that the capillary force
The KENTEX has been successfully operated since it started, and the performance and reliability of the installed sensors and instruments have been good except for the humidity sensors. Regarding the humidity sensors only one is still operative and the rest have malfunctioned. The lack of hydro data due to the failure of the humidity sensors were compensated for by a core-drilling method. The profiles of temperature at several locations in the bentonite blocks for 760 days are shown in Fig. 9. The temperature data obtained were relatively stable and uniform, and they responded sensitively to a variation of the electric power supplied to a heater. As shown in the figure, the temperature distributions had a similar trend regardless of the position of the thermocouples. The temperature rapidly reached a quasi-steady state within about 2 weeks, and after then slowly approached a constant value. The closer it was to the wall of a confining cylinder, the shorter the time was for the temperature to reach a steady state. This is probably
1404
Figure 9. Temperature profiles at various locations in the bentonite blocks.
Figure 10. Evolution of the relative humidity and water content distributions of core samples.
due to an increase in the thermal conductivity at the groundwater-wetted part close to the wall (i.e., hydration surface) of the confining cylinder. Fig. 10(a) is the humidity profile at a height of 0.34 m (D) and a radial distance of 0.246 m as a function of time, which presents the output of the only-functioning humidity sensor. The humidity sharply increased during an initial hydration, and then it decreased followed by a gradual increase as time passed. This behavior may be explained as follows: first, the water rapidly intruded through crevices between the bentonite blocks and/or bentonite block and a sensor or its cable (which may be a preferential path for a water inflow), abruptly wetting the blocks; secondly, the wetted blocks desiccated as the water moved outwards due to a thermal gradient, leading to a decrease in the humidity; and thirdly, after the hydration overcame the drying process, the blocks were wetted again by the gradual inflow of water from the hydration surface. Fig. 10(b) and 10(c) represent the water content distributions ((b) a height of 0.61 m (the midpoint of E and F), and (c) a height of 1.03 m (just above H)) determined by using the core-sampling method. The water contents close to the hydration surface were higher than those in the heater part, and they approached a saturation value ( 31.4%) beyond a point of about 0.32 m apart from the heater. In this figure, the abnormally high values of the water content near the hydration surface appears to be experimental errors which resulted from the inflow of water into the geotextile placed between the bentonite blocks and the wall of the cylinder during the core sampling.
Figure 11. Total pressure distributions in the bentonite blocks at various measurement points.
The profiles of total pressure in the bentonite blocks are shown in Fig. 11. The pressure data acquired from the pressure sensors ranged from 0.8 to 16 kg/cm2. The pressure profiles, as shown in this figure, reveal the general trend that there is a gradual increase in the pressure values as time passes and, at a constant time, the values near the hydration surface are higher than those at the heater side. This is probably attributable
1405
to the build-up of the swelling pressure caused by the evolution of the saturation front from the hydration surface. However, it can be seen at an initial stage of the operation below about 100 days that the pressure profiles had different patterns depending on the location of the pressure measurement. The pressure profile at a point close to the hydration surface increased immediately when the bentonite blocks were contacted with the water (Figure 11(a)), while those at points farther apart from the hydration surface followed a pattern of an initial period during which there was a small positive reading (Figures 11(b) and 11(c)) due to the thermal expansion of the bentonite blocks near the heater and then a continuously increased reading as the bentonite blocks were saturated with the intruding water. All the pressure profiles of Fig. 11 had two peaks at the early stage, which seemed to be a system-intrinsic behavior occurring due to the water in-flowing rapidly through the crevices between the bentonite block(s) and the sensor or sensor cable at the start of the test. 5 CONCLUSIONS AND FURTHER WORK
The temperature reached a steady state in a short time after the test started. The temperature increased as the heater was approached. The water content was higher close to the hydration surface than it was in the heater part. The relative humidity data suggested that a hydration of the bentonite blocks might occur by different drying-wetting processes depending on their position. The total pressure increased continuously due to the evolution of the saturation front in the bentonite blocks and thereby increase of the swelling pressure. There was also a contribution from the thermal expansion of the bentonite blocks near the heater. Final conclusions will be drawn by conducting modeling work to explain the observed T-H-M behavior and to verify the hypothesis of the T-H-M processes in the bentonite blocks. ACKNOWLEDGEMENT This work has been performed under the Nuclear R&D Program by the Ministry of Education and Science. REFERENCES
Cho, W.J., Lee, J.O., and Chun, K.S. 1999. The temperature effects on hydraulic conductivity of compacted bentonite. Applied Clay Science, 14, 47. Fredlund, D.G., and Rahardjo, H. 1993. Soil mechanics for unsaturated soils. John Wiley & Sons Inc., NY, 517. Kang, C.H., Kim, J.W., Chun, K.S., Park, J.H., Cho, W.J., Choi, W. J., Lee, J.O., Lee, Y.M., and Hwang, Y.S. 2002. High level radwaste disposal technology development/Geological disposal system development. KAERI/RR-2336/2002, Korea Atomic Energy Research Institute.
The material property measurements of compacted bentonite for the buffer of the KRS lead to the following conclusions: The thermal conductivity increased with an increasing dry density and water content. The hydraulic conductivity decreased with increasing the dry density, while it increased a little with an increasing temperature. The suction was sensitive to the water content, and the swelling pressure increased with an increasing dry density. The current T-H-M behavior in the bentonite blocks of the KENTEX allows us to draw preliminary and qualitative conclusions.
1406
Water infiltration and methane emission through three different cover layers of an experimental Municipal Waste Landfill at Muribeca, Recife, Pernambuco, Brazil
R.L. Lopes
Department of Education and Natural Resources Technology, Federal Institute for Education, Science and Technology of Rio Grande do Norte (IFRN), Brazil
M.C.M. Alves J.F.T. Juc
Department of Civil Engineering, Federal University of Rio de Janeiro (UFRJ), Brazil Department of Civil Engineering, Federal University of Pernambuco (UFPE), Brazil
ABSTRACT: Studies in landfill covers have been done in order to control CH4 emissions across layers without compromising leachate generation. The purpose of this paper is to present preliminary Landfill gas (LFG) emission investigation and infiltration conducted in three different cover layers in an Experimental Cell at Muribecas Landfill in Brazil: Methanotrophic Layer (mix clay/compost layer), Capillary Barrier Layer (clay/gravel layer) and Conventional Layer. Results show that capillary barrier has major capability to retain biogas because of the distribution gravel layer below the soil layer and saturation maintenance above air incoming point in most of the observation period. Methanotrophic cover layer followed the same emission probably pattern due to methane oxidation. On the other hand, conventional layer was more efficient for minimizing infiltration than the other two, and the capillary barrier presented much higher infiltration in periods where the rainfall was greater than 25 mm in 3 days. 1 INTRODUCTION from the 3 top covers built on an experimental cell in Brazil. 2 GENERAL DESCRIPTION OF THE EXPERIMENTAL COVERS
Most of biogas from biodegradation processes of Municipal Solid Waste (MSW) in Brazil goes to atmosphere through simple burning. Projects of recovering of energy out of Sanitary Landfills are limited to large landfills in Brazil. Cover layer is though the system that acts as a barrier to the incoming of liquid into the landfill as well as a barrier to out coming of biogas to the atmosphere. Cover layers are submitted to climate changes all over the year. So accompanying their performance for a longer period of time is necessary to establish adequate relations between climate and cover material properties related to infiltration and gas emission. Three different cover layers were put on in an experimental cell constructed in Muribeca Landfill, in the Metropolitan Region of Recife, Pernambuco, Brazil: (MET) mix clay/compost layer (methanotrophic layer), (BAC) stratified clay and gravel layer (capillary barrier layer) and (CONV) conventional layer. Complementary water infiltration was also studied in each cover type using lisimeters. This study presents the first results on the investigation of water infiltration and biogas emission
The Experimental Cell dimensions are 65 m 85 m 9 m deep, filled up with 36,659 t of waste from June/2007 to February/08. The investigation was conducted in the top cover layer of the cell, which is divided in three types of cover as shown in Figure 1, according to the following characteristics: Methanotrophic layer (MET): layer composed of 30 cm of compacted soil under a mix of soil and compost layer. The proportions of the compost are of 50% volume of the mixture, which corresponds to 75% soil and 25% compost in natural weight. The compost used was taken out from the compost unit of the landfill plant. This depth of the MET layer varies from 0.40 to 0.75 m and covers an area o 590.2 m . Capillary barrier (BAC): layer composed of 0.20 m of gravel under a compacted soil layer with total thickness varying from 0.50 to 0.90 m.
1407
Metanotrophic layer
0,3-0,45m 0,3m
(MET)
0,1-0,2m 0,3m
75% soil + 25% compost Compacted soil
Waste
Capillary layer
0,55-0,7m 0,2m
(BAC)
Compacted soil
geotextile
0,3 - 0,55m 0,2m
gravel
Waste
Convencional layer
Figure 2.
Lisimeter for infiltration measurements.
0,75-0,85m
Compacted layer
0,5-0,65m
Waste
Figure 1.
Cover types of the experimental cell.
The superficial area is 500.3 m . Geotextile was used in the soil-gravel interface. Conventional layer (CONV): compacted sandy clay varying from 50 to 80 cm in thickness with superficial area of 534.8 m . 3 METHODOLOGY
For the soil characterization, several tests carried out such as soil classification, granulometry, compaction curve, Atterberg limits, water permeability (NBR-7181/88; NBR-6459/84; NBR-7180/88; NBR7182/86; NBR-14.545/00) and characteristic curve according to the ASTM D5298 (1994) procedures. It is relevant to mention that physico-chemical analysis of the compost used in the methanotrophic layer were taken such as pH, water content, volatile solids, C, N, P and K. The same type of soil was used in the three experimental covers. The infiltration assessment was done within June and December 2009. The infiltrated water it was daily measure in a lisimeter with dimensions of 60 60 45 cm, made of 3 mm carbon steel plates carried with anticorrosion treatment. The lisimeter is located below the top cover as illustrated in Figure 2. The water infiltrated was collected through a tap installed on the lateral slope. Next to the lysimeter water content and temperature sensors were installed within 3 depths.
Rainfall data were obtained from INMET (Curado StationRecife/PE) and from another station located inside the leachate treatment station of Muribecas landfill. The methodology used to evaluate the gas emission through the top covers consisted in CH4 superficial flux quantification using a static flux chamber and evaluation of biogas and soil parameters in depth until the contact with the solid waste. Fluxes of CH4 and CO2 through the experimental covers at Muribecas landfill were measured using the static flux chamber technique described by Maciel, (2003). Various parameters were measured like concentration of the gases CH4, CO2 and O2, internal temperature (between the chamber and the top cover), external temperature, internal and atmospheric pressures. All measurements were taken every 5 minutes during the test. The overall duration of the test ranges between 30 to 60 minutes and the gas flux is determined by the initial inclination of the curve CH4 mass versus time. CH4 mass flow rates and coefficient of linear determination (R2) were obtained to every single test carried out in the three experimental top covers. By the end of each test, a ring of soil was taken to further analysis in the lab to determine the in-situ soil density, water content and dry soil density. In total, 65 plate tests were carried out on the top covers between September 2008 and August 2009. To complete this investigation, 25 additional tests were carried on the cells slopes (lower and upper) and on the edges of the cell aiming to map the CH4 flow liberated to the atmosphere. 4 RESULTS AND DISCUSSION
4.1 Soil characterization The soil used in this research was obtained from the borrow area that supplies material for the cover
1408
cells of Muribecas landfill. The compost used in the mixture was taken out from the composting unit of the landfill plant. Geotechnical characterization as well as compaction data for soil, compost and the soil/compost mixture are presented in Table 1. According to the Unified Soil Classification System (USCS), the soil is classified as silty sand (SM). The water permeability determined in saturated condition in the Tri-Flex Permeameter, w was 2.0 108 m/s. The compost has average quantity of volatile solids of 46.8% 1.6% and the measured water content was 53.0% 1.4%. The initial pH was 7.75, chemical composition in g/Kg: 15.65 of Ntotal, 2.06 of Ptotal and 3.70 of Ktotal, besides 260 g/Kg of C. C/N relation was 16.6. Figure 3 presents experimental data taken from the soil characteristic curve and a correspondent preview from Gardners model (1956). Experimental data indicates an air incoming spot at a suction of approximately 500 kPA which corresponds to a volumetric water content of approximately 0.3 m3/ m3 for both capillary barrier and conventional layer.
Table 1. data. Geotechnical characterization and compaction Soil/compost (50/50) v/v 50.9% 21.9% 23.6% 2.0% 1.6% 41.5 12.3 12.4 33.1%
This value leads to a saturation degree around 70%. The characteristic curve from the mix soil and compost is in course. Studies about unsaturated condition soil in landfill cover layer are important to understand the behavior of the material and establish water content intervals where biogas emissions as well as water infiltration are minimized. 4.2 Infiltration
Geotechnical characterization Size 0.075 0.0750.42 0.422.0 2.04.75 4.76 LL IP Compaction data dmx(kN/m3) Wopt
The infiltration analysis in the 3 top covers has been made daily by verifying the amount of water infiltrated in the lysimeter installed in the interface soil/solid waste. Results of infiltration during the studied period are shown on Table 2 and Figure 4. Although the short evaluation period, it has been observed that the methanotrophic layer, due to its porosity characteristics, presents a higher infiltration during the period of high rainfall when compared to the conventional cover. The minimum infiltration has the same magnitude values for all covers during the low precipitation periods when the saturation degree becomes lower and, consequently, also does the unsaturated permeability. The capillary barrier did not work properly between June and August, although it has presented water content similar to the conventional layer. During this period, precipitation is intense in this region leading to great percolation volume into waste mass.
Soil 55.1% 23.2% 20.2% 1.0% 0.7% 48.0 11.2 16.9 16.5%
Compost 1.0% 34.6% 44.8% 13.9% 5.7% nd nd nd nd
Table 2. Infiltration in the 3 top covers Month jun/09 jul/09 aug/09 sep/09 oct/09 nov/09 dec/09 Rainfall (mm) 333,0 387,4 288,4 83,3 16,3 49,9 47,8 MET 32,0% 12,2% 15,8% 0,0% 1,6% 0,6% 0,9% Infiltration BAC 96,7% 86,7% 75,3% 0,0% 0,0% 0,5% 0,3%
Rainfall Infiltration MET Infiltration BAC Infiltration CONV 400 350 300 250 200 200,00 150,00 100,00 50,00 0,00 1-jun 1-jul 1-ago 1-set month 1-out 1-nov 1-dez 150 100 50 0 Infiltration (mm)
CONV 12,3% 0,1% 0,0% 0,0% 0,0% 0,0% 0,0%
0,45 0,40 Volumetric water content (m3/m3) 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 1 10 100 1000 10000 Matric Succion (kPa)
Rainfall (mm)
Gardner Model (1956) Experimentaldata
Rainfall x Infiltration
450,00 400,00 350,00 300,00 250,00
Figure 3.
Soil water retention curveCONV/BAC.
Figure 4. Infiltration in the top covers from June to December 2009.
1409
It was observed that these periods matched with accumulated rain over 25 mm in 3 days. This is basically due to the configuration of this layer, which contains 20 cm of gravel below soil cover. That leads to an interruption of the suction in this interface and allows the direct infiltration until the decrease of water content below the field capacity of the material. These events of free drainage were observed during the period when saturation degree of the cover surface was over 90%. Quantity of infiltrating water only started reducing in September; when the saturation of the cover reached values lower than 75%. Regarding to the months of September and December, the high evaporation rate linked to a low precipitation volume caused a water deficit, lowering the water content in all top covers. During these months it is noticed the brutal decrease of infiltration due to this decrease of water content, which coincides with the high temperatures, leading to a similar infiltration performance on the three covers. Gee, et al (2006) quotes in his project the disadvantages of capillary barriers, the high cost, and mentions that when saturated, those systems originate flux instability, leading to high rates of fluid transportation inside the landfill. Hupe, et al (2003) while investigating 3 systems of top covers to stabilized landfills with compost on the surface of the capillary barrier also found out that this barrier was overcharged during the intense precipitation periods. It caused the suction rupture. In this project as well, the authors got to the conclusion that the compost on the surface of the cover increase the capacity of water storage and that variations on temperature and precipitation rates on the surface are quickly reflected on variations of surface moisture, but not significant to the underground. 4.3 Methane flux
The emission rates through the top cover of the experimental cell were obtained between September 2008 and August 2009. It is important to point that this scenario of emissions reflects both dry and rainy seasons and there was no forced extraction of gas during the tests, in order to simulate the conditions of small and medium landfills in Brazil, that do not have this device in their plants. During all the observed period methane flow on the top cover varied from 0.0 to 1,332.11 g/m .day. The greatest values were observed on the conventional cover. This variation is consistent with those reported in literature presented by Boeckx et al, (1996), Czepiel, et al, (1996), Tanaka et al, (1997) e Mocet et al, (2003). Regarding to other landfills located in the same region where gas drains operates with free
discharge in atmospheric pressure, average methane flow values of 362.9 g/m .day (MACIEL, 2003) and 400 g/m .day (MARIANO, 2008) were found. From all 65 flux tests in this research, only 6 presented measurement over 240 g/m .day. Two of them performed in the region close to the vertical gas drains turning it possible to conclude that the impermeabilization around the drains region wasnt efficient enough to minimize the fugitive emissions. Tests that aimed to evaluate the leaks around the drains were not taken into account to determine superficial gas emissions. Table 3 presents the results of flux measured in each period of evaluation as well another parameters of soil grouped according to the top cover type. According to data exposed on Table 3, in all periods the BAC layer presented the lowest average rate of emissions although this layer is less thick than the others two ones. This may be caused by better distribution of biogas due the presence of rocks below the soil, which minimizes hot spots, as well as due to high values of saturation average degree (SD). The BAC and MET layers have the same magnitude of saturation degree average and they are larger when compared to CONV layer of saturation degree. It was verified that the methanotrophic layer presented emission in the same magnitude as the capillary barrier. This can be attributed to the oxidation process of methane due to physic-chemical characteristics of the material used (soilcompost mix). The methanotrophic differs from the other layers in the amount of organic matter and pH values. Organic matter and pH values are 16% and 7.5, respectively for MET layer. These circumstances allow a favorable environment to methanotrophic bacterial growth. On the other hand much lower values are found in the other two layers (7% and 6.0, respectively). Average moisture content in depth in MET layer is greater than the other two ones, as shown in Table 3, probably due to the higher water retention
Table 3. Fugitive CH4 emissions and soil parameters.
Layer Period Range CH 4 superficial emission 2 (g/m .day) 0 - 74.1 0 - 23.4 0 - 3 1.6 0 - 63.4 0 - 6.7 0 2.1 - 984.7 64.3 - 137.6 0 - 194.9 WC (%) 20.8 7,5 20.5 4.7 26.6 1.4 16.9 0.6 19.2 6.2 21.6 1.4 14.9 1.0 16.4 4.3 21.5 1.6 SD (%) 60.4 17.9 61.6 13.4 82.3 5.4 58.9 2.1 67.1 21.6 80.1 6.0 52.1 3.4 56.4 15.1 75.1 5.6 90% 92% 78% CD (%)
Sep-Dec/08
MET
Jan-Apr/09 May-Aug/09 Sep-Dec/08
BAC
Jan-Apr/09 May-Aug/09 Sep-Dec/08
CONV
Jan-Apr/09 May-Aug/09
WC water content measured in various depths of cover layer, SD Average Saturation degree measured in various depths of cover layer, CD Compaction degree.
1410
capacity of the soil/compost layer. This condition offers a more propitious environment for growth of methanotrophic microorganisms favoring methane oxidation within cover soil. Additionally higher temperatures were measured throughout MET layer depth which may indicate oxidation reactions. These observations were also analyzed on projects carried by Boeckx et. al. (1996), Borjesson et al., (1997), Cabral et al., 2007 e Gebert et. al. (2007), where they analyzed the influence of temperature, pH and moisture to the oxidation potential in oxidative covers. Higher hot spots measured on the conventional and methanotrophic covers indicates the existence of punctual fluxes as verified in situ. That was attributed to the presence of big cracks observed during the period of high temperatures, besides to leaking around the gas drains and close to the superficial drainage, where cover layer is less thicken. 5 CONCLUSIONS
barrier does not work properly, due to reaching of internal breakthrough suction value, which turns to be of inadequate usage in regions with high rainfall rates. The methanotrophic layer presented monthly infiltration ranging from 12% to 32% for high precipitation period. In months with precipitation below 100 mm, few infiltration could be measured through methanotrophic layer. The same tendency was also observed in the other two types of coverage. The conventional layer as might be expected has the lowest infiltration rate. Considering the short monitoring time the results presented so far related to water balance are not conclusive. Nevertheless one can point to an association of methanotrophic layer and capillary barrier as a promising design to alternative cover layers due to the capacity of minimizing gas emission as well as to present acceptable water infiltration. The recommended profile should be from top to bottom, compost and soil mix layer followed by a compacted clay layer and a permeable layer for better biogas distribution. ACKNOWLEDGEMENT The authors gratefully acknowledge the CHESF for sponsoring the experimental cell and EMLURB for the usage of Muribecas Landfill area and installation of the cell. The authors also acknowledge CAPES for sponsoring exchange among three Brazilian universities for this research field. REFERENCES
Boeckx, P.; vanCleemput, O.; Villaralvo, I. (1996). Methane emission from a landfill and the methane oxidizing capacity of its covering soil. Soil Biology & Biochemistry 28, 13971405. Borjesson, G.; Chanton, J.; Svensson, B.H., (2001). Methane oxidation in two Swedish landfill covers measured with carbon-13 to carbon-12 isotope ratios. Journal of Environmental Quality 30, 369376. Cabral, A.; Arteaga, K.; Rannaud, D.; Ait-Benichou, S.; Pouet, M.F.; Allaire, S.; Jugnia, L.B.; Greer, C. (2007). Analysis of methane oxidation and dynamics of methanotrophies within a passive methane oxidation barrier. In: Proceedings Sardinia 2007, Eleventh International Waste Management and Landfill Symposium, Cagliari, Itlia, 2007. Czepiel, P.M.; Mosher, B.; Crill, P.M.; Harriss, R.C. (1996). Quantifying the effect of oxidation on landfill methane emissions. Journal of Geophysical Research, vol. 101, nd11, pp. 16.72116.729. 1996. Gebert, J.; Groengroeft, A.; Miehlich, G. (2007). Methane oxidation in a biofilter system influence of water content, temperature and salt concentration. In: Proceedings Sardinia 2007, Eleventh International
These preliminary studies point to a need to evaluate alternatives landfill cover layers that take into account the mitigation of methane emissions to the atmosphere, while allowing no significant increasing of leachate generation. In small and medium municipalities in Brazil, landfill gases are usually thrown into the atmosphere without any treatment, leading to environmental pollution. In addition to that, Brazilian standards do not give much emphasis on the technical aspects of the layers of final cover of landfills. This study has demonstrated that the capillary barrier and methanotrophic layer had the same order of magnitude of biogas emissions. Although compaction degree of methanotrophic layer is lower than capillary barrier, additional characteristics like water content and presence of organic matter, may have worked to improve its performance. Conventional monolithic layers used as a solution in small landfills have methane emissions 10 times larger than the other two kinds of tested cover layers. The presence of the gravel layer below the compacted layer in capillary barrier cover contributes to a more uniform gas distribution minimizing the existence of pressure spots and the maintenance of water content. This fact probably contributed to minimize fugitive emissions in this cover layer. When saturation degree was close to 90% in capillary barrier cover there was no gas emission, although high infiltration rates are allowed. The short time analysis for infiltration allows concluding that for intense precipitation events (above 25 mm accumulated in 3 days), the capillary
1411
Waste Management and Landfill Symposium, Cagliari, Itlia, 2007. Gee, G.W.; Benson, C.H.; Albright, W.H. (2006). Comment on evaluation of evapotranspirative covers for waste containment in arid and semiarid regions in the Southwestern USA. Vadose zone journal 5:809812. Hupe, K.; Heyer, K.-U.; Becker, J.F.; Traores, O.; Noetzel, S.; Stegman. N.R. (2003). Investigations of alternative landfill surface sealing systems in test fields. In: Proceedings Sardinia 2003, Ninth International Waste Management and Landfill Symposium, Cagliari, Itlia, 2003. Maciel, F.J. (2003). Estudo da gerao, percolao e emisso de gases no aterro de resduos slidos da Muribeca/PE. Master Dissertation. Federal University of Pernambuco. Recife-PE. p. 173.
Mariano, M.O.H. (2008). Avaliao da reteno de gases em camadas de cobertura de aterros de resduos slidos. Doctorate Thesis. Federal University of Pernambuco. Recife-PE. p. 243. Morcet, M., Aran, C.; Bogner, J.; Chanton, J.; Spokas, K.; Hebe I.; (2003). Methane mass balance: a review of field results from three French landfill case studies. In: Proceedings Sardinia 2003, Ninth International Waste Management and Landfill Symposium, Cagliari, Itlia, 2003. Tanaka, N.; Matsuto, T. & Lee, H.S. (1997). On-site survey of methane and carbon dioxide flux from MSW landfills in japan, In: Proceedings Sardinia 1997, sI Sixth International Waste Management and Landfill Symposium, vol. 4, pp. 7180 Cagliari, Itlia, 2003.
1412
Evaluation of landfill gas emission in experimental cover layers in Brazil

F.J. Maciel
Department of Civil Engineer, Federal University of Pernambuco, Brazil
R.L. Lopes
Department of Education and Natural Resources Technology, Federal Institute for Education, Science and Technology of Rio Grande do Norte, Brazil
J.F.T. Juc
Department of Civil Engineer, Federal University of Pernambuco, Brazil
ABSTRACT: Landfill gas (LFG) emissions are considered an important parameter in the environmental impact from Municipal Solid Waste (MSW). The purpose of this paper is to present the LFG surface emission investigation conducted in an experimental cell (36,659 t capacity) at the Muribeca Landfill, Brazil. This investigation involved more than 49 static flux chamber tests, associated with soil/gas characterization along the cover depth, in the three different top cover layers: (I) homogeneous layer, (II) mix clay/compost layer (methanotrophic layer), and (III) stratified clay and gravel layer (capillary barrier layer). The studies indicated that the average CH4 flux in the capillary barrier was 15.9 g/m2 d, the methanotrophic layer was 22.2 g/m2 d, and the conventional layer was 161.5 g/m2 d. The lowest flux observed in the capillary barrier allowed to conclude that this design permitted better gas distribution under the clay that influence positively the LFG emission reduction. 1 INTRODUCTION depth, that were performed in three different top cover layers: (I) homogeneous layer, (II) mix clay/ compost layer (methanotrophic layer), and (III) stratified clay and gravel layer (capillary barrier layer). Complementary water infiltration has been also studied in each cover type using infiltrometers and soil suction and water content parameters. 2 GENERAL DESCRIPTION OF THE EXPERIMENTAL COVERS
The Municipal Solid Waste Management (MSWM) situation in Brazil is still precarious, once there are around 4000 open dumps in operation which contribute to the contamination of the environment and the bad life quality of the population. Landfill gas (LFG) emissions are considered an important parameter in the environmental control of a MSW landfill. The LFG liberation to the atmosphere may cause a number of environmental impacts due to the presence of methane, carbon dioxide, among other trace gases in its composition. Organic matter decomposition processes (aerobic and anaerobic) associated with local climate conditions, cover layer characteristics and gas drainage system will determine the amount of pollutants released to the environment. The aim of this paper is to present the LFG surface emission investigation conducted in an experimental cell (36,659 t capacity) at the Muribeca Landfill, Brazil. This cell was constructed to evaluate the landfill gas potential, waste biodegradation and energy recovery. This investigation involved more than 70 static flux chamber tests, associated with soil/gas characterization along the cover
The Experimental Cell dimensions are 65 m 85 m 9 m of thickness and it was filled up with 36,659 t of waste from June/2007 to February/08. The investigation was conducted in the top cover layer of this cell, which is divided in three types of cover as illustrated in Figure 1. Profile 1: methanotrophic layer: layer composed of 0.30 m of compacted soil under a mixture of soil and compost layer. The proportions of the compost are of 50% in the mixture volume, which corresponds to 75% of soil and 25% of compost in weight. The compost used was taken from the compost unit of the landfill. This mix layer varies from 0.40 to 0.75 m and covers an area of 590.2 m .
1413
Metanotrophic layer
0,3-0,45m 0,3m
75% soil + 25% compost Compacted soil
0,1-0,2m 0,3m
Waste
Capillary layer
0,55-0,7m 0,2m
Compacted soil
geotextile
0,3 - 0,55m 0,2m
gravel
Waste
Convencional layer
0,75-0,85m
Compacted layer
0,5-0,65m
Waste
Figure 1.
Cover types of the experimental cell.
Profile 2: Capillary barrier: layer composed of 0.20 m of gravel under a compacted soil layer with total thickness varying from 0.50 to 0.90 m. The superficial area is of 500.3 m . On the interface soil-gravel it was used a geotextile. Profile 3: Conventional layer: compacted sandy clay with thickness varying from 0.50 to 0.80 m with superficial area of 534.8 m . 3 METHODOLOGY
method, gas volume concentration (CH4, CO2 and O2), temperature, and pressure are measured internally in the chamber. The ambient temperature and atmospheric pressure is also monitored. All those measures are taken every 5 minutes during the test. The duration of the test is between 30 to 60 minutes and the gas flux is determined by the initial inclination of the CH4 mass curve versus time. Therefore, the CH4 mass flux rates and coefficient of linear determination (R2) were determined for every test carried out in the three experimental top cover layers. After the chamber removal, a soil sample is taken with a volumetric ring for further analysis in the lab to determine the in situ soil density, and water content. Posteriorly, soil samples were taken from each 0.10 m of the cover by inserting a metallic tube sampler (0.04 m of diameter) until the interface with the solid waste. In each depth it were also measured gas concentration and soil temperature. The soil samples were also used for pH, water content and volatile solids determination in laboratory. This procedure was done in 12 flux chamber tests. The tests were executed in three different periods according to the following distribution: (I) 31 flux tests were carried on the top cover layer between September and December 2008; and (II) 18 flux tests were carried on the slope (lower and upper) and edge of the cell aiming to map and estimate the CH4 emission to the atmosphere. This mapping was analyzed using isoflux curves determined by Kriging interpolation technique.
RESULTS AND DISCUSSIONS
4.1 Soil characterization The soil used in this research was obtained from the borrow area that supplies materials for the cells` cover of Muribeca landfill. The granulometric soil composition is: 23% of clay, 22% of silt, and 53% of sand (32% of fine sand). According to the Unified Soil Classification System (USCS), the soil is classified as sandy clay (SC). The Atterberg limits are: LL 48% and IP 11.2%. An optimum moisture content of 16.5% for a maximum dry density of 16.9 kN/m3 was obtained from the Standard Proctor compaction test (ASTM D698). The water permeability was determined in saturated condition in the Tri-Flex Permeameter, where the value found for w was 2.0108 m/s. Figure 2 presents experimental data taken from the soil-water characteristic curve and a correspondent preview from Gardners model, (1956). Experimental data indicates an air entrance point at approximately 500 kPa. The mixture compost-soil used in the methanotrophic layer contained initially 34.7% of the
The methodology used to evaluate gas emission through the top cover layers consisted of: (I) soil and compost characterization in laboratory; (II) CH4 superficial flux quantification using a static flux chamber and (III) biogas and soil parameters evaluation along the cover in depth until the contact with the waste. For the soil characterization, several tests were carried out such as granulometric analysis, compaction curve, Atterberg limits, water permeability and the soil-water characteristic curve according to the ASTM procedures. The same soil was used in the three experimental cover layers. It is important to mention that physic-chemical analysis of the compost used in the methanotrophic layer included: pH, water content, volatile solids, C, N, P, and K. CH4 and CO2 fluxes through the experimental covers were measured using the static flux chamber methodology described by Maciel (2003). In this
1414
0,45 0,40 Volumetric water content (m3/m3) 0,35 0,30 0,25 0,20 0,15 0,10 0,05 0,00 1 10 100 1000 10000 Matric Succion (kPa)
Gardner Model (1956) Experimental data
Figure 2.
Soil-water characteristic curve.
particles with granulometry range from 0.075 to 0.42 mm (fine sand), 52.4% from 0.42 to 2 mm (medium size sand) and 7.3% with granulometry from 2 to 4.8 mm (thick sand). The fine fraction (0.075 mm) of the compost was below 1.0%. The average volatile solids content was 46.8% ( 1.6%) and the water content was 53.0% ( 1.4%). The initial pH was 7.75, and composition in g/Kg was 15.6 of Ntotal, 2.1 of Ptotal and 3.7 of Ktotal, besides 260 g/Kg of C, presenting a relation C/N of 16.6, which characterizes a stable compost. 4.2 Methane flux
Methane flux through the top cover layers varied from zero to 1332.1 g/m d. This variation is coherent with the values presented in the literature (Boeckx et al., 1996; Czepiel, et al., 1996; Tanaka et al., 1997), and correspond to the dry season of the year (September to December, 2008), and a situation of no forced gas extraction (compressor was turned off during the tests) in the vertical gas wells. The results obtained in this research were superior to those found by Maciel (2003) at Muribeca Landfill (highest of 362.9 g/m d) and Mariano (2008) in the Aguazinha Landfill located in the same metropolitan region (highest of 400 g/m d). That occurred probably due to the fact that gas compressor (forced extraction) was turned off during the tests in the Experimental Cell and the gas wells became underestimated. In Aguazinha and Muribeca Landfills there is no gas extraction (vertical drains open to the atmosphere), although the drainage system was dimensioned for this situation (tube with diameter 0.60 m). It was determined four points with high flux rates (240 g/m d). Two of them were verified in the region close to the vertical gas drains and the other two close to the central pluvial
drainage system. It is possible to conclude that the impermeabilization around the drains region was not efficient to minimize the fugitive gas emissions at the Experimental Cell. The results of CH4 superficial emission investigation, excluding the fluxes verified close to the vertical drains, allowed to obtain the average CH4 flux rate for the three cover layers: 15.9 g/m d (capillary barrier), 22.2 g/m d (methanotrophic layer) and 161.5 g/m d (conventional layer). It is important to analyze the results with the geotechnical parameters to evaluate the layers efficiency, and the main factors that influenced the results. Table 1 presents the range of flux and the average of the superficial cover soil parameters for each cover layer. It is possible to observe that the capillary barrier presented the lowest superficial emission rate where the soil has the highest density and compaction degree. The superficial water content was very similar for all layers. For the conventional cover, fugitive emissions were 10 times superior compared to the others. A positive factor for the capillary barrier efficiency is the presence of the gravel layer as an element to distribute the gases under the cover, minimizing the converging of gas to a certain spot and formation of confined bags of concentrated gas under the cover. This factor may have contributed to minimize the gas emissions in through this cover. The magnitude of the emissions verified on the methanotrophic layer was similar to the capillary barrier. Although soil properties (compaction degree and soil dry density) of the methanotrophic layer were more favorable to emissions then the conventional cover, the main factor that may explain those lower emission rates is the biological factor (oxidative) due to the presence of organic matter (compost) mixed to the soil. 4.3 Methane mapping The investigation of CH4 superficial emissions at the Experimental Cell was completed with the execution of 18 tests on the slopes (lower and upper) and on the edges of the cell aiming to map and estimate the total CH4 flux emitted to the atmosphere.
Table 1. CH4 fugitive emissions and top cover soil parameters.
Range CH4 superficial emission (g/m2.day) 0 - 74.1 0 - 63.4 2.1 - 984.7 Soil average superficial water content (%) 10.8 3.2 10.3 3.9 10.6 3.9 Dry Average density (kN/m2) 13.80 3.2 14.7 1.2 14.2 0.8 Average Compaction Degree (%) 76.80 12.0 86.60 7.1 83.90 4.9
Layer Methanotrophic Capilary barrier Conventional
1415
The procedure adopted to evaluate CH4 emissions was by analyzing the 2D mapping of emissions based on isoflux cuvers. The results of each test were interpolated using the software Surfer and applying the Kriging interpolation. Figure 3 presents the 2D mapping of CH4 superficial emissions at the Experimental Cell. The isoflux curves is represented in units of Nl/m h. It is observed that the superficial CH4 fluxes on the slopes (lower and upper) are higher than the average measured on the top cover and the edges. Between the factors that might have occasioned this behavior, it can be pointed out: (I) difficulty of soil compaction on the slopes using a belt tractor and (II) predominance of the horizontal gas transport inside the solid waste mass related to the vertical flow. It is important to mention that Jain et al. (2005) used a horizontal gas permeability of the waste varying from 1 to 100 times the vertical permeability in tests of injection of air in the landfill. The presence of plastic in the MSW is on of the factors that contribute to the horizontal transport of fluids inside the cell. 4.4 CH4 oxidation parameters In parallel to the flux measurements, tests in depth were carried out to evaluate the CH4 oxidation potential in the cover layers. Between the 12 tests performed in this investigation, 05 presented CH4 concentration under the cover superior to 50% (3 tests on the conventional layer and 2 on the methanotrophic); 05 tests with CH4 15% (2 on the methanotrophic, 1 on the conventional and 2 on the capillary barrier) and 02 tests with values between 15 and 35% (both on the capillary barrier).
It is observed that none of the tests carried on the capillary barrier registered a concentration of CH4 under the cover higher than 35%, which demonstrates the efficiency of the drainage layer (gravel) when distributing the gas under the top cover. This factor contributed directly to the lowest superficial flux of CH4 observed among the experimental covers as the gas percolation through the soil can be regulated by diffusion (Ficks Law). The gas concentration profiles observed in tests allowed verifying a reduction in the CH4/CO2 relation beginning on the interface soil/waste until the surface. In the region around 0.30 m from the surface, this reduction was greater (from the interface to surface this relation felt from 1.3 to 0) on the methanotrophic layer. This variation was from 1.4 to 1.0 at the capillary barrier and from 1.4 to 0.9 on the conventional barrier. That confirms the main oxidation activity in the first 0.30 m of the cover, as constated by Cabral et al. (2007), Huber-Humer et al. (2007) e Berger et al. (2005). The methanotrophic layer has also higher porosity in this range of depth because of the mix soil/compost. That allows the entrance of oxygen, as well as it has higher concentrations of volatile solids, water content in depth, saturation degree, and pH close to neutrality; it all provides a good environment for the methanotrophic microorganisms as cited by Boeckx et al. (1996). Results of soil analysis in depth are as shown on Table 2. The soil temperature profiles indicated a greater variation on the methanotrophic layer with average of 36.1C, which is superior to the capillary barrier (34.7C) and the conventional (33.2C). As the CH4 oxidation is caused by an exothermic reaction, the results permitted to indicate a greater oxidation activity on the methanotrophic layer, followed by the capillary barrier, and finally the conventional. The literature is not conclusive in what concerns the best temperature range of the soil for oxidation activity. In Boeckx et al. (1996) studies, the optimum soil temperature range was between 20C and 30C. Borjesson et al. (2001) suggest that temperatures between 25 to 35C can promote oxidation of 85% of methane. De Visscher et al. (2001) cited by Yoon et al. (2005) verify that greater oxidation rates in temperatures of 35C. Gebert et al. (2007) proves that oxidation rates grow exponentially to temperatures below 38C,
Table 2.
Layer
Average of soil parameters in depth.

T (.C) pH Water Content (%) Volatile Solids (%)
Methanotrophic
36.1 5.1 34.4 1.9 33.2 2.0
7.45 0.55 6.08 0.86 5.74 1.18
20.8 7.5 16.9 0.6 14.9 1.0
7.45 0.55 7.33 1.47 7.43 1.06
Figure 3. CH4 isoflux as fugitive emissions of the Experimental Cell (Maciel, 2009).
Capilary barrier Conventional
1416
although oxidative activity was identified for all the range of temperature analyzed (3C to 45C). The methanotrophic layer presented also an average pH of 7.45 while the other two layers tended to an acid pH (lower than 7 in 75% of the samples). In general, the methanotrophic microorganisms are neutrophilic. Yoon et al. (2005), citing other authors, affirm that the optimum pH to develop bacteria are between 6.7 and 8.0, although other studies have not registered a significative oxidation in pH between 3.5 to 8.0. Cabral et al. (2007) affirm that the optimum pH is 5.5 to 8.5 and Hanson & Hanson (1996) citing Born et al. (1990) observe that the rates of oxidation were similar to soils with pH between 3.5 and 8.0. Although there was no specific study on the influence of pH on the oxidation of CH4 for this soil, the value obtained on the methanotrophic cover was even closer to neutrality. It may be a good indicative of better conditions to the development of methanotrophic bacteria. Relating to the water content in depth, it is observed on Table 2 that the average results along all depth of the methanotrophic layer was 20.8%, which is superior to the capillary barrier (16.9%) and the conventional cover (14.9%). The higher water content in the methanotrophic layer may be cause by the organic matter of the mixture (soilcompost), which is around 2 times superior compared to the other covers. It is important to mention that the accumulated precipitation in the three days before the tests date was no higher than 0.7 mm, so precipitation was not an influence factor to the variations of water content during the dry period. The influence of water content on the maximization of methanotrophic activity has been evaluated in many projects. Teixeira (2008) citing Boeckx et al. (1996) say that the optimization of the oxidative activity occurs to moisture close to half the field capacity of the soil. Teixeira et al. (2009) observed that as high as the saturation degree of the soil is, lower is the counting of methanotrophic in samples collects in the first 0.15 m do the top cover. The content of volatile solids along the depth of the methanotrophic layer (16.3%) was superior to the conventional layer (7.5%) and to the capillary barrier (7.3%). The highest quantity of organic matter is an indicative of better conditions to develop CH4 oxidation activity, because it helps to retain humidity and is favorable to the microbial growth. Borjesson & Svensson (1997) verified that the rates of CH4 oxidation are directly associated to the organic matter in the soil. The presence of compost in the methanotrophic layer was also favorable to the vegetation growth on this layer. The presence of vegetation is a positive factor to
the CH4 oxidation because it provides water and nutrients to bacteria; besides, the roots help the entrance of O2 into the soil. It is important to mention that was not possible to establish a difference in the volatile solids content in the two regions of the methanotrophic layer (upper with mixture soil/compost and lower with compact soil) after a 5 months period as the study carried by Cabral et al. (2007). This happens, possibly, by the leaching of organic matter in the mixture zone (soil/ compost) to the lower layer (homogeneous soil). 5 CONCLUSIONS
Methane is the main gas that contributes to greenhouse effect in landfills. It is usually emitted to the atmosphere through the leakage of the gas collection system or as fugitive emission through the cover layer. In small to medium size landfills in Brazil the cost/benefit to economical exploitation of this gas is very low, which leads to higher superficial emissions to the atmosphere. Brazilian standards do not give much emphasis on the technical aspects of the landfill final covers. The flux of methane found was 15.95 g/m2 d to the capillary barrier, 22.16 g/m2 d in the methanotrophic layer and 161.49 g/m2 d in the conventional layer. The results showed that the lowest average flux rate was obtained in the capillary barrier, where the topsoil was a little more dense and compact and that the magnitude of emissions in methanotrophic layer was similar to the capillary barrier. It is important to mention that the existence of the gravel layer under the capillary barrier cover functioned as a distribution of gas thus minimizing the existence of pressure and gas concentration spots under the layer. This fact may have contributed to minimize fugitive emissions in this cover layer. Emissions from methanotrophic layer of about 7 times less than conventional layer can be attributed to the existence of appropriate environment for methane oxidation. In this study, it was found that methane oxidation was influenced by the moisture content of soil and presence of organic matter, which govern together with pH and temperature, the methanotrophic activity. Increased porosity in the methanotrophic layer allows the oxygen entrance favoring the CH4 oxidation. Analysis of soil parameters (moisture, VS, pH, temperature), composition and flux rates of CH4 in chamber tests permitted to propose, on an experimental basis, a new layer conception which can be constituted by a gravel layer in the bottom, an intermediate homogeneous clay layer, and a soil compost mixture on top. It is important to emphasize the need for transition materials
1417
between those layers in order to minimize negative effects, such as the fissures of the soil surface. This researched allowed to conclude that the fugitive emissions in landfills can be significantly reduced by managing properly the materials of the cover layer and carrying out a complete monitoring investigation (gas, temperature, physical-chemical, and geotechnical parameters soil, compost, among others). This fact is particular important for the small to medium size landfills or opem dumps that is widespread in Brazil. ACKNOWLEDGEMENT The authors gratefully acknowledge the CHESF for sponsoring the project and EMLURB for the participation of all the installation phases of the cell. REFERENCES
Berger, J.; Fornes, L.V.; Ott, C.; Jager, J.; Wawra, B. & Zanke, U. 2005. Methane oxidation in a landfill cover with capillary barrier. Waste Management 25: 369373. Boeckx, P.; Van Cleemput, O. & Villaralvo, I. 1996. Methane emission from a landfill and the methane oxidizing capacity of its covering soil. Soil Biology & Biochemistry 28: 13971405. Borjesson, G. & Svensson, B.H. 1997. Seasonal and diurnal methane emissions from a landfill and their regulation by methane oxidation, Waste Management and Research 15: 3354. Borjesson, G.; Chanton, J. & Svensson, B.H. 2001. Methane oxidation in two swedish landfill covers measured with carbon-13 to carbon-12 isotope ratios. Journal of Environmental Quality 30, 369376. Cabral, A.; Arteaga, K.; Rannaud, D.; Ait-Benichou, S.; Pouet, M.F.; Allaire, S.; Jugnia, L.B. & Greer, C. 2007. Analysis of methane oxidation and dynamics of methanotrophies within a passive methane oxidation barrier, In: Proceedings of the Eleventh International Waste Management and Landfill Symposium, Cagliari, Italy. CD. Czepiel, P.M.; Mosher, B.; Crill, P.M. & Harris, R.C. 1996. Quantifying the effect of oxidation on landfill
methane emissions, Journal of Geophysical Research 101: 16.72116.729. Gebert, J.; Groengroeft, A. & Miehlich, G. 2007. Methane oxidation in a biofilter system influence of water content, temperature and salt concentration. In Proceedings of the Eleventh International Waste Management and Landfill Symposium S. Margherita di Pula. Cagliari. Italy. CD. Hanson, R.S. & Hanson, T.E. 1996. Methanotrophic bacteria. Microbiology Reviews 60 (2): 439471. Huber-Humer, M.; Prantil, R. & Lechner, P. 2007. Improvement of the upper waste layer to foster methane oxidation, In: Proceedings of the Eleventh International Waste Management and Landfill Symposium, Cagliari, Italy. CD. Jain, P.; Powell, J.; Townsend, T.G. & Reinhart, D.R. 2005. Air permeability of waste in a municipal solid waste landfill. Journal of Environmental Engineering 131 (11). Maciel, F.J. 2003. Estudo da gerao, percolao e emisso de gases no aterro de resduos slidos da Muribeca/PE (in portuguese). Master Dissertation. Federal University of Pernambuco. Recife-PE. 173 p. Maciel, F.J. 2009. Gerao de biogs e energia em aterro experimental de resduos slidos urbanos (in portuguese). Doctorate thesis. Federal University of Pernambuco. Recife-PE. 350 p. Mariano, M.O.H. 2008. Avaliao da reteno de gases em camadas de cobertura de aterros de resduos slidos (in portuguese). Doctorate thesis. Federal University of Pernambuco. Recife-PE. 243 p. Tanaka, N., Matsuto, T. & Lee, H.S. 1997. On-site survey of methane and carbn dioxide flux from MSW landfills in Japan, In Proceedings of The Sixth International Landfill Symposium, Sardinia, Italy, vol. 4: 7180. Teixeira, P.F. 2008. Oxidao biolgica de metano em coberturas de resduos slidos urbanos: dinmica do processo e aspectos geotcnicos (in portuguese). Doctorate Thesis. So Paulo State University. Teixeira, C.E.; Torves, J.C.; Finotti, A.R.; Fedrizzi, F.; Marinho, F.A.M. & Teixeira, P.F. 2009. Estudo sobre a oxidao aerbia do metano na cobertura de trs aterros sanitrios no Brasil (in portuguese). Revista Engenharia Sanitria e Ambiental 14 (1): 99 a 108. Yoon, J.I.; Nam, K. & Kim, J.Y. 2005. An innovative cover design using oxygen releasing compounds to reduce methane emissions from solid waste landfills, In: Proceedings of the Tenth International Waste Management and Landfill Symposium, Cagliari, Itlia. CD.
1418
Behaviour of wet fill using residual soil

F.A.M. Marinho & F. Massad
University of So Paulo, So Paulo, Brazil
ABSTRACT: There are situations where compacting the soil above the optimum water content are desired bringing some benefits for the use of compacted soils. This is the case of landfill liners. The compaction of a soil above the optimum leads to a decrease in hydraulic conductivity and induces a greater deformability of the mass (minimizing the formation of cracks). In other situations, such as embankments of roads, dams, etc., the compaction above the optimum water content may not be desired. However, regions or periods where climatic conditions are not favourable, compaction in the wet branch of the curve of compression are the only solution that can meet the deadlines of construction. Compaction above the optimum water content leads to a lower dry unit weight and to a higher degree of saturation. The combination of low density and high degree of saturation may lead to the development of undesirable pore water pressure during construction. At the end of construction of an embankment the pore water pressure development may induce stability problems. Compacted soils are unsaturated and when they are compacted at the optimum water content the soil matrix suction varies between 50 kPa to 500 kPa. The initial soil suction of a compacted layer varies with the increase of total stress induced by construction. The pore water pressure, initially negative, decreases and may reach a point where it becomes zero or positive. With the nowadays ability to directly measure matrix soil suction by using high capacity tensiometers, it is possible to determine the path followed by the suction during loading. Results obtained using a residual soil of gneiss indicated that the negative pore water pressure, generated during the compaction of a soil layer, reduces with the increase of total stress but can remain negative up to considerable equivalent heights. This behaviour was observed even if the soil is compacted above optimum water content. 1 INTRODUCTION ment within the embankment. Obviously embankments placed over compressible soil are subjected to others mechanism of failure, and this should be considered. The soil and geological/geotechnical conditions of the project govern the drainage of water during the loading. The boundary condition and the soil permeability will control the drainage characteristics. Drainage condition can be defined as follow: DRAINED conditionis the situation in which ground water can flow into or out of the mass of soil at the time that the soil is subjected to changes in loading. At drained condition changes in load do not cause changes in pore water pressure. UNDRAINED conditionis the situation in which the water has no time to leave or enter the soil mass when the soil is subject to a change in stress. At undrained condition the change in stress will cause variations in pore water pressure. The drainage condition depends on the soil type associated with the rate of loading. Compacted soils, even when compacted at the wet side of the compaction curve, present negative pore water pressure. The presence of suction increases the factor of safety in terms of slope
Compaction above the optimum water content induce a lower degree of compaction and a higher degree of saturation. The combination of low density and high degree of saturation may lead to the development of unwelcome pore water pressure during the construction of embankments. At the end of construction of an embankment the pore water pressure should not reach values that may create an unstable condition for the embankment and its foundation. However, the undesired pore water pressure development may occur already during the construction period, leading to instability even before it reaches the final height. The end of construction analysis is generally performed in new embankments where: a clay soils is used. the embankment is constructed over a soil with high compressibility. the soil is compacted at optimum water content or above it. when the embankment can reach saturation by loading (compacted layers) or by rain. The mechanism of failure of an embankment can be activated by pore water pressure develop-
1419
stability. Vaughan et al. (1978) mentioned that the generation of high pore-water pressure depends on the structure of the compacted material and it is important to assess the condition of the structure in the field. This can be inferred by monitoring the soil suction in the field during construction. There are many situations where the soil from the borrow pit used for the embankment presents a high water content, requiring a great effort to reduce the water content to values close to the optimum water content. The preliminary tests presented in this paper aim to contribute to implement procedures to construct embankments using soil with a water content well above the optimum water content. 2 WET COMPACTION AND THE SHEAR STRENGTH
Wet compaction is a condition where the soil is compacted well above the optimum water content. Massad et al. (1978) presented a study for Rio Verde dam, constructed in Paran State, Brazil. Due to the climatic condition of the site wet compaction was used. Studies were developed at that time, with the objective of defining how the shear strength of the soil varied with loading/drainage condition. Triaxial tests UU and CU were conducted. The first case simulates the condition and loading where there is no time for consolidation or drainage during shearing. The CU tests on the other hand simulate the condition where there would be time for consolidation, but the subsequent load is placed very fast do not allowing drainage during shear. Two soils were tested by Massad et al. (1978), one was a sandy clay (Ip 27%; wl 65%) and the other a sand-clay silt (Ip 25%; wl 56%). Figure 1 presents the results obtained by Massad et al. (1978) for the sand-clay silt. Figure 1a presents the results of the UU tests and Figure 1b the data related to the CU tests. It was found that shear strength under UU condition varied with both the compaction water content and with dry density. For the CU condition the compaction water content did not show significant influence on the shear strength. Similar results were also obtained by Casagrande & Hirschfeld (1960). For the same w and compaction degree Figure 1 also shows UU shear strength greater than the corresponding CU value. Massad et al. (1978) made a conjecture, at that time without a proof, that suction developed in the UU specimens was responsible for this behaviour. And that the end of construction may not be the most critical situation in terms of embankment stability. The reduction in void ratio induced by the CU tests created a structure that generated more porewater pressure at failure.
To better define the variations found by Massad et al. (1978) Figure 2 presents an illustration of how the resistance should vary at the wet side of the compaction curve according to the type of test performed. Please note that this trend may vary with the soil tested. For an accelerated construction process, the higher the compaction water content the lower the shear strength at failure. The investigation of the development of pore water pressure during loading was studied by Casagrande and Hirschfeld (1960). They suggested three tests, all at constant water content, to investigate this subject: (a) specimen subjected to increasing hydrostatic confining pressure, called PH test; (b) specimen subjected to constant 13, called PN test; and (c) specimen loaded in a consolidation apparatus, PC test. The soil tested by Casagrande and Hirschfeld (1960) was a clay (Ip 15% and wl 34%) with the colloidal fraction consisting largely of montmorillonite type clay minerals. The authors presented results showing that for specimens close to saturation (S 85%) the development of pore water pressure had two responses. For low pressures the relation between pore-water pressure and applied stress was slightly flatter than 45o,
100
600 kPa 500 kPa
Confining pressure 100 kPa UU Test S = 100% Compaction degree(%)
100
500 kPa
Confining pressure 100 kPa CU Test S = 100%
Compaction degree(%)
95
400 kPa
95
400 kPa
Shear Strength
90 0 2 4
300 kPa
300 kPa
90
Shear Strength
0 2 4 6 8
w(%) Condition where each compacted layer cannot consolidate after the placement of the following layer.
w(%) Condition where each compacted layer can consolidate after the placement of the following layer.
UU Test
CU Test (b)
(a)
Figure 1. Triaxial test results for Rio Verde dam. (Modified from Massad et al. 1978).
UU Test
Effect on shear strength
CU Test
Effect on shear strength
Compaction Degree
100 % Compaction Degree
100 %
INCREASE
INCREASE
DECREASE
DECREASE
0%
w (a)
0%
w (b)
Figure 2. Schematic diagram for the change in shear strength according to the test performed.
1420
and for high stress the mentioned relation follow a 45 degree line, suggesting an almost saturated behaviour. Casagrande and Hirschfeld (1960) did not measure the suction of the specimens. The results presented in this paper are only from PH tests. 3 THE ROLE OF SUCTION
Throughout the compaction curve the water content varies and as a consequence the suction also varies. Figure 3a shows a compaction curve where point A represent situations before the optimum water content, point B at optimum water content and point C above it. In Figure 3b is shown the variation of suction with water content (call Soil Water Retention Curve), where it is observed that the lower the water content the greater the suction. The arrows indicate the path followed during a constant water content loading for each point. The rate that the pore water pressure reaches the zero depends on the initial degree of saturation and structure of the soil. In another words it depends on the hydraulic conductivity of the soil. Vaughan (1982) presents considerations on wet compaction and illustrates the effect of the compaction water content associated with the load using the representation shown in Figure 4. In this figure it is shown the initial negative pore-water pressure (suction) and its variation with the increase in total stress. It is observed that the suction decreases with increasing total stress. The point where the suction is zero and positive pore-water pressure start to develop depend on the initial value of suction. The same phenomenon was observed by Sandroni (1985) and Lins and Sandroni (1994). Associated with this variation of total stress and suction Vaughan (1982) also presents an outline of how the shear strength varies for three different initial conditions. For the condition where the suction is high the total stress increment is not sufficient to reduce the suction to zero. For lower suction the increase in total stress can reduce the suction to zero, inducing a 0 condition at low total stress.
Figure 4. Response of the pore-water pressure and shear strength to the total stress applied (Vaughan, 1982).
Marinho et al. (2003) presented the porewater pressure development for the same residual soil used for this paper. The results presented by Marinho et al. (op. cited) focused on specimens compacted at optimum or below optimum water content, and showed that the positive pore-water pressure development occurred only for confining pressure higher than 200 kPa. 4 MATERIAL AND METHOD
Compaction Degree
100 %
B C A
A B C
0% (a)
w (b)
Suco
Figure 3. Compaction curve and the soil water retention curve.
The soil tested is a residual soil of gneiss from the city of So Paulo. Some of the geotechnical characteristic of the soil is presented in Table 1. In order to evaluate the development of porewater pressure due to increase in total stress specimens statically compacted below the optimum, at optimum and above optimum water content were tested. The tests were performed under constant water content. Table 2 presents the characteristic of the specimens tested. Figure 5 presents the compaction curve for the soil. It is also shown the initial conditions of the statically compacted specimens. It can be seen that the points are slightly below the compaction curve. Although very rigid control was used to compact the specimen along the compaction curve a small difference occurred. This difference does not affect the conclusions obtained from the tests performed. Soil water retention curves (SWRC) obtained from specimens compacted at optimum water content, below optimum and above optimum, are presented in Figure 6. The SWRC was obtained using three methods as shown in Figure 6. For specimens tested for this paper the suction was measured using a high capacity tensiometer (HCT) and they are indicated in Figure 6.
1421
Table 1.
Geotechnical characteristics of the material.

100
Suction Plate
Pressure Plate
W1 W3 W4 W2 O1 D1
Filter Paper
Particle density (g/cm3) Medium sand content (%) Fine sand content (%) Silt content (%) Clay content (% 2 m) Liquid limit (%) Plasticity limit (%) Plasticity index (%) Optimum water content (%) Maximum dry unit weight (kN/m3) Table 2. Specimen D1 O1 W1 W2 W3 W4 Characteristics of the specimens tested. wc (%) 22.8 24.8 28.0 27.5 27.7 31.1 S (%) 76 84 90 88 89 87
2.71 7 27 33 33 47 34 13 25 15.0
90 80 70 60
S (%)
50 40 30 20 10 0 1 10 100 1000 10000

Static Compaction Optimum W/C Dry of Optimum] Wet of Optimum
Suco (kPa)
Initial suction (kPa) 206 108 42 40 38 14
Figure 6. Soil water retention curve and state of the specimens tested.
15.1 14.9
O1 D1
Dry Unit Weight (kN/m3)
14.7
W3
14.5 14.3 14.1 13.9 13.7
W2 W1
After compaction the soil suction is measured using the HCT. Following equilibration of suction the triaxial cell was filled with water and confining pressure is applied under undrained condition to air and water. Increments of confining pressure were applied at intervals of 5 minutes. In some tests the maximum confining pressure was left for some time and the pore water pressure was monitored. 5 RESULTS AND COMMENTS
W4
13.5 17
19
21
23
25
27
29
31
33
Water Content (%)
Figure 5. Compaction curve and initial condition of the specimen tested.
A triaxial cell equipped with a high capacity tensiometer (HCT) installed at the base was used. All tests, except test W4, were performed using a top cap fitted with a dry low air entry pressure porous material. This top cap is connected to a valve that may be used to control the air drainage. It should be called attention to the fact that there is a certain amount of air connected to the sample. This air is at atmospheric pressure at the beginning of the test. In order to compare the effect of this set up, a blind top cap was used during test W4. The test to infer the pore-water pressure generation due to increase in isotropic pressure is performed according to the following procedure:
As it is well know the increase of total stress under undrained condition generates positive increments of pore water pressure. If the soil is saturated the total stress increment is completely transmitted to the pore water. However, if the soil is under a degree of saturation below 100% it is partially transmitted to the soil pore water. It is expected that during loading the air should dissolve and the soil eventually reaches saturation. If the soil is close to saturation the relation between the total stress applied and the pore water pressure response is close to one. Figure 7 presents the development of pore water pressure due to the increase of isotropic pressure. An equivalent height, assuming a unit weight of 20 kN/m3, is also indicated as a reference. Results of six tests are shown in Figure 7. It can be seen the effect of the initial suction (or degree of saturation) on the pore water pressure response. The results shown in Figure 7 are particularly interesting because it shows the negative initial suction before any confining pressure is applied. The use of the HCT allowed the monitoring of the development of the pore water pressure, from the negative value to easily measured positive figures.
1422
Equivalent Height (m)

400 300 200 100 0 100 200 300 0 10 20 30 40 W1 W2 W3 W4 O1 D1 50
200
400
600
800
1000
Confining Pressure (kPa)

Figure 7. Response of the pore water pressure to confining pressure.
From Figure 7 some interesting behaviour can be observed: The specimen compacted at the dry side of the compaction curve started with a suction of approximately 200 kPa, and the specimen compacted 6% above optimum still presented a negative pore water pressure of 14 kPa. The results indicated a downward concavity at the beginning of the test, more specifically while the pore water pressure was negative. The confining pressure at which the pore water pressure reached zero is directly proportional to the initial suction. After the suction is reduced to zero the curve shown a upward concavity. When the maximum confining pressure was left for more than 30 minutes it was observed a movement of water toward the top cap. The presence of air along the system, from the top cap to the valve, seems to be responsible for that behaviour. The test performed using a blind top cap allowed to verify that behaviour. Figure 8 presents the results of two tests, W1 and W3. The difference between the procedure used for these two tests are only the blind top cap used for test W1. From the results shown in Figure 7 it can be seen that the presence of the blind top cap has no effect on the short term behaviour of the pore water pressure development. When the maximum confining pressure is left for some time, the pore water pressure is reduced on the test performed with an open top cap. In Figure 8 the response of
the pore water pressure for the two tests shows the differences between the two procedures used. The importance of this behaviour is related to the pore water pressure dissipation in the field. If there is a possibility for the air to dissipate, the reduction of the positive pore water pressure developed during the construction process is much quicker. 6 CONCLUSIONS
The possibility of performing tests to infer the ability of the soil to keep the pore water pressure below zero is important to make viable the development of embankments using soil with a water content well above the optimum water content. The wet compaction is viable provided tests are performed in order to investigate the development of pore water pressure during construction. Using the high capacity tensiometer it was possible to infer the development of pore water pressure for a compacted residual soil starting from the as compacted condition and hence recording the initial suction. The confining pressure above which the pore water pressure starts to become positive was obtained. From the preliminary set of tests performed it was shown that the residual soil can be compacted with a water content of 3% above the optimum water content and still present negative pore water pressure up to a confining pressure of approximately 50 kPa. For compaction water content of
1423
250 200
W1 W3
Blind Top cap
Pressure (kPa)
150 100 50 0
50 100
Top cap with air drainage
(a)
0 10 20 30 40 50 60 70 80 90
Time (min)
700
Confining Pressure (kPa)
600 500 400 300 200 100 0

100
W1 W3
(b)
0 10 20 30 40 50 60 70 80 90
Time (min)
Figure 8. Effect of the air drainage on the long term behaviour. Marinho, F.A.M., Standing, J.R. and Kuwagima, R.M. (2003). Soil suction development under isotropic loading and unloading in a compacted residual soil. Solos e RochasRevista Brasileira de Geotecnia., v.26, pp. 115128. Massad, Rocha, J.L.R. e Barros, J.M.C. (1978). Caractersticas geotcnicas de solos de aterros midos, compactados sobre fundao compressvel. XII Seminrio Nacional de Grandes Barragens, So Paulo (In Portuguese). Sandroni, S.S. (1985). Estimativa de poro-presses positivas em macios de terra compactada durante a fase de construo. XVI Seminrio Nacional de Grandes Barragens. Belo Horizonte. pp. 697717 (In Portuguese). Vaughan, P.R. (1982). Design and Construction With Wet Fills. Special Lecture presented at the Brazilian Society for Soil Mechanics. So Paulo- SP. Vaughan, P.R., Hight, D.W., Sodha, V.G. and Walbancke, H.J. (1978). Factors controlling the stability of clay fills in Britain. Clay Fills. Institution of Civil Engineers. pp. 205217.
6% above optimum the soil remains with some suction even when the confining pressure is about 15 kPa. These findings confirm that the end of construction may not lead to the most critical situation in terms of embankment stability. ACKNOWLEDGEMENT The first author is grateful to FAPESP and CNPq for the financial support. REFERENCES
Casagrande and Hirschfeld (1960). Stress-deformation and strength characteristics of a clay compacted to a constant dry unit weight. Research Conference on Shear Strength of Cohesive Soils. ASCE. pp. 359417. Lins, A.H.P. and Sandroni, S.S. (1994). The development of pore-water pressure in a compacted soil. XII ICSMFE, New Delhi. pp. 177180.
1424
Unsaturated behaviour of mine tailings in low precipitation areas

L.A. Oldecop, L. Garino & J.J. Muoz R. Rodriguez C. Garca
Instituto de Investigaciones Antissmicas (IDIA), Universidad Nacional de San Juan, San Juan, Argentina Departamento de Recursos geolgicos, Instituto Geolgico y Minero de Espaa (IGME), Madrid, Spain Departamento de Ingeniera Geolgica y Cartogrfica, Universidad Politcnica de Cartagena, Murcia, Spain
ABSTRACT: Tailing deposits are singular geotechnical structures, which are intended to last indefinitely in time. One of the most important challenges in relation to such waste deposits is to predict their behaviour in the long term, e.g. several centuries. In this context, it seems useful to observe the behaviour of old abandoned tailings deposits. Since tailings are very fine grained materials, the unsaturated phenomena play a significant role in the hydraulic and mechanical behaviour of the deposit. In this paper, some examples of old abandoned tailing deposits are examined, located in low precipitation regions. The in situ measured water content distributions show that, the degree of saturation remains relatively high, even though the low rain amounts and high evaporation rates. It is concluded that the evolution of water content should be taken into account in the deposit design process. 1 INTRODUCTION management of the decant pond. Moreover, the severity of the consequences of failure is also related to tailings water content. Catastrophic mud avalanches developed only in high water content tailings or when special circumstances determined that large amounts water entered the deposit. Mine tailings are stored with an initial amount of moisture which depends mainly on the processing technology used, i.e., conventional slurry, paste or filtered tailings. The deposition method, management techniques of the deposit and the climate are determinant factors for evolution of moisture distribution within the deposit. In tailing deposits located in arid regions a large portion of the stored material tends to develop an unsaturated condition. Since tailings are very fine grained materials, the unsaturated phenomena play a significant role in the hydraulic and mechanical behaviour of the deposit. In this paper, some examples of old abandoned tailing deposits are examined from the point of view of their geotechnical behaviour. It is believed that useful lessons can be obtained from the study of these geotechnical structures, although they are significantly smaller in size than those which are currently being built. Particular attention is given to the unsaturated phenomena and their role in the stability and general performance of the deposits.
Tailings deposits are singular geotechnical structures. At least theoretically, they are intended to last indefinitely in time, a fact which is very unusual from the engineering point of view. Moreover tailing deposits which are being built at present times have extraordinary large volumes, exceeding all previous experiences in civil engineering fills of any kind. Many of these large tailings deposits are being built as part of metalliferous mining projects in arid regions of South America, mainly along the Andes range in Per, Chile and Argentina. One of the most important challenges, in relation to these types of waste deposits, is to predict their behaviour in the long term, maybe several centuries or thousands of years. In this context, it seems useful to observe the behaviour of old abandoned tailings deposits, in order to gain knowledge about their long term behaviour and the evolution trends of their geotechnical properties. Failure case histories (Chandler & Tosatti, 1995; Bligth, 1997; Fourie et al, 2001; Blight & Fourie, 2003) show that one of the most significant variables regarding the deposit stability is the water content of the stored tailings. Most failures of tailing dams (ICOLD, 2001) were related to rainfall, insufficient drainage provisions or improper
1425
STUDY OF OLD ABANDONED TAILING DEPOSITS
Figure 1 displays the geographical location of the mining zones and the distribution of the studied deposits in each zone. The cases studied, a total of nine tailing deposits, are located in two different regions: the Castao Viejo mining district in the central Andes mountain range, Argentina and Sierra de CartagenaLa Unin, in Murcia, Spain. The studied tailing deposits were inactive for periods raging 15 to 45 years. No capping neither other reclamation works where done upon the abandonment and
hence, they experienced free interaction with the environment, since the end of mining operations. The Sierra de Cartagena- La Unin mining zone has about 50 Km2 and was exploited since the roman presence in the Iberian Peninsula (III bC) for the extraction of iron, lead and zinc. The latest period of mining activity ranged from 1940 to 1991. Since the 90s all mining works were abandoned and most tailing deposits were left as they were. Eighty nine tailing deposits were identified with a mean height of 10.5 m. The mining operations in the Castao Viejo mining district were devoted to the extraction of lead, zinc and copper. The most intense exploitation
Figure 1. Geographical location of the two mining zones and distribution of the studied tailing deposits. a) Sierra de Cartagena-La Unin, Murcia, Spain. b) Castao Viejo mining district, province of San Juan, Argentina.
1426
Depth [m]
period lasted from 1956 to 1964 and since then all mining activity halted. Twenty four small tailing deposits exist in the zone with heights ranging 5 to 15 meters. In both mining zones, tailings were transported and deposited in the traditional way, i.e pumped or gravitational flow of slurry from the extraction plant to the deposit and discharge by mobile spigots. Upstream construction method was used with pond decantation. The climate in both regions is of the arid type. Figure 2 shows mean monthly precipitation records and the mean monthly evapotranspiration (ETP) computed by the Thornwaite method (TM) and the evaporation pan method (PEM), based on actual evaporation measurements (Smajstrla et al, 2000). Both environments determine high water-deficit conditions. In Cartagena, the total mean annual precipitation is 339 mm and the total annual ETP is 904 mm computed with TM, and 724 mm when computed with PEM. In Castao the water-deficit is even harsher, with a total mean annual precipitation of 265 mm and ETP of 1276 mm computed with PEM. TM method yield lower ETP values for the Castao region but it should be noted that the TM usually under-predicts the ETP in very dry climates.
FIELD EXPLORATION DATA
The selected tailing deposits were studied in the field. Borings were done from the top of nine tailing deposits, logging the usual geotechnical parameters. Though the rather dry climates in which they are located, significant amounts of water were encountered in all the studied tailings. Figure 3 and 4 display the obtained profiles of degree of saturation. The degree of saturation in the Cartagena deposits has low values at surface, but they rapidly increase in
Degree of saturation (S) 0,00 0 2 4 6 8 10 0,20 0,40 0,60 0,80 1,00
Dep 1 Dep 2 Dep3 Dep 4 Dep 5
a) 200
180 160 140 120 100 80 60 40 20 0 Jan
Rainfall ETP (Thornwaite method) ETP (Evaporation pan method)
12 14
Dep 6 Dep 7
mm/month
Figure 3. Profiles of degree obtained in Cartagena tailing deposits (Filled symbols: non-vegetated deposits; Hollow symbols: vegetated deposits).
May. Mar. Aug. Sep. Nov. Nov Apr. Oct. Dec. Dec Feb. Jun. Jul.
Month
b) 200
180 160 140 120 100 80 60 40 20 0 Jan
Rainfall ETP (Thornwaite method) ETP (Evaporation pan method)
mm/month
May
Mar
Apr
Aug
Month
Figure 2. Climatic data. Rainfall and evapotranspiration. a) Cartagena, Spain. b) Castao, Argentina.
Sep
Feb
Jun
Oct
Jul
Figure 4. Castao N 8 tailings deposit. Boring log and degree of saturation profile.
1427
the first 23 meters depth. From there on, S becomes higher than 0.7 (in all cases except one). It interesting to note that the lower S profiles (0.7) correspond to vegetated deposits (in Fig. 3, Deposit 3: shrubs, Deposits 5 and 6: trees). The highest values of S, almost approaching 1 (though the phreatic level was not reached), occur in non-vegetated deposits and in those deposits where ponding of rain water occurs in the top (Deposits 1 and 7). The profiles of degree of saturation obtained in the Castao tailing deposits are shown in Figures 4 and 5. The general tendency is similar
to the observed in the data of Cartagena, with S increasing with depth almost up to 1. However, no phreatic water was encountered within the depth of the boreholes. At site N 8, the phreatic level was located at 1.5 meters below the base of the deposit, in alluvial gravels. Moreover, deposit N 8 contains some coarse grained layers (sand) and in correspondence with them, the degree of saturation falls. A suction profile corresponding to the boring in N 8 deposit is shown in Figure 6. The situation displayed corresponds to a strong evaporation condition. However between 1.5 and 2 meters depth a downward flow front can be identified, probably associated to a past (recent?) rain event. 4 STABILITY ISSUES
Figure 5. Castao N 9 tailings deposit. Boring log and degree of saturation profile.
Suction [MPa]
0,01 0,00 -1,00 -2,00 0,1 1 10 100
-3,00
-4,00 -5,00 -6,00
-7,00
Suction
Hydrostatic line
Figure 6. Suction profile corresponding to borehole N 8 compared to the hydrostatic line.
Most of the studied tailing deposits do not display stability issues, in spite of their steep angled external slopes. The construction method involved excavation and re-filling (with or without compaction) of the coarse fraction (sand) of previously discharged tailings, allowing them to air-dry first. The resulting slope angle was always close to the equilibrium angle atrest. So most of these deposits have slopes with angles between 40 and 45. Moreover, some of them suffered the effects of hydraulic erosion, which produced some vertical cuts, as the one shown in Figure 7, corresponding to deposit N 8 in Castao. In this case, a 7 meters high, nearly vertical cut remains stable, mainly due to the effect of matric suction and probably also by some cementation effect due to precipitated salts. On the other hand, in some few cases in Cartagena, the signs of slope failures occurred in the past could be identified. The most likely triggering mechanism for these failures is the increase in water content of tailings, due to infiltration of rain water. Signs of water ponding in the top of the deposit and platforms can be observed just in these cases. Several relicts of the slumped tailings can be seen in the direction of sliding reaching some 100 m away from the slope toe. In one case (Figure 8) the slumped mass crossed a small stream channel and penetrated a nearby forest. These facts suggest that the observed failures were of the flow type. The previous observations lead to the conclusion that the stability of these tailing deposits mainly depends on maintaining an unsaturated condition. But the presented field data also suggest that it is not necessary that a very dry condition develops, for ensuring enough strength in order to produce stable slopes. In fact Rodriguez (2002) showed that, in tailings materials, the highest apparent cohesion due to matric suction develops for saturation degrees around the air entry value condition (i.e., S 0.8). However, it is worth to remember that this strength is instantly lost when the degree of saturation approaches 1.
Depth [m]
1428
intervention occurred during those relatively long periods of time, could yield interesting data about how the deposit interacts with the environment. Some basic assumptions were made for the analysis: The initial time for the analysis is the moment of abandonment. The initial water volume contained in the profile is computed by assuming that the whole profile was initially saturated (S 1). This is reasonable since the height of the deposits is moderate (less than 14 m) in relation with the material air entry value (estimated in, at least, 200 kPa). Current water volume in the profiles are computed on the basis of the measured water contents. Consolidation and under-drainage has all already occurred during the construction period. Infiltration is computed by the curve number method SCS (1957), for those deposits in which the geometry allows runoff. When the geometry of the deposits causes rain water to pond in the top, all the precipitation value is assumed to infiltrate in the deposit.
Figure 7. Nearly vertical cut produced by hydraulic erosion in N 8 tailings deposit, in Castao.
40 slopes
am
The water balance is performed by the following equation: Vw Vw 0 Ia T ETPa T (1)
Slumped mass
Forest
50 m
Figure 8. Plant view of El Descargador (N 4) tailings deposit (Cartagena), showing the scarps of four slope failures and the extent of the slumped masses of tailings.
LONG TERM WATER BALANCE
In order to get an assessment of the long term behaviour of the tailings deposits, a water balance was performed for all the nine cases studied, ranging the whole period in which they remained abandoned. It seems that making use of the fact that no human
where Vw current water volume in the whole profile, Vw 0 initial water volume in the profile, Ia mean annual infiltration, ETPa mean annual evapotranspiration, T time period elapsed between the moment of abandonment and borehole execution, fraction of mean annual ETP to be applied in order to make the balance to close. is taken as the variable of the analysis, measuring how effective was the ETP as the driving force of the tailings drying process. The analysis results are presented in table 1. Since TM and PEM yield similar ETP values in the case of Cartagena, only the former was applied. Table 1 results show that real evapotranspiration (ETP) mean rates in bare deposits of Cartagena are around 40% the predicted ones with the Thornwaite method. The ratios are a bit higher (4555%) for vegetated deposits, because plants enhance soil moisture extraction. This was already noted when observing the degree of saturation profiles. In the tailings deposits of Castao, the results based on the TM are similar to Cartagena results while the PEM results indicate that the overestimation of ETP may be even higher in more arid climates. The analysis performed shows that, in all cases, in the long term, the ETP is overpredicted by either of the applied methods. The most likely reason for this is that in a water-deficit climate the drying
St re
1429
Table 1. DEP
Water balance for a 1 m2 tailings column. Vw0 (litres) Ia (mm) ETPa (mm) T (years) 15,0 15,0 15,0 15,0 15,0 15,0 15,0 Vw (litres) 4110,2 3711,5 3361,3 1281,0 2100,9 5015,6 2777,9 1455.5 2718.0 1455.5 2718.0 ETPreal / ETPteor 0,40 0,42 0,45 0,40 0,55 0,53 0,41 0.34 0.41 0.18 0.21 Observations Ponding/No vegetation Runof/No vegetation Ponding/Vegetation (shrub) Runof/No vegetation Runof/Vegetation (trees) Ponding/Vegetacin (trees) Ponding/No vegetation Runof/No vegetation Ponding/Vegetation (shrub) Runof/No vegetation Ponding/Vegetation (shrub)
Cartagena, Spain (Thornwaite Method) 1 4520,0 339,0 923,3 2 4551,5 329,0 923,3 3 4521,0 339,0 923,3 4 1890,0 329,0 923,3 5 4780,0 329,0 923,3 6 7262,5 339,0 923,3 7 3379,0 339,0 923,3
Castao, Argentina (Thornwaite Method) 8 2568,5 205.8 673.6 45,0 9 3103,4 264.7 673.6 45,0 Castao, Argentina (Evaporation PAN Method) 8 2568,5 205.8 1276.1 45,0 9 3103,4 264.7 1276.1 45,0
process of tailings involves not only stage I drying (Gowing et al, 1990), in which the evaporation front is at the soil surface, but also stages II and III, controlled respectively by capillary rise and vapour diffusion from the evaporation front to the soil surface. This effect is apparent in the saturation degree profiles. An additional factor slowing the drying process could be the transport of salts due to capillary rise and their precipitation at the deposit surface (Newson & Fahey, 2003). 6 CONCLUSIONS
safety of the deposit and hence its evolution should be taken into account in the deposit design process and monitoring programs. REFERENCES
Bligth, G.E. (1997). Destructive mudflows as a consequence of tailing dyke failures. Proc. Instn. Civ. Engng, 125, 918. Blight, G.E. y A.B. Fourie (2003). A review of catastrophic flow failures of deposits of mine waste and municipal refuse. Int. Workshop on Occurrence and Mechanisms of Flow in Natural Slopes and Earthfills, Sorrento, Italia. Chandler, R.J y Tosatti, G. (1995). The Stava dams failure. Italy, July, 1985. Proc. Instn. Civ. Engng. 113, 6779. Fourie, A.B., G.E. Blight y G. Papageorgiou (2001). Static liquefaction as a possible explanation for the Merriespruit tailings dam failure. Can. Geotech. J, 38, 707719. Gowing, J.W., Konukcu, F. & Rose, D.A. 2006. Evaporative flux from a shallow watertable: The influence of a vapourliquid phase transition. Journal of Hydrology 321, 7789. ICOLD (2001). Tailings dams. Risk of dangerous occurrences. Lessons learnt from practical experiences. Bulletin 121, UNEP. Newson, T.A. & M. Fahey (2003). Measurement of evaporation from saline tailings storages. Engineering Geology, 70, 217233. SCS, 1957. Hydrology National Engineering Handbook. USDA Soil Conservation Service. Smajstrla, A.G., Zazueta, F.S., Clark, G.A. & Pitts, D.J. (2000). Irrigation Scheduling with Evaporation Pans. Bulletin 254. Department of Agricultural and Biological Engineering, Florida Cooperative Extension Service, Institute of Food and Agricultural Sciences, University of Florida.
Except in a shallow superficial layer, the saturation degree is relatively high in the studied deposits. It seems a general fact that, no matter how dry is the climate, the loss of water of the tailings deposit will not continue indefinitely, but rather reach a stable state. For depths greater than a couple of meters, the drying process will not go on further than a certain point lying between saturation and the air entry value condition (0.7 S 1). Capillary water produces an apparent cohesion which ensures the stability of the studied deposits. Hence stability relays on maintaining the unsaturated state. Regarding the design of modern tailing deposits, a desirable and realistic design target would be, not to expect indefinite drying of tailings, but to reach a suction value (and the corresponding saturation degree) near to the air entry value. All capping and reclamation works should help to maximize runoff/infiltration ratio in order to avoid recharge of water. Water content of tailings seems to be a fundamental variable for the long term behaviour and
1430
Evaluation of a capillary barrier at the Rocky Mountain Arsenal

L.O. Williams D.L. Hoyt
US Environmental Protection Agency, Denver, USA Pacific Western Technologies, Denver, USA
G.A. Hargreaves S.F. Dwyer
US Environmental Protection Agency, Denver, USA Dwyer Engineering, Albuquerque, USA
J.G. Zornberg
The University of Texas at Austin, Austin, USA
ABSTRACT: A monitoring program is currently being implemented to assess the performance of a capillary barrier constructed at the Rocky Mountain Arsenal, a high-profile hazardous waste site located northeast of Denver, Colorado, USA. Monitoring includes regulatory-required pan lysimeters to measure basal percolation as well as water content reflectometer probes to measure soil moisture profiles. The moisture content profiles allowed continued evaluation of the cover performance in response to natural precipitation and irrigation. The results to date indicate that the comparatively low density of the cover soils, selected to aid in vegetation establishment, appears to have been detrimental to the overall cover performance. 1 INTRODUCTION The RMA site, which is regulated under the US Environmental Protection Agencys (USEPA) Superfund program, was once considered to contain the most contaminated square mile on earth (Frumkin 2005). A primary remedy component at this site involves in-situ consolidation of contaminated soils to be covered with a total of six unsaturated soil covers. These six covers, which span over 160 ha of RMA, are deemed alternative covers, as they were required to be equivalent to a prescriptive RCRA-Subtitle C cover (FWENC 1996). Based on the site-specific conditions and studies available at the time (e.g. Melchior 1997), a quantitative threshold percolation of 1.3 mm/year was adopted in 1998 for the RMA alternative covers (RVO 1998). The first RMA alternative cover consisting of a capillary barrier was constructed in 2007 over an 8.5 ha area in central RMA. Known as the Shell Cover, it includes a 1.22 m-thick soil layer underlain by a capillary break composed of a nonwoven geotextile placed over coarse gravel (chokestone layer). A comparatively low relative compaction (ranging from 75% to 85% of the maximum dry density from Standard Proctor tests) was specified for the cover soils in order to promote vegetation growth.
Unsaturated soil covers [e.g. evapotranspirative (ET) covers] have been selected and constructed in multiple solid waste, mining, and hazardous waste containment facilities around the world. However, the selection of parameters governing the infiltration of moisture into ET cover systems is still controversial. Specifically, soil parameters that minimize the downward moisture infiltration are often compromised by soil parameters that maximize vegetation and upward moisture movement due to transpiration. Critical parameters that designers must balance include the soil type and its placement conditions, primarily the soil density (e.g. Zornberg et al. 2003). An increased soil density leads to a decreased hydraulic conductivity that slows infiltration. However, the increased soil density may also limit the type and quality of vegetation. In order to provide insight into the performance of unsaturated soil covers, this paper presents the results of a monitoring program implemented at the Rocky Mountain Arsenal (RMA), a highprofile hazardous waste site located northeast of Denver, Colorado, USA.
1431
The post-construction monitoring program for the Shell Cover includes regulatory-required gravity lysimeters to measure basal percolation as well as water content reflectometers (WCRs) to measure moisture within the cover soils. Interpretation of the moisture content profiles within the Shell Cover is presented herein to evaluate the covers performance in response to natural precipitation and irrigation. 2 2.1 BACKGROUND INFORMATION Description of the capillary barrier system
Vegetative surface
Barrier soil with organic amendments Barrier soil Nonwoven geotextile Chokestone Biota barrier Foundation soil Waste
1.219 m
0.457 m varies
The climate in Denver is semiarid, with an average annual precipitation of 396 mm and an average pan evaporation of 1,394 mm (as quantified for the 1948 to 1998 period). The wettest months of the year (April to October) are also the months with the highest pan evaporation, which makes the RMA site well suited for the use of ET cover systems. The RMA remedy required a field demonstration to prove equivalence of the alternative design with a prescriptive cover before design and construction of the final alternative covers. The equivalence demonstration at RMA involved a field demonstration complemented with comparative numerical analyses (Kiel et al. 2002). Four ET test covers, approximately 9.14 m by 15.24 m, were constructed on a rolling plain at RMA in the summer of 1998. Data collected for these test plots included basal percolation, precipitation, moisture content and overland runoff. Basal percolation was collected in pan lysimeters, which involved a geocomposite drainage layer underlain by geomembrane. Precipitation was monitored using an all-season rain gauge. Surface water was collected in polyethylene geomembrane swales constructed around the cover perimeters. WCRs were used to measure volumetric moisture content profiles (RVO 1998). Success of the equivalence demonstration was initially based solely on lysimeter data, which provided a measurement that could be compared directly against the 1.3 mm/year criterion. (Kiel et al. 2002). With this approach, the lysimeter measurements for all test plots at RMA satisfied the quantitative percolation criterion over the demonstration period of 19982003. However, subsequent scrutiny of the WCR data clearly indicated that the design criterion had been achieved because a capillary break had developed within the constructed test plots due to the presence of a geocomposite drainage layer below the soil. The final design for the Shell Cover involved a capillary barrier (Figure 1). As shown in the figure, the cover includes three components: a chokestone (gravel) layer, a geotextile, and a soil layer. A nominal 0.457-m thick biota barrier, consisting of crushed high-durability concrete obtained from runways at an adjacent airport demolition site, was
Figure 1. Capillary barrier at the Shell Cover at the RMA, Denver, Colorado, USA.
constructed over the contaminated soils and foundation soils. The biota barrier is designed to prevent burrowing animals from digging into the waste. The capillary break component of the cover, which was constructed directly over the biota barrier, includes a 2.54 cm-thick chokestone layer (coarse gravel with a maximum particle size of 19 mm). The capillary break component also includes a nonwoven geotextile that also helps minimize the migration of soil particles into the chokestone layer. The ET component of the cover involved a 1.22 m-thick soil layer, placed at a density ranging from 75% to 85% in relation to the maximum dry density according to Standard Proctor maximum tests. Soil amendment was added to the top 0.3 m of the soil layer to facilitate vegetation. The cover was revegetated with native grasses characteristic of a short-grass prairie and compatible with the habitat of the surrounding National Wildlife Refuge. 2.2 Description of the monitoring program
The Shell Cover was irrigated after seeding to promote initial establishment of the vegetation. The post-construction monitoring program included three gravity lysimeters to measure basal percolation. Each lysimeter included nests of WCR probes, installed to measure the real-time moisture profiles within the soil. The data evaluated in this paper includes soil moisture content, soil temperature, percolation, precipitation, and irrigation collected from July 9, 2007 to Dec. 31 2008. Lysimeters 1, 2, and 3 are located within the Shell Cover at the northern downslope edge, upslope portion, and southern downslope edge, respectively (Figure 2). Each lysimeter includes a total of five nests with eight moisture sensors, as shown in Figure 3a. Nest 1 is located outside the lysimeter area (to the right of the lysimeter when facing downslope). Nests 2, 3, and 4 are located inside the lysimeter area, towards the downslope portion, central portion, and upslope portion of the lysimeter, respectively. Finally, Nest 5 is located
1432
was uniform, as shown in the figure by an approximately linear cumulative irrigation. The daily and cumulative precipitation recorded at the site for years 2007 and 2008 are also shown in Figure 4. A total cumulative precipitation of 359 mm was received at the site in 2007 (164 mm in the 3rd and 4th Quarters), while a total cumulative precipitation of only 211 mm was received in 2008. The steep portions of the cumulative precipitation curve correspond to rain or snow events, while the flat portions of the cumulative precipitation curve correspond to dry periods. It is of particular interest that the cumulative irrigation and precipitation on the Shell Cover in 2007 was 738 mm, nearly double the average annual precipitation for Denver (396 mm).
Figure 2. Lysimeter locations in Shell Cover.
3.2 Temperature Figure 5 shows the temperature data collected at the site from July 2007 to December 2008. This data is consistent with expectations as it shows that surficial sensor T1 registers the highest temperature among all probes during the summer, while showing the lowest temperature among all probes during the winter. As expected, the deepest sensor T6 has the lowest temperature among all probes during the summer, while it shows the highest temperature among all probes during the winter. Of particular relevance is the assessment of the sensors that record below freezing
Daily Precipitation & Irrigation (mm)
Daily Rainfall Daily Irrigation C. Rainfall C. Irrigation
35 30 25 20 15 10 5 0
1/1/07 4/2/07 7/2/07 10/1/07 12/31/07
350 300 250 200 150 100 50
outside the lysimeter area (to the left of the lysimeter when facing downslope). The location of the eight probes in each nest is illustrated in Figure 3b. Probes 1 and 2 (duplicates) are located 0.152 m below ground surface. Probe 3 is located 0.356 m below ground surface. On the other hand, Probes 4, 5, and 6 are located 0.66 m, 0.457 m, and 0.254 m above the geotextile. Finally, Probes 7 and 8 (duplicates) are located 0.051 m above the geotextile. In cases where the cover thickness exceeds 1.219 m (minimum cover thickness), the distance between Probes 4 and 5 was increased. Six temperature sensors were installed in the Shell Cover at depths corresponding to the locations of the moisture sensors. 3 3.1 ANALYSIS OF WEATHER DATA Irrigation and precipitation
3/31/08
6/30/08
9/29/08
12/29/08
Figure 4. irrigation.
Daily and cumulative precipitation and
The Shell Cover was irrigated from July 2, 2007 to September 15, 2007 as shown in Figure 4. The average daily irrigation was 7 mm/day with a cumulative irrigation total of 379 mm. The irrigation rate
Figure 5. records.
Precipitation, irrigation and temperature
1433
Cumulative Precipitation & Irrigation (mm)
Figure 3. Distribution of nests: (a) Plan view of lysimeter 3, (b) Probes in lysimeter 3-03.
40
400
temperatures (0C). As shown in the figure, sensors T1 and T2 reach freezing temperatures in the first winter (December 19, 2007 to February 22, 2008) as well as during the second winter (since December 16, 2008). None of the deeper temperature sensors reached freezing temperature. The multiple temperature probe depths are critical for interpreting the WCR data because freezing temperature in the cover soils affects moisture readings in at least three aspects, as follows: The dielectric constant of ice is significantly lower than the dielectric constant of liquid water. Consequently, the WCRs do not detect water in ice form. The WCR readings drop significantly when the temperature falls below freezing in response to the change in water phase and not to a real moisture decrease. Freezing of the ground surface shuts down upward water migration from beneath the frozen depth. Consequently, moisture stored within the soil cover at the time of surface freezing can only migrate downwards or remain unchanged throughout the frozen ground period. Downward infiltration may take place once thawing (of either accumulated snow or of ice within the ground surface) begins. 3.3 Analysis of Overall Weather Conditions
Figure 6. Volumetric moisture content time series.
Figure 5 also shows the combined precipitation, irrigation, and temperature records. This data shows a sequence of events over the 20072008 winter that likely contributed to a detrimental performance of the cover. Specifically, this information shows: An initial period of heavy irrigation just after construction, which contributed to the high moisture content throughout the cover; A subsequent period of rain and snow that may also have contributed to increased infiltration into the cover (or at least prevented its recovery); and A final period of frozen ground surface, which may have prevented evaporation (i.e. cover recovery) throughout the winter season, even if the 2007 winter was comparatively dry.
conducted using the volumetric moisture content values corrected for temperature. The layout of the probes in Nest 303 and the color codes are shown in Figure 3b. Figure 6 also includes the cumulative precipitation and infiltration values as a reference. Inspection of the data presented in Figure 6 reveals very good, essentially identical, agreement between the readings from duplicate Probes 1 and 2, as expected since they are located at the same depth. Also, Probes 1 and 2 show sharp moisture changes over short periods of time. These moisture fluctuations are expected of locations near the ground surface. Probes 1 and 2 show an increase in moisture in response to precipitation, which is particularly clear for the summer 2008 precipitation events. The readings from duplicate Probes 7 and 8 also show good agreement, particularly before summer 2008. A discrepancy in the magnitude of these readings is observed after summer 2008 (approximately 15%). This discrepancy is attributed to local heterogeneities and to the high sensitivity of the calibration curves of WCR for high moisture content values. Readings from Probes 7 and 8 show smooth changes with time. This type of response is expected of locations towards the bottom of the cover. It should be noted that Probes 7 and 8 show continuously high values of moisture from July 2007 to mid-May 2008 (exceeding 40%). This response has important implications, as follows: Increasing moisture content at the bottom of the cover indicates that capillary action may be developing; particularly after mid-September 2007, when the bottom probes show increasing moisture content while the remaining probes show approximately constant values. It is speculated that the fluctuations are due to deficiencies in the temperature correction calibrations (note that the fluctuations are countercyclical with the yearly temperature fluctuations). The significant increase in moisture content at the bottom of the cover, detected soon after the initiation of the irrigation period, suggests a poor performance of the ET component of the cover. That is, the infiltration moisture fronts reach the bottom of the cover quickly, which is
ANALYSIS OF MOISTURE DATA
Nest 3 of Lysimeter 3 (Nest 303), located at the center of Lysimeter 3 (Figure 3a), was selected as baseline for subsequent comparison and is the only one presented due to paper size constraints. 4.1 Evaluation of Time Series
The volumetric moisture content time series for Nest 303 is shown in Figure 6. All analyses were
1434
not desirable. A well-performing ET component would store and subsequently release moisture without triggering the development of the capillary barrier unless the season is exceptionally wet. It is acknowledged, though, that the Shell Cover was irrigated to encourage germination of grass seeds and vegetation was not expected to be fullyestablished during the initial monitoring period. A total percolation of 0.795 m3 was collected during the monitoring period in Lysimeter 3 (possibly soon after the irrigation period). This percolation is consistent with the fact that the moisture front reached the bottom of the cover. Even though breakthrough did occur, the potential development of the capillary break likely minimized the amount of percolation. Probes 3, 4, 5, and 6 in Nest 3-03 are the intermediate probes from top to bottom within the nest. Moisture content data shows that the moisture content increases from the top to the bottom of the cover. These probes show marked increases in moisture in September 2007, becoming progressively smoother from the top probe to the bottom probe which is consistent with the expected response. This increase in moisture is attributed to the advancing moisture front induced by the heavy irrigation. As shown in Figure 5, freezing temperatures reached the depth corresponding to Probe 3 (0.356 m below ground). Freezing of the ground surface can explain some of the trends observed in the data collected from early January 2008 to late February 2008. Specifically, the decrease in moisture content shown by Probes 1 to 3 during ground freezing occurs because WCRs do not account for the ice fraction of the volumetric moisture content. Also, the sudden increase in moisture in early January in most of the probes may be due to a downward moisture migration triggered by shutdown of ET after freezing of the ground surface. The likely sequence of events that triggered moisture changes during June 2007 to July 2008 is: Breakthrough may have occurred in late September 2007. A sudden increase of moisture can be observed in Probes 4, 5, and 6 around this time, as well as a smooth increase in the moisture of Probes 7 and 8 (lower probes).
Subsequent to the breakthrough, there is some indication from Probes 4, 5, and 6 that the soil may have started drying out towards the end of 2007. The recovery trend of Probes 4, 5, and 6 is halted when the ground surface becomes frozen in late December, with moisture readings remaining approximately constant until late March 2008 in spite of a comparatively dry winter. Once the ground surface temperature increased beyond freezing values in early March 2008, the anomalous drop in moisture in the surficial probes is no longer observed. Indeed, Probes 1 and 2 initiate a trend of decreasing moisture (from early March to July 2008). This trend is only altered by a moisture increase (late May 2008) that is triggered by precipitation events. Recovery of the cover initiated with decreasing moisture in Probes 1, 2 and 3 in early March, followed by Probes 4, 5, and 6 in mid-April 2008, and the lower Probes 7 and 8 in early June 2008. Recovery continued towards the end of 2008. Probes 3, 4, 5, and 6 reached comparatively low moisture content by July 2008 (12 to 15%, which is consistent with the moisture content of the cover before irrigation in July 2007). This response indicates that the cover can recover after the dry summer season. A yearly cycle in which the bulk of the cover reaches comparatively low moisture content (e.g. 10 to 12%) is expected to provide evidence of the capability of the cover to recover from a wet season. 4.2 Evaluation of moisture profiles
The volumetric moisture content profiles for Nest 303 are shown in Figure 7. The moisture profiles shown in the figures correspond to the average moisture contents for time periods of approximately 2 days, 2 weeks, and 2 months, respectively. The moisture profiles, averaged every 2 days during the initial 12 days, show an initial moisture profile of reasonably uniform moisture content of approximately 12% in the middle portion of the cover (Figure 7a). The initial moisture profile shows higher moisture content towards the surface (less than 20%), which is likely due to the cover irrigation. Also, the initial moisture profile shows comparatively
Figure 7.
Average moisture content profiles during: (a) the initial 12 days; (b) the initial 12 weeks; (c) the initial year.
1435
high moisture towards the base of the cover. It is possible that the moisture content at the base of the cover was already high when irrigation started, or that moisture at the base increased rapidly during irrigation due to preferential flow mechanisms. As shown in the figure, a moisture front is observed to advance from the ground surface into the cover. The moisture profiles, averaged every 2 weeks during the initial 12 weeks, show the continued advance of the moisture front from the ground surface into the cover (Figure 7b); reaching the base of the cover approximately eight weeks after irrigation began. It is of concern that the moisture front reaches the base of the cover so quickly. While a low soil cover density was selected to promote vegetation growth, it is clear that the low soil density led to high hydraulic conductivity and, consequently, fast infiltration. Available soil hydraulic conductivity data indicates that a slightly higher relative compaction could have lowered hydraulic conductivity values by a factor of ten with a resulting ten-fold decrease in the moisture front velocity. The capillary break appears to develop when the moisture front reaches the base of the cover. This is shown by a continued increase of moisture, now reflected from the base of the cover upwards. That is, the moisture profile bulges at the cover base once it is reached by the moisture front. The moisture profiles, averaged every 2 months during the initial 12 months, indicate the further development of the capillary break (Figure 7c). It should be noted that the initial moisture profile (JulySeptember 2007) corresponds to the average of the profiles shown in Figure 7a. The four moisture profiles ranging from September 2007 to July 2008show a significant increase in moisture throughout the entire profile. The moisture content increases with depth, with particularly high moisture values at the bottom of the cover. After breakthrough, the profiles remain unchanged at a high moisture value and without showing signs of cover recovery for approximately 8 months. The last moisture profile shows that the cover has begun to recover, as indicated by a comparatively low moisture content throughout the cover system, with the sole exception of the bottom probes (Probes 7 and 8). 5 CONCLUSIONS
front, potential development of a capillary break, and potential recovery of the unsaturated soil barrier are evident in the moisture content profiles. A capillary barrier effect may have been triggered once the moisture front reached the bottom of the cover. While this indicates potentially good performance of the capillary barrier component of the system, it also indicates a poor performance of the ET component. While the specified low soil cover density was intended to aid vegetation growth, it also led to a fast infiltration into the cover. The fast infiltration (only 8 weeks to reach the base of the cover) is significantly shorter than the approximately 6 month-long dry seasons in the Denver area. Ground freezing delayed recovery of the cover, despite a comparatively dry winter season. The additional, temporary infiltration barrier induced by ground freezing affected the cover performance detrimentally by preventing ET. Overall, it is recommended that the design density of the covers soils be selected to ensure that the soil layer prevents a moisture front from reaching the cover base before the time period between dry seasons. Also, alternative covers proposed for areas where seasonal temperatures drop below freezing should acknowledge the significant decrease in evaporation that takes place during ground freezing. Despite the potential difficulties triggered by the low soil cover density and by conducting irrigation immediately before winter, moisture profiles appear to indicate that the Shell Cover has begun to recover, and is expected to continue. REFERENCES
Foster Wheeler Environmental Corporation (FWENC). 1996. Record of Decision for the On-Post Operable Unit. June. Frumkin, H. (2005). Environmental Health: From Global to Local. Jossey-Bass, 1108 pp. Kiel, R.E., Chadwick, D.G., Lowrey, J., Mackey, C.V. & Greer, L.M. (2002). Design of evapotranspirative (ET) covers at the Rocky Mountain Arsenal. Proceedings: SWANA 6th Annual Landfill Symposium. Melchior, S. (1997). In-situ studies of the performance of landfill caps (Compacted clay liners, geomembranes, geosynthetic clay liner, capillary barriers). Land Contamination and Reclamation. 5(3), 209216. RVO (1998). Final RCRA-Equivalent Cover Demonstration Project Comparative Analysis and Field Demonstration Design Scope of Work. RVO (2009). Soil Cover Moisture Monitoring System, Data Evaluation Summary, October 2007 to December 2008. Zornberg, J.G., LaFountain, L. & Caldwell, J.A. (2003). Analysis and Design of Evapotranspirative Cover for Hazardous Waste Landfill. Journal of Geotechnical and Geoenvironmental Engineering, ASCE, 129(5), 427438.
The moisture content profiles provided key information for evaluating important aspects of the Shell Cover performance at RMA. The following moisture migration mechanisms were identified using moisture data collected during 18 months of monitoring: A front of increasing moisture content advancing from the top of the cover towards the bottom was observed. The speed of the advancing moisture
1436
Numerical study on rainfall infiltration into landfill earthen final cover

Tony L.T. Zhan, G.W. Jia, Y.M. Chen & D.G. Fredlund
MOE Key Laboratory of Soft Soils and Geoenvironmental Engineering, Zhejiang University, Hangzhou, China
ABSTRACT: The infiltration rate at which a given soil can absorb rainfall is highly influenced by rainfall patterns. One-dimensional numerical simulations of infiltration process into landfill earthen final covers were conducted to investigate the effects of rainfall patterns on the infiltration rate under different initial conditions. The numerical simulation model conducted by SVFlux was firstly verified by experimental data. The results indicate that the peak rainfall intensity occurs earlier, the cumulative infiltration will be larger no matter what the initial condition is. The advanced pattern (A1) generates the largest amount of infiltration under certain conditions. The time to ponding increases as the initial matric suction increases. The rainstorm with short duration will generate less infiltration compared to the rainfall patterns lasting the whole day. The numerical analyses suggest that rainfall pattern should be taken into account in the calculation of infiltration into earthen final covers. 1 INTRODUCTION infiltration subjected to different rainfall patterns. The empirical rainfall-runoff models can not calculate the infiltration into the earthen final covers accurately according to rainfall patterns. Numerical simulations are conducted to evaluate the performance of the earthen final covers in the cover design. The commercial software used in numerical simulations includes SEEP/W, HYDRUS, UNSAT-H, Vadose/W and SVFlux (Scanlon 2002). It is difficult to simulate the interactions between the ground surface and the atmosphere, especially the rainfall infiltration rate. For SEEP/W, it is assumed that if the rainfall intensity is greater than the saturated permeability, there will be ponding and the pore water pressure at the ground surface is zero. It can not simulate the interactions because there is no ponding in the initial period according to the observation. For SVFlux, if the pore water pressure at the surface is negative, all the rainfall will infiltrate into the ground, if the pore water pressure at the surface reaches zero, the infiltration flux will be equal to the saturated permeability when the rainfall intensity is higher than the saturated permeability (SoilVision 2006). The infiltration will be computed in an interactive manner, which can reveal the real infiltration process. Few studies have been conducted to investigate the influences of rainfall patterns on the infiltration. Usually, the rainfall intensity is assumed to be uniform or central triangular in one day in VADOSE/W (GEO-SLOPE 2004). The objectives of the study are to investigate the effects of rainfall patterns on the infiltration process and total
Landfill earthen final covers are at present widely used to cap the waste disposal in the world, such as US, Germany and Canada. The landfill earthen final cover uses plants and soil to produce an effective, economical capping system. It is designed to work with forces of nature rather than attempting to control water flow with barrier layers (e.g. compacted clay liner and geomembrane). It uses two natural processes to control infiltration of water into the waste. Firstly, the soil stores the infiltrating water. Secondly, the evapotranspiration removes the water from the water reservoir. Therefore, the amounts of infiltrating water highly influence the performance of earthen final covers. The infiltration rate is primarily controlled by rainfall characteristics (e.g. rainfall intensity and rainfall patterns), soil properties (e.g. the saturated permeability and soil-water characteristic curves) and the initial conditions. When the rainfall intensity is smaller than the saturated permeability, all the rainfall will infiltrate into the soil. When the rainfall intensity is greater than the saturated permeability, as time elapsed, the infiltrated water will increase the water content at the ground surface and then the rainfall intensity will be greater than the infiltration capacity and runoff will be generated. Normally, the infiltration rate under constant rainfall intensity is estimated by empirical models (e.g. Green-Ampt 1911, Horton 1938, Philip 1957, Holtan 1961, and Mein-Larson 1973). However, little attempt has been conducted to investigate the
1437
amount of infiltration through series of numerical simulations. 2 2.1 THE NUMERICAL MODEL AND ITS VERIFICATION The geometry and boundary condition of the Numerical Model
with sand was chosen as the soil used to construct the earthen final cover in the paper. Figure 2 shows the soil-water characteristic curve of silt with sand. Figure 3 shows the soil permeability of silt with sand. The soil-water characteristic curve function was defined using van Genuchten equation (van Genuchten 1980): 1 r ( s r ) n ( ) 1
11 / n
Volumetric water content/%
One dimensional numerical analysis through SVFlux was conducted to investigate the effects of rainfall patterns on the infiltration process and total amount of infiltration. The thickness of 1.5 m was chosen as the vertical depth of the numerical model (see Figure 1) according to the Alternative Cover Assessment Project (ACAP) report (Albright et al 2004). The initial condition is designated initial pore water pressure. The upper boundary can be described as follows: if the pore water pressure at the ground surface is negative, the infiltration is equal to the rainfall intensity; when the pore water pressure at the ground surface reaches zero, the upper boundary condition switches from flux boundary to zero hydraulic head boundary. The lower boundary condition of the numerical model is unit gradient boundary. The unit gradient boundary means that theres only gravity gradient acting at the lower boundary and velocity of flow is equal to the soil permeability at the lower boundary. 2.2 Soil properties and rainfall patterns
where r is the residual water content, s is the saturated water content, and and n are empirical parameters.The unsaturated hydraulic conductivity (K) was defined by the van Genuchten-Mualem equation:
n / n 1 11 / n K ( ) = K s ) 1 (1 2
50 40 30 20 10 0 10-1
= 0.00176/cm, n=1.29, s=0.45r=0
(silt with sand, Roesler et al (2002))
Roesler et al. (2002) summarized the hydraulic properties of soils used in the ACAP project. Silt
100
101
102
103
104
105
106
Matric suction/kPa
Figure 2. sand.
10-3 10-6
Permeability /cm/s
The soil-water characteristic curve of silt with
10-9 10-12 10-15 10-18 10-21 10-1 100 101 102 103 104 105 106
= 0.00176/cm, n=1.29, ks=1.4 x 10-5cm/s
(silt with sand, Roesler et al (2002))
Matric suction /kPa
Figure 1. The geometry and boundary condition of the numerical model.
Figure 3.
The permeability curve of silt with sand.
1438
In which is the relative saturation, which is defined as

Infiltration(t) /mmmin-1
1.0 0.8 0.6 0.4 0.2 0.0

i-t Zhu (2006) Experimental Data i-t SVFlux Simulation Results i-t Zhu (2006) Simulation Results
r s r
Figure 4 shows six representative rainfall patterns of 24 h duration used in the previous literature. They consist of four basic types of rainfall patterns: advanced type (A1 and A2), centralpeaked type (C), delayed type (D1 and D2), and uniform type (U) (Ng et al. 2001; Tsai 2008). The advanced patterns have relatively high rainfall intensity during early part of the rainfall event, whereas the delayed type is just the opposite. The central-peaked pattern has relatively high rainfall intensity in the center part, while the uniform type reveals constant rainfall intensity throughout the rainfall duration. A1, C and D1 are adopted in the analysis. Rainstorm is also a common rainfall pattern. Usually, rainfall storms last for several hours and there are several rainstorms in one day depending on the weather. 2.3 Verification of the numerical model
50
100 150 Time /min
200
Figure 5. Infiltration curves of the column tests and numerical simulations.
2.4
The program of parametric study
Series one-dimensional infiltration tests were conducted by Zhu and Chiu to investigate the infiltration pattern under constant rainfall intensity greater than the saturated permeability of soil (Zhu 2006, Chiu 2009). The test results were then compared with the numerical simulations obtained by solving Richards equation. SVFlux was used to simulate the column infiltration tests, and the numerical results were shown in Figure 5. It shows that the numerical results obtained by Zhu and SVFlux are identical, so SVFlux can be employed to do the numerical analysis under different rainfall patterns.
The saturated permeability of silt with sand is 1.4 105 cm/s (i.e. 0.012 m/d). In the analysis, the total amount of rainfall in one day is chosen as 0.024 m (i.e. 2 times of the saturated permeability) and 0.048 m (i.e. 4 times of the saturated permeability). The total amount of rainfall will distribute according to the four typical rainfall patterns and rainstorms. The initial matric suction designated to the one dimensional soil column is 40 kPa, 100 kPa and 200 kPa. 3 INTERPRETATION OF THE NUMERICAL SIMULATION RESULTS
3.1 The influence of rainfall pattern on infiltration process The infiltration rate is primarily controlled by rainfall characteristics (e.g. rainfall intensity and rainfall patterns), soil properties (e.g. the saturated permeability and soil-water characteristic curves) and the initial conditions. Figure 6 shows the rainfall intensity and the corresponding infiltration curve under four representative rainfall patterns. It shows that the infiltration curves consist of two stages: (1) before ponding stage; (2) after ponding stage. Before ponding, all the rainfall will be absorbed into the soil surface (i.e. the infiltration rate is equal to the rainfall intensity). The time to ponding (tp) increases as the initial matric suction increases no matter what the rainfall pattern is. After ponding, the infiltration rate decreases with time and approaches to the saturated permeability at last. The saturated permeability is the minimum infiltration capacity. After ponding (i.e. ground surface is saturated), though the rainfall supply is sufficient, the infiltration rate is controlled by the matric suction gradient at the ground
1.2 1.0 0.8
q/qmax
0.6 0.4 0.2 0.0 0 4 8

U A1 C D1 A2 D2
Time /h
12
16
20
24
Figure 4.
Representative rainfall patterns.
1439
surface. As time elapsed, more water infiltrate into the soil then reduce the matric suction gradient, therefore, the infiltration capacity reduces. For A1 and C rainfall pattern, as time elapsed when the infiltration capacity reduces to the rainfall intensity, the ponding vanishes and all the rainfall will be absorbed by the soil. 3.2 The influence of rainfall pattern on cumulative infiltration
infiltration between the A1 and D1 pattern is up to 0.0035 m (7.3% of the total amount) when the initial matric suction is 40 kPa. Series of numerical simulation suggests that difference of infiltration rate between rainfall patterns are between 7.3% and 17.4%, so it can not be ignored in the analysis. Therefore, the rainfall pattern should be taken into account to calculate accurate infiltration and then the percolation. 3.3 Infiltration process of rainstorms The rainstorm last for 4 hours (i.e. 1/6 day). The rainfall intensity is 0.144 m/d, which is 12 times of the saturated permeability. The total amount of rainfall is 0.024 m, and the corresponding uniform rainfall intensity is 0.024 m/d. Figure 8 shows the rainfall intensity and the corresponding infiltration curve under different initial matric suction (i.e. 40 kPa, 100 kPa and 200 kPa). The time to ponding increases with initial matric suction increases. Since the rainfall intensity is quite high, the duration needed to reach surface saturation is short. After ponding, the infiltration capacity reduces quite quickly. The reducing rate of infiltration with time reduces with time. 3.4 The cumulative infiltration of rainstorms Figure 9 shows the cumulative infiltration curves of D1 pattern and rainstorm. The initial matric suction for the soil column is 40 kPa, 100 kPa and 200 kPa, respectively. Based on the previous analysis, we know that D1 pattern will generate the minimum infiltration. Figure 9 shows that the rainstorms will generate less infiltration under the same initial condition compared to the D1 pattern. It can be concluded that the high intensity rainfall with short duration is not the adverse weather for
Rainfall intensity and infiltration rate / m/d 0.16 0.14 0.12 0.10 0.08 0.06 0.04 0.02 0.00 0.0 0.2 0.4
1 Storm rainfall 1 Storm (-40kPa) 1 Storm (-100kPa) 1 Storm (-200kPa)
Figure 7 shows the cumulative infiltration curves under 4 representative rainfall patterns with daily rainfall of 0.048 m. The initial matric suction for the soil column is 40 kPa, 100 kPa and 200 kPa, respectively. The numerical simulation results indicate that for the total amount of infiltration, A1 pattern U pattern C pattern D1 pattern, which suggests that the earlier the peak rainfall intensity occurs, the greater the total amount of infiltration is. The cumulative of infiltration of A1 pattern is close to the uniform pattern. The difference of infiltration between the A1 and D1 pattern is up to 0.00835 m (17.4% of the total amount) when the initial matric suction is 200 kPa. The difference of infiltration between the A1 and D1 pattern is up to 0.00651 m (13.5% of the total amount) when the initial matric suction is 100 kPa. The difference of
0.10 0.08
U rainfall pattern A1 rainfall pattern C rainfall pattern D1 rainfall pattern U (-40kPa) A1 (-40kPa) C (-40kPa) D1 (-40kPa) U (-100kPa) A1 (-100kPa) C (-100kPa) D1 (-100kPa) U (-200kPa) A1 (-200kPa) C (-200kPa) D1 (-200kPa)
Infiltration rate /m/d
0.06 0.04 0.02
ksat
0.00 0.0 0.2 0.4 0.6 0.8 1.0
Time /d
Figure 6. Rainfall intensity and the corresponding infiltration curves.

0.05
A1 rainfall pattern C rainfall pattern D1 rainfall pattern U rainfall pattern
Cumulative Infiltration /m
0.04 0.03 0.02 0.01 0.00 0.0
U (-40kPa) A1 (-40kPa) C (-40kPa) D1 (-40kPa) U (-100kPa) A1 (-100kPa) C (-100kPa) D1 (-100kPa) U (-200kPa) A1 (-200kPa) C (-200kPa) D1 (-200kPa)
ksat
0.6 0.8 1.0
0.2
0.4
0.6
0.8
1.0
Time /d
Time /d
Figure 7. The cumulative infiltration curves of different rainfall patterns.
Figure 8. Rainfall intensity and the corresponding infiltration curves of rainstorm.
1440
0.025 Cumulative infiltration /m 0.020 0.015 0.010 0.005 0.000 0.0

D1 (-40kPa) D1 (-100kPa) D1 (-200kPa) 4 hours storm (-40kPa) 4 hours storm (-100kPa) 4 hours storm (-200kPa)
by the National Natural Science Foundation of China (NSFC). REFERENCES

Albright, W., Benson, C., Gee G., Roesler, A., Abichou T., Apiwantragoon P., Lyles B. and Rock S. 2004. Field water balance of landfill final covers. J. Environ. Qual. 33 (6): 23172332. Chiu, A.C.F., Zhu, W. and Chen, X.D. 2009. Rainfall infiltration pattern in unsaturated clayey silt. Journal of Hydrologic Engineering, American Society of Civil Engineers, 14(8), 882886. GEO-SLOPE International Ltd. 2004. Seepage modelling with SEEP/W, users guide version 6.16. GEOSLOPE International Ltd., Calgary, Alta. Green, W.H. and Ampt, C.A. 1911. Studies on soil physics: flow of air and water through soils. J Agric Sci 4:124 Holtan, H.N. 1961. A concept of infiltration estimates in watershed engineering. ARS4151, U.S. Department of Agricultural Service, Washington, DC. Horton, R.I. 1938. The interpretation and application of runoff plot experiments with reference to soil erosion problems. Soil Science Society of America Proceedings. 3: 340349. Mein, R.G. and Larson C.L. 1973. Modeling infiltration during a steady rain. Water Resour Res. 9(2):384394. Ng, C.W.W., Wang B. and Tung, Y.K. 2001. Threedimensional numerical investigations of groundwater responses in an unsaturated slope subjected to various rainfall patterns. Canadian Geotechnical Journal. 38(5): 10491062. Philip, J.R. 1957. The theory of infiltration: 1 The infiltration equation and its solution. Soil Sci., 83(5), 345357. Roesle,r A.C., Benson C.H. and Albright W.H. 2002. Field Hydrology and Model Predictions for Final Covers in the Alternative Assessment Program-2002. Geo-Engineering Report No. 0208, University of Wisconsin, Madison, WI. 279 pp. [Online]. Available at: http://www.acap.dri.edu/ Scanlon, B.R., Christman M., Reedy, R.C., Porro I., Simunek, J. and Flerchinger, G.N. 2002. Intercode comparisons for simulating water balance of surficial sediments in semiarid regions, Water Resour. Res. 38(12): 13231339. SoilVision Systems Ltd. 2006. SVFlux Users Manual. Saskatoon, Saskatchewan, Canada. Tsai, T.L. 2008. The influence of rainstorm pattern on shallow landslide. Environ Geol. 53:15631569. van Genuchten M. Th. 1980. A closed form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Science Society of America Journal. 44 (5): 892898. Zhu, W., Chen, X.D. and Zhong, X.C. 2006. Observation and analysis of rainfall infiltration. Rock and Soil Mechanics. 27(11): 18731879.
0.2
0.4 0.6 Time /d
0.8
1.0
Figure 9. The cumulative infiltration curves of rainstorms and D1 pattern.
the landfill earthen final cover because the infiltration capacity is lower than the rainfall intensity and more rainfall water will be runoff rather than infiltration. 4 CONCLUSIONS
Series one-dimensional numerical simulations of infiltration process into landfill earthen final covers were conducted to investigate the effects of rainfall patterns on the infiltration rate subjected to different initial conditions and the same daily precipitation. The numerical simulation model conducted by SVFlux was firstly verified by experimental data. The numerical simulation results indicate that the peak rainfall intensity occurs earlier, the cumulative infiltration will be larger no matter what the initial condition is. The advanced pattern (A1) will generate the largest amount of infiltration compared to other patterns. The delayed pattern (D1) will generate the minimum amount of infiltration compared to other patterns. The time to ponding increases as the initial matric suction increases. The rainstorm will generate less infiltration compared to the rainfall patterns lasting the whole day. The numerical analyses suggest that rainfall pattern should be taken into account in the calculation of accurate infiltration into earthen final covers. ACKNOWLEDGEMENT The authors would like to acknowledge the financial support from the National Key Technology R&D Program funded by the Ministry of Science and Technology of China (2006BAJ06B02) and research grants (50538080 and 50878194) provided
1441
New areas
LNAPL retention in partially saturated silty sand

M. Alferi
E. Romero
A. Dominijanni & M. Manassero

ABSTRACT: The paper presents an experimental programme aimed at studying the retention properties of partially saturated silty sand in terms of the simultaneous storage of water and light hydrocarbon (LNAPL). Specifically, this study pursues a better understanding of the multi-phase flow and transport phenomena in oil-polluted unsaturated soils. Using a mixture of two natural soils (Ticino Sand and Pontida Silt), soil samples were prepared at different initial dry densities and oil and water contents. Soltrol was used as analogue of the LNAPL. The investigation was focused on both microscopic and macroscopic point of views by using mercury intrusion porosimetry and vapour transfer technique. Clear aggregation phenomena on this mixture were observed on addition of small amounts of Soltrol, which have important consequences on the retention properties. The aggregation induces a double-porosity soil structure, which is not easily erased on subsequent hydraulic cycling. 1 INTRODUCTION Delage (2003) and Manassero et al. (2005). Other experimental studies on LNAPL contaminated sands were carried out by Lenhard and Parker (1987 and 1988), Busby et al. (1995), Sharma and Mohammed (2003) and Kechavarzi et al. (2005). The work presented is part of a research programme aimed at evaluating the applicability and the efficiency of a new remediation technique based on the application of high vacuum for the multiphase extraction of hydrocarbons. Preliminary results of the experimental activity focused on measuring the retention curve of a silty sand contaminated by LNAPL are presented in the paper. 2 2.1 EXPERIMENTAL ACTIVITY Materials
Soil contamination by hydrocarbons is nowadays a relevant problem. The improvement of existing remediation techniques for polluted soils and the development of more efficient new methods is a challenging goal of research. A better understanding of flow and transport processes in contaminated soils, with special attention to partial saturated conditions, is necessary for a correct evaluation of the performances of such remediation techniques. A contaminated partially saturated soil is a system composed by four species: soil, air, water and contaminant. The contaminant can be distributed between solid, aqueous and gaseous phase and can be present as a non miscible liquid called NAPL (Non-Aqueous Phase Liquid). This paper focuses on LNAPLs (Light Non Aqueous Phase Liquid) which are lighter than water and, as a result, float on the water table. In spite of the relevance of the topic, few experimental studies are available in the literature, the most of which referred to two fluid system (air and water, air and contaminant and water and contaminant). With reference to three fluid system (air, water and contaminant), works focused on the experimental determination of the retention properties of a silty sand were presented by Cui and
Laboratory tests were carried out on artificially prepared (statically compacted) samples (50% of Ticino Sand and 50% of Pontida Silt with an average density of solid particles of 2.70 Mg/m3). Samples were prepared at different dry densities and different liquid contents as shown in Table 1. Contaminated samples were prepared using Soltrol170 Isoparaffin Solvent (Chevron Phillips Chemical Company LP), a mixture of branched alkanes with carbon numbers ranging from C10 to C16 having a very low solubility in water and
1445
Table 1.
Initial sample properties.

0.6
Freeze Drying
Pore Size Density Function
Air Drying
Dry density, d Sample (Mg/m3) 1 2 3 4 5 Table 2. 1.70 1.85 2.00 1.85 2.00
Initial void Water Oil ratio, content, content, e0 () ww (%) wo (%) 0.588 0.459 0.350 0.459 0.350 10 15 11 11 7 3 3
Total degree of saturation, S (%) 45.9 88.1 84.9 87.3 83.4
AIR + WATER Stot = 45.9%

0.4
Sample 1:
0.2
Properties of the fluids used in the study.

0.6
(Mg/m ) at 20C
3
(Pa s) at 20C 1.00 103 2.03 103
a,j (N/m) at 20C 72.7 103 24.2 103
Pv at 25C (kPa) 3.17 0.08
AIR + WATER Stot = 88.1%

0.4
Sample 2:
Water 0.998 Oil 0.782
0.2
density lower than water (Lenhard and Parker, 1988). Water and Soltrol physical properties (i.e., density , dynamic viscosity , air-liquid interfacial tension a,j and vapor pressure Pv) are summarized in Table 2. Samples are prepared by adding to the dried soil mixture water and successively oil. The amounts of water and oil were calculated so that samples with the same density had a similar degree of saturation. The mixture was statically compacted in oedometric rings having a diameter of 50 mm and 20 mm of height. In the following, the subscripts a, w and o will be used in order to indicate air, water and oil, respectively. 2.2 Microstructural analysis
0.6
AIR + WATER + OIL Stot = 87.3%

0.4
Sample 4:
0.2
0 1 10 100 1000 10000 100000 1000000
Pore Diameter, d (nm)
Mercury intrusion porosimetry (MIP) tests were carried out on an AutoPore IV 9500 porosimeter (Micrometrics Instrument Corporation) applying mercury pressure between 0 and 223 MPa (apparent pore diameter from 360 m to 7 nm). MIP technique requires dehydrated samples measuring approximately 1 cm3 (sample 1, 2 and 4). Sample 1 is representative of a soil mixture with a low water degree of saturation; sample 2 is representative of a soil mixture with a high degree of saturation; finally sample 3 is representative of a contaminated soil mixture. Sample dehydration was obtained by means of freeze-drying and air-drying. Air drying technique has been used to investigate shrinkage effects on the pore network. The results of MIP tests are shown in Figure 1 in terms of pore size distribution (PSD). With reference to freeze-dried samples, quite different structures are detected. Samples 1 and 4 show
Figure 1. Pore size distribution of sample 1; sample 2 and sample 4.
a bimodal porosimetry (macro and micro pores), whereas sample 2 is clearly mono-modal. The addition of Soltrol (sample 4) maintains the double porosity as in sample 1, in spite of the total degree of saturation is similar to that of sample 2. These differences in soil structure affect the retention curve, as described in paragraph 3. Comparing the PSDs of freeze-dried and airdried samples a shrinkage effect appears. Considering the samples with a dual-porosity structure, it can be observed that for sample 1 the shrinkage affects only the macro-pores, while the sample 4, containing oil, shows shrinkage not only concentrated at the macroscopic but also at the microscopic level. In the light of the results, the addition of Soltrol maintains the double structure of the wetted soil (ww 11%). During drying process, the Soltrol
1446
Total suction, (MPa)
stored in the macropores of sample 4 limits the evaporation of water held at the macro-structure. In this way, the demand of water required by the applied suction is also supplied by the water held at the micro-porosity. This phenomenon can explain the shrinkage of micropores for sample 4. Nevertheless, this fact should be further investigated. 2.3 Vapor transfer technique
1000
Sample 4 Sample 5
100
10
The first part of the macroscopic analysis has been carried out using a chilled-mirror psychrometer (WP4 PotentiaMeter, Decagon Devices INC., see Cardoso et al. 2007). The equipment determines the relative humidity in equilibrium with the sample by estimating the dew-point temperature of the humid air. Compacted samples were carefully prepared and, starting from the initial water content (Table 1), progressively dried under controlled temperature and humidity conditions. According to the different vapor pressures of water and Soltrol (see Table 2), we can assume that the oil content remains constant during the test (Manassero et al., 2005). Figure 2 and 3 show the results obtained for samples 1, 2 and 3 and samples 4 and 5, respectively. As observed in the figures no appreciable density effects are detected. The residual water content for samples 1, 2 and 3 is about 0.6%, which is approximately 3% lower than the residual liquid content of samples 4 and 5. Considering that in sample 4 and 5 the oil content is 3%, this fact ensures that no significant evaporation of Soltrol occurs. In order to better evaluate the effect of oil content on the retention curves obtained by vapor transfer technique, a new set of samples has been prepared,
1000
0.1 0 4
Total Liquid Content, wtot (%)
12
16
Figure 3. Retention curve of samples prepared with water and oil.

1000
100
Total Suction, (MPa)
woil = 0% woil = 1% woil = 2% woil = 3%
10
0.1 0 4
12
16
Figure 4. Suction measurements for different initial oil content.

Sample 1 Sample 2 Sample 3
100
10
adopting a constant water content (ww 11%) and different oil content (wo equal to 0%, 1%, 2% and 3%). In Figure 4 the obtained retention curves are reported. Suction is controlled only by the water content. As a result, approximately the same total suction (about 1 MPa) is measured at the initial conditions, in spite of the different initial total liquid contents. At the end of the drying path, samples reach a residual liquid content variable in accordance with the different oil content. 3 AN INSIGHT INTO RETENTION PROPERTIES OF TWO AND THREE FLUID SYSTEMS
Total Suction, (MPa)
0.1 0 4
Water Content, ww (%)
12
16
Figure 2. Retention curve of samples prepared only with water.
MIP can be used to derive the retention curve of the soil observing that the mercury intrusion process
1447
is similar to air-intrusion during the drying path of the retention curve. Thus, the injection of nonwetting mercury is equivalent to the ejection of the wetting phase by the non-wetting front advance of air, for the same diameter of pores being intruded (Romero and Simms, 2008). To obtain matric suction (s), the mercury pressure must to be corrected with a corrective factor, , defined as:
On the other hand, the fitting of the bimodal retention curve was obtained applying a modified form of van Genuchten equation as follow: w wres e m wsat wres e0 m 1 m 1 m s 1 1 m 1 eM e0 1 M s M
1
1
f cos f Hg cos Hg
(2)
(6)
where f and Hg are the interfacial tension of the wetting fluid (water and oil) and mercury respectively, while f and Hg are the contact angles of the wetting fluid and mercury. The liquid content or degree of saturation must be estimated taking into account the volume of pores not intruded by mercury and the residual degree of saturation of the non-wetting fluid. Moreover, the wetting fluid is held by capillarity at low suction (2 MPa) and the derivation of suction-liquid content relationship should be limited to the low suction range, in which capillarity dominates the retention properties of fluids (Romero and Simms, 2008). To complete the retention curve at high suction values, WP4 data are used under the assumption that osmotic suction is negligible for the considered soil. The corrected values of the liquid content (w) can be calculated as: w Srel wsat wres wres (3)
where e0 is the initial void radio of the soil, em is void radio related to micro-porosity, eM is void ratio due to macro-porosity, m, m and M and M the fitting parameters for the micropores and for the macropores respectively. Equation 6 is obtained applying van Genuchten expression (Eq. 5) both for the micro and the macrostructure, determining two sets of parameters that are related by taking into account the void ratio of the soil. The values of em and eM can be estimated directly from the pore size density. It can be stressed out that a relevant difference between m and M should be observed as these two parameters are related to the structure of the soil, while m and M are expected to be very similar. The procedure here described has been applied to sample 2 (air/water system) and 4 (air/water/oil system) finding how the variation of the pore structure can affect the retention properties of the soil. 3.1 Two fluid system (air/water system)
where wsat is the liquid content corresponding to the complete saturation condition, wres in the residual liquid content derived from WP4 measurements in the range of suction between 1 and 2 MPa and Srel is the relative saturation defined as: Srel S Sres 1 Sres (4)
with Sres wres/wsat residual degree of saturation. The van Genuchten equation (van Genuchten, 1980) was used to fit experimental data with dominant unimodal soil structure: w wres 1 wsat wres 1 s
1 1
The retention curve of the two fluid system composed by air and water was assessed referring to sample 2. MIP data were processed and suction values corresponding to the applied mercury pressure were estimated taking 0.196 (considering the airwater interfacial tension in Table 2 and a contact angle, w, of 180). Water contents were calculated using equation 3 and then fit according to equation 5. The obtained retention curve is shown in Figure 5 and curve fitting parameters are reported in Table 3. 3.2 Three fluid system (air/water/oil system) The three-fluid system (i.e. air/water/oil system), requires further hypotheses in order to realize the back analysis from MIP data. Two conditions can be distinguished: (1) removal of water maintaining constant oil content and (2) removal of oil with constant water content. In the first case, matric suction (sw) is defined as the difference between air pressure (ua) and water pressure (uw), while in the second case the suction (so) represents the
(5)
where wsat is the saturated liquid content, wres is the residual liquid content, s is the matric suction and and are the fitting parameters.
1448
1000000
WP4
1000000
100000
MIP Fitting
100000
Constant Oil Content, wo = 3% WP4 MIP

Fitting
Matric Suction, sw = ua-uw (kPa)
1000
Matric Suction, sw = ua-uw (kPa)
10000
10000
1000
100
100
10
10
0.1 0 4 8 12 Total Liquid Content, wtot (%) 16 20
0.1 0 4
12
16
20
Figure 5. Table 3. (kPa-1) 0.006
Retention curve for sample 2. Curve fitting parameters for sample 2. 1.407 wsat 0.163 wr 0.000 R2 0.9972
Figure 6. Retention curve for sample 4 assuming constant oil content (wo 3%) during drying path. Table 4. Curve fitting parameter for sample 4 (water drying path with constant oil content wo 3%). Micro-porosity m (kPa1) m em Macro-porosity M (kPa1) M eM wsat wres R2 0.001 1.888 0.213 5.246 1.891 0.162 0.213 0.028 0.9937
difference between air and oil (uo) pressures. To estimate the total degree of saturation (Stot), the different fluid densities are required: w w Stot e0 s w o w o
3
Other parameters
(7)
100
where s is the soil density (2.7 g/cm for the considered soil) and w and o are the densities of water and oil respectively (see Table 2). The above-mentioned concepts were applied to estimate the retention curve of sample 4. In Figure 6, the retention curve obtained considering the removal of water maintaining constant the oil content (i.e., wo 3%) is presented. Suction values were calculated scaling the applied mercury pressure with 0.196. Water content from MIP data was estimated from equation 7 and then combined with WP4 data. The bimodal retention curve was fit according to equation 6 and the fitting parameters are reported in Table 4. Starting from the same porosimetry, it is possible to estimate oil removal when the initial water content is maintained constant (Figure 7). The variation of oil content was calculated according to equation 7, while the evaluation of the corresponding suction values was done applying a constant 0.087. This value was directly estimated by measuring capillary rise of Soltrol and considering its contact angle and wettability.
Constant Water Content, ww= 11%

MIP van Genuchten
Suction, so = ua-uo (kPa)
10
0.1 0 4 8 12 Total Liquid Content, wtot (%) 16
Figure 7. Retention curve for sample 4 assuming constant water content (ww 11%) during drying path.
In this case suction is defined as the difference between air and oil pressures. The curve was fit with the unimodal equation and fitting parameters are summarized in Table 5.
1449
Table 5. Curve fitting parameter for sample 4 (oil drying path with constant water content ww 11%). (kPa1) 1.755 2.325 wsat 0.157 wr 0.112 R2 0.9928
ACKNOWLEDGEMENTS The first author acknowledges Universitat Politcnica de Catalunya and, in particular, the Geotechnical Laboratory of the Department of Geotechnical Engineering and Geoscience to make available the experimental devices used for this study. REFERENCES
Busby, R., Lenhard, R.J. & Rolston, D.E. 1995. An investigation of saturation: capillary pressure relations in two and three-fluid system for several NAPLs in different porous media. Groundwater J. 33(4): 570578. Cardoso, R., Romero, E., Lima, A. & Ferrari, A. 2007. A comparative study of soil suction measurement using two different high-range psychrometers. Proc. 2nd Int. Conf. on Mechanics of Unsaturated Soil; Experimental unsaturated soil mechanics, T. Schanz (Ed.)., Springer Verlag 112: 7993. Cui, Y.J., Delage, P. & Alzoghbi, P. 2003. Retention and transport of a hydrocarbon in a silt. Gotechnique 53 (1): 8391. Kechavarzi, C., Soga, K. & Illangasekare, T.H. 2005. Two-dimensional laboratory simulation of LNAPL infiltration and redistribution in the vadose zone. J. Contam. Hydrol. 76: 211233. Lenhard, R.J. & Parker, C.J. 1987. Measurement and prediction of saturation-pressure relationships in three-phase porous media systems J. Contam. Hydrol. 1: 407424. Lenhard, R.J. & Parker, J.C. 1988. Experimental validation of the theory of extending two-phase saturation-pressure relations to three fluid phase system for monotonic drainage path. Water Resourc. 24(3): 373380. Manassero, M., Rabozzi, C., Ribotta, L. & Musso, G. 2005. On the retention properties of a polluted sandsilt mixture. Proc. Advanced Experimental Unsaturated Soil Mechanics, Tarantino, Romero & Cui (Ed.). Taylor & Francis Group, London: 459465. Romero, E. 1999. Characterization and thermo-hydromechanical behavior of unsaturated Boom-clay: An experimental study. Ph. D. Thesis, Universitat Politcnica de Catalunya, Barcelona. Romero, E., Gens, A. & Lloret, A. 1999. Water permeability, water retention and microstructure of unsaturated compacted Boom clay. Eng. Geol. 54: 117127. Romero, E. & Simms, P.H. 2008. Microstructure Investigation in unsaturated soils: a review with special attention to contribution of mercury intrusion porosimetry and environmental scanning electron microscopy. Geotech. Geol. Eng. 26(6): 705727. Sharma, R.S. & Mohamed, M.H.A. 2003. Patterns and mechanism of migration of light non-aqueous phase liquid in an unsaturated sand. Gotechnique 53(2): 225239. Smettem, K.R.J. & Kirkby, C. 1990. Measuring the hydraulic properties of a stable aggregated soil. J. Hydrol. 117: 113. Van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. 44: 379392.
Suction controlled tests for oil matric suction are in progress in order to verify the retention curve here estimated. Comparing the retention curves plotted in Figure 5 and in Figure 6 and 7, it is clear that important changes are induced as a consequence of the addition of Soltrol at low suction values (i.e., the more significant range of suction referring to application of these relations to the evaluation of soil clean up). 4 CONCLUSIONS
The paper presents the preliminary results of an experimental programme aimed at studying the retention properties of a partially saturated silty sand contaminated by LNAPL (Soltrol170). The experimental activity was focused on microscopic (mercury intrusion porosimetry) and macroscopic (dew-point psychrometer) point of view for specimens prepared with and without Soltrol. MIP shows that changes of the soil pore structure occur when: (1) decreasing the dry density of the soil, (2) reducing the initial water content and (3) adding some amount of Soltrol. In all the abovementioned cases, the soil presents a double-porosity structure that affects the retention curve. Psychrometric tests (WP4 dew-point potentiometer) were carried on drying at constant oil content. The results show that oil content remains constant during the test, determining the total residual saturation. Psychrometric data and mercury intrusion porosimetry were used to estimate suction-liquid content relationships. Experimental data were fit using unimodal and bimodal van Genuchten equations. The retention curve of two fluid system (i.e., unimodal retention curve) and three fluid system (i.e., bimodal retention curve) showed relevant differences, in particular at low suction values. The addition of small amounts of LNAPL induces aggregation that is not easily removed by water. This aggregation affects the air entry value of the polluted material producing relevant variations on the retention properties of the soil. Tests with suction controlled oedometer are envisaged to better understand the retention properties of the soil at low suction values and to verify the suction-liquid content relationship.
1450
Water retention curve for evolving marl under suction cycles

R. Cardoso
Instituto Superior Tcnico, Lisbon, Portugal
G. Della Vecchia C. Jommi
Politecnico di Torino, Torino, Italy Politecnico di Milano, Milano, Italy
E. Romero
ABSTRACT: Data of the water retention domain measured on upper Jurassic marl from Portugal are presented. Under drying and wetting cycles physical degradation was observed, which causes an evolution of the microstructure and of the water retention domain of the material. Classical water retention models from the literature are not able to account for evolving retention domain with degradation, unless the parameters are adapted as the number of drying and wetting cycles increases. A different approach is proposed here, based on microstructure data, which quantify the changes in the microstructural void ratio. The model is able to describe the evolving retention domain, by accounting for the changes in the amount of water stored in the material as a function of its void ratio. In this way the effects of hydro-mechanical coupling and degradation of the material on its hydraulic behaviour may be correctly reproduced. 1 INTRODUCTION of the marl. Retention data from three suction cycles are analysed and interpreted with the aid of microstructural data coming from Mercury Intrusion Porosimetry (MIP). Possible approaches for modelling the evolution of the water retention domain are discussed. Selected literature models are verified against the experimental data. Their limitations suggested the development of an enhanced framework, in which the evolution of the water retention curve, WRC, is tracked by means of few parameters, having a sound physical basis, and accounting for the changes in the pore size distribution emerging from the MIP data inspection, and water exchanged between the different structural levels. The extension of the meaning of the WRC provided by its physical interpretation justifies its use to characterise the soil water retention domain. 2 CHARACTERISATION OF THE MATERIAL
The large majority of existing models for unsaturated hard soils and evolving rocks consider the mechanical and the hydraulic behaviour as uncoupled. However, the influence of suction on the mechanical behaviour of the materials, as well as the influence of mechanical loading on their hydraulic properties is well reported in the literature (see, for instance, Vanapalli et al. 1999, Kawai et al. 2000). In natural highly structured hard soils and weak rocks, the pattern is even more complicated by the weathering effects resulting from the coupled action of mechanical and hydraulic cyclic loadings. In this framework, physical degradation of an upper Jurassic marl from Portugal was studied through an extensive set of experimental tests (Cardoso 2009). The experimental results highlighted strong degradation of the natural marl occurring as a result of a wetting process (Cardoso & Alonso 2009), affecting both the compressibility and the retention characteristic of the material. To model properly the behaviour of the marl, the effects of the evolving structure of the material must be taken into account (Cardoso 2009). Here the attention is focused on the retention behaviour
The marl studied comes from the upper Jurassic formation of Abadia (Arruda dos Vinhos, Portugal). The material has a relatively high plasticity (wL 49% and PI 25%) consistent with the
1451
nature of the minerals present (mainly chlorite, kaolinite, bentonite, illite and gypsum, with no smectite detected, quartz, feldspar, calcium carbonate and mica). The solid unit weight is 27.0 kN/m3 (Gs 2.75), in situ the porosity ranges between 18% and 39% and the water content between 8% and 22%, corresponding to relatively high degrees of saturation (70%100%). An extensive set of experimental tests was performed to characterise the behaviour of the material (Cardoso 2009). As for the hydraulic characteristics, the natural marl has a saturated permeability of ksat 7.5 1014 m/s. The compressibility index against suction changes (Alonso et al. 1990) is s 0.008. Void ratio shows a rather strong dependence on water content. With reference to the data reported in Figure 1, it is interesting to observe that for the samples investigated, the sensitivity of void ratio to changes in water content is clear for water contents above 10%. Below this value, the void ratio changes only slightly with the amount of soil water. The water retention curve was determined by means of WP4 measurements. The data collected during three drying-wetting cycles are presented in Figure 2 as a function of water content. Water content was adopted in the representation because no volume control was provided during the tests.
0.70 0.60 void ratio 0.50 0.40 0.30
water content (%)
10
12
14
16
18
20
Figure 1.
1000.00 100.00 Total suction (MPa) 10.00 1.00 0.10 0.01
Void volume dependence on water content.

Before the cycles - drying Before the cycles - wetting After 2 cycles - drying After 2 cycles - wetting After 3 cycles - drying
The data plotted in Figure 2 demonstrate that the retention domain evolves as the number of drying-wetting cycles increases. In the first drying branch, performed on a sample of natural material, the air entry value seems to be above a value of suction of 1.0 MPa. The following drying branches clearly have a much smaller air entry value, tending to 0.1 MPa. Reduction of the air entry value is accompanied by a clear increase in void ratio, which can be appreciated if the value of water content at saturation is tracked. The inflexion point for the drying branches seats around 10 MPa in the first cycle and near 4 MPa in the following cycles. These observations, all together, indicate that physical degradation is taking place, and that the first drying-wetting cycle induces rather strong structural changes. The evolution of the saturated water content for increasing number of cycles seems to suggest the development of some macroporosity which could be attributed to microcracking (Cardoso 2009). To provide a quantitative description of the microstructure evolution, mercury intrusion porosimetry tests (MIP) were performed. The pore size distributions of three relevant samples are presented in Figure 3. Three fragments with different water contents were analysed: (i) natural water content of the samples stored in the laboratory (w 14%, e 0.507, with relative humidity equal to 95%, corresponding to a suction of 12 MPa at 25C), (ii) almost dry conditions imposed by vapour equilibrium technique (w 1%, e 0.451, with relative humidity equal to 11%, corresponding to a suction of 310 MPa at 25C), and (iii) fully saturated case (w 21%, e 0.651). The data show that the natural material has a fairly uniform pore size, identified by the marked peak at 148 nm in Figure 3 If the material is dried, starting from its natural water content, the unimodal distribution keeps unaltered. The diameter corresponding to the peak decreases a little, being 105 nm for w 1%.
0.6 w=21% 0.5 w=14% w=1%
Pore size density
After 3 cycles - wetting
0.4 0.3 0.2 0.1 0
1
0 5 10 15 20 25 water content (%)
10
100
1000
10000
100000
Pore diameter (nm)
Figure 2.
Water retention curve for three cycles.
Figure 3. MIP results: effects of drying or wetting on the material pore size distribution.
1452
On the contrary, after a wetting path, the pore size distribution no longer shows a unimodal pore size distribution. The original peak shifts a little toward a value around 100 nm. Part of the voids clearly swell and a second peak at 672 nm emerges. The data for the saturated material indicates that swelling occurred increasing the size of the smaller pores since the entire curve slightly displaces to the right. However the higher peak at a diameter slightly smaller than 1 m may indicate that microstructure evolves as a consequence of saturation. The occurrence of a third small peak at a diameter of about 10 m can be attributed to microcracking, confirmed by Environmental Scanning Electron Microscope (ESEM) photographs (Cardoso 2009). Disregarding the latter, after saturation the microstructure of the marl may be considered to be characterised by a double structure. The volume increment at decreasing suction, which was confirmed by the elaboration of ESEM photographs (Cardoso 2009), seems to be mostly recovered during a drying path. The effect of a closed wetting-drying path on the marl microstructure is shown in Figure 4, where the dry (w 1%) and full saturated (w 21%) samples from Figure 3 are compared with a sample dried in a RH 11% environment after being fully saturated (w 1% after saturation and drying). Shrinkage caused by drying reduces the dimensions of the pores developed on full saturation back to values comparable to those of the original material. Nonetheless, the micropores increase slightly their final size. The peak in the smallest dimension displaces to the right on the diameter axis. The larger macropores (microcracks) do not disappear. It is even more marked, although shifted a little to the left. All together the results are consistent with an increment in the saturated water content, hence in void ratio, with suction cycles observed in the WRC shown in Figure 2.
0.6 0.5 w=21% w=1% w=1% (after full saturation) 0.4 0.3 0.2 0.1 0
MODELLING THE WATER RETENTION DOMAIN WITH LITERATURE EQUATIONS
Two well known expressions from the literature (van Genuchten 1980 and Romero & Vaunat 2000) were adopted to fit the data presented in Figure 2. The first one was conceived for non active soil (soils with a unimodal pore size distribution or soils for which the volume changes of the microstructure do not significantly affect global volume changes), and uses the (normalised) degree of saturation as a convenient variable to describe the amount of water stored in the pore space. Linear scaling with void ratio provides homothetic retention curves at varying void ratio. The proposal of Romero & Vaunat (2000) stems from the evidence that in compacted soils two levels of porosity are present, and that the two levels show different sensitivity to mechanical loading. In the model by Romero & Vaunat, a van Genuchtens form is retained for the interaggregate porosity, which scales linearly the effect of inter-aggregate void ratio. The intra-aggregate pore space is constant and mainly depends on the physico-chemical characteristics of the active constituents of the soil. The expression proposed by van Genuchten (1980) relates the degree of saturation degree, Sr, to the suction,, by means of Equation 1, where P and are material parameters. By relating degree of saturation to water content w (Eq. 2), by means of specific gravity Gs and void ratio e (ratio between the volume of voids Vv and the volume of solids Vs), the definition of the water retention curve in the suction versus water content plane may be obtained (Eq. 3).
1 1 Sr 1 P
(1)
Sr
Gs w e
1 1 wsat 1 P
(2)
Pore size densit y
1 e 1 w 1 Gs P
(3)
10
100
1000
10000
100000
Pore diameter (nm)
Figure 4. MIP results: effects of drying the sample after previous full saturation.
In the case of expansive materials, the use of Equation (3) is not trivial, as during a dryingwetting cycle void ratio changes depending on water content (Fig. 1). In the case of this investigation, the volume changes were not measured in the determination of the WRC. To overcome the lack of experimental data, the calibration of the parameters was made assuming a constant value for void ratio for each cycle. Its
1453
value was adjusted to fit the curve in the low suction range. As the marl exhibits more evident structural changes near saturation, the assumption of constant void ratio is not totally realistic. Moreover, the values of wsat after each cycle depend on the irreversible volume changes caused by degradation. Therefore, the data can be reproduced reasonably with a van Genuchten model only if each cycle is considered independently from the others. The void ratio used in the calculation of the saturated water content for the first cycle was approximately the value measured in the sample before the test. The values for the other cycles were obtained numerically. The WRC obtained with this procedure are represented in Figure 5. The fitted parameters are collected in Table 1 (D-drying, W-wetting). Although the final curves may represent well the experimental data, a set of 4 calibrated parameters is needed to describe the water retention domain for each cycle. The parameters are just fitting values as they can not be inferred from physical observation. The model by Romero & Vaunat (2000) was proposed for compacted deformable clays to account for the relationship between suction range and the
pore size affected by water exchange in the hydraulic path. The water retention capabilities of the soil are related with the different dimensions of pores characterising compacted clayey soils, and depend on the changes in the dimensions of the pores caused by mechanical actions. In the low suction range (high water content and high saturation degree) water is exchanged in the larger inter-aggregate pores. For deformable materials the changes in void ratio affects mostly these pores, hence water exchanges in the low suction range. Water in the micro intra-aggregate voids has reduced mobility due to the small dimensions of these pores and also to the chemical nature of the minerals. This amount of water is exchanged only at very high suctions, and it does not affect significantly the volumetric behaviour of the soil. Romero & Vaunat (2000) assumed that the intra-aggregate pore space is constant, depending only on the clay activity. The entire water retention domain is given by an expression similar to that proposed by van Genuchten, but a correction factor C is introduced to enhance the flexibility of the model at high suctions: 1 w wmax C 0 1 0 n
m
1000.000
Rock - drying Rock - wetting Rock after 2 cycles - drying Rock after 2 cycles - wetting Rock after 3 cycles - drying Rock after 3 cycles - wetting
(4)
100.000
with 0 ln 1 a C ( 0 ) 1 ln(2 )
Total suction (MPa)
10.000
1.000
(5)
0.100 2nd Cycle ? 1st Cycle ? 0.001 0 4 8 12 16 20 24 28 32
0.010
3rd Cycle ?
Figure 5.
Curves adjusted with van Genuchten model.
Table 1. Fitting parameters for Van Genuchten model. Cycle 1st cycle (e 0.431) 2nd cycle (e 0.594) 3rd cycle (e 0.618) D W D W D W P (MPa) 4.0 1.8 0.5 0.2 0.4 0.2 0.30 0.28 0.23 0.21 0.23 0.22
where wmax is the water content at saturation for total suction 0, n is related to the slope of the curve at the inflexion point, m is related to the residual water content and is mainly the inverse of the air-entry value of the soil. The fitting parameter a in the correction factor C is the maximum suction corresponding to a water content w 0. Equation 5 makes the curve tending to a linear relationship between the logarithm of and w in the high-suction zone, with intersection parameter, a, at w 0. Although Equation 4 appears as a simple correction of the original proposal by van Genuchten, it proves able to model the reduced influence of void ratio on the volume changes of the small pores. Calibration of the model on the marl data, gives the result presented in Figure 6, with the fitted parameters listed in Table 2 (D-drying, W-wetting). The model provides better adjustment to the experimental data especially for low values of suction, thanks to the two additional calibration parameters. Nonetheless, the parameter a could be kept
1454
1000.000
100.000
10.000
1.000 3rd Cycle 0.100
0.010 1st Cycle 0.001 0 4 8 12 16 20 ? 2nd Cycle ? 24 28 ? 32
Figure 6. Curves adjusted with the equation proposed by Romero & Vaunat (2000). Table 2. Fitting parameters for Romero & Vaunat model. Cycle 1st cycle (wsat 17%) 2nd cycle (wsat 25%) 3rd cycle (wsat 31%) D W D W D W a (MPa) 1000 1000 1000 1000 1000 1000 (MPa1) 0.03 0.04 2.0 4.0 5.5 11.0 m 1.30 1.60 0.50 0.55 0.50 0.55 n 0.60 0.50 0.55 0.55 0.60 0.55
leaving the inter-aggregate pore structure empty. In their work, Romero & Vaunat had assumed a fixed value for wmicro. More recently, Della Vecchia (2009) enhanced the framework to account for swelling and shrinking of the aggregates depending on clay activity. He proposed a new model for the water retention curve in which intra-aggregate porosity may evolve with the current amount of water, while the inter-aggregate porosity depends on void ratio. The capability of the latter proposal to account for an evolving microstructure was exploited here to model the behaviour of the marl. As already discussed, the marl is characterised by a pore size distribution with a peak which is weakly sensitive to drying wetting cycles. Below a water content value of w 10% (see Fig. 1) the void ratio is rather insensitive to changes in the water content. Sensitivity of the void ratio to changes in the water content becomes apparent for water contents above 10%, when a second higher peak is developed. Deeper inspection of the data in Figure 1 suggests that for w 0.1 a linear increment of void ratio with water content may describe the evolution of the material microstructure at least as a first approximation: e 0.7wGs (6)
Total suction (MPa)
constant, as it represents the maximum attainable suction in the laboratory. Considering the other three parameters, to follow the evolution of the water retention domain with increasing number of cycles, only two must change significantly. They are the parameter governing the air entry value and one of the other two governing the shape of the curves. It is worth noting that both models may represent with a sufficient degree of accuracy the experimental data. Nonetheless, calibration of the parameters is purely numerical and changes in void ratio during the drying wetting process are not considered. 4 MODELLING A RETENTION DOMAIN WITH EVOLVING PORE SIZE DISTRIBUTION
Therefore, the amount of matrix pore volume over solid volume, em Vm/Vs, is assumed to be the sum of a constant part, e*m, representing the void m , ratio at low water contents, plus a contribution, e which evolves with the water content:
* m , em em e
(7)
where m 0, e for w 0.1, (8a) for w 0.1 (8b)
* m w Gs em e ,
with the material parameter describing the sensitivity of the matrix pore volume to changes in the water content above the cut-off value 0.1. Given this definition, the expressions proposed by Della Vecchia (2009) may be adapted to describe the water content as a function of suction. Two different equations describe the range of water contents below and above the cut-off value 0.1: For w 0.1, the main drying and the main wetting curves may be written in the form:
max * log * bem 1 m w 1 , Gs log max b log * * m m
To better represent the evolution of void ratio due to mechanical loading paths, Romero & Vaunat (2000) suggested to scale Equations 4 and 5 to describe the inter-aggregate pore range, hence for water content values in the range [wmax; wmicro] where wmax was already explained and wmicro is the water content which saturates the intra-aggregate pores

(9)
1455
where *m is the suction corresponding to e*m, max is the maximum suction attainable, and is b a parameter ruling the average slope of the curve for high values of suction. - For w $ 0.1, the current water contents depends on the microstructure and on the void ratio,
log 1 m 1 w em (e em ) 1 2 Gs

1 ( )n 1
m
(10)

where the suction m is defined as

max * bem b log * m * m exp max * log e be m m *
m
b

(11)
and e is the current void ratio, with , m and n parameters of the model. Among the last three parameters, which have in principle the same meaning as the equivalent parameters introduced in the model by van Genuchten (1980) or Romero & Vaunat (2000), only two are independent one from the other. Parameter comes by imposing the two curves (Eqs. 9 and 10) to be continuous together with their first derivatives. For all cycles a unique value of and of the parameters m and n naturally leads to a correct dependence of the air entry value on void ratio. The original model by Della Vecchia (2009) was developed for clays with an intrinsic double structure, where the volume changes of the microstructure (expansive clay minerals) influence global volume changes. For the marl studied the volume changes of the porosity of the matrix are associated to the development of a second pore mode, which is apparent from the elaboration of the MIP data. This second pore mode is mathematically treated as belonging to the macropore range, as it is sensitive to mechanical actions (Cardoso 2009). As a first approximation, the data can be modelled by assuming that the curves for high values of suction are the same, as they are not much affected by cyclic changes in the water content. A value of e*m 0.275 was chosen, corresponding to water content w 10% to discriminate the voids which volume remains practically unchanged. The parameter b was calibrated to reproduce the correct curvature of the retention domain for high values of suction. The experimental data allowed for choosing a value of D* 8.0 MPa for the drying branch and of W* 3.0 MPa for the wetting branch, corresponding to e*m. These values correspond to the values of suction to which all the
wetting and drying branches of the water retention domain in the different cycles converge. The maximum attainable suction was given the value max 1000 MPa. As for the calibration of Equation 11, in order to reduce to a minimum the total numbers of parameters, the relationship between n and m suggested by van Genuchten, n 1/(1 m), was exploited. Only the parameter m had to be calibrated (a value of m 0.25 was chosen). The entire set of parameters of the model is collected in Table 3. The three cycles were simulated by just changing the void ratio, in such a way that the final value of water content (for suction tending to zero) represented fully saturated conditions. Void ratio of 0.430, 0.610 and 0.630 were adopted for the three cycles respectively, corresponding to saturated water contents of 16%, 22%, and 23%, respectively. The comparison between the simulation and the experimental data is shown in Figure 8. The results in Figure 8 show that the model proposed is able to provide a pretty good description of the evolving retention domain of the marl, at increasing number of cycles. The quality of the prediction is comparable to that of the previous equations adopted, if not better. Nonetheless, it is important to remark that a single set of parameters was adopted in this case for all cycles, the difference standing only in the saturated water content at the end of the cycle.
Table 3. Fitting parameters for the model proposed. e*m () 0.275 max (MPa) 1000 W* (MPa) 3.0 D* (MPa) 8.0 b () 20 () 0.7 m () 0.25
1000.000
100.000
Total suction (MPa)
10.000
1.000
0.100
0.010 1st Cycle ? 0.001 2nd Cycle
3rd Cycle ?
12
16
20
24
28
32
Figure 7.
Curves adjusted with the equation proposed.
1456
CONCLUSIONS
REFERENCES
Alonso, E.E., Gens, A. & Josa, A. 1990. A constitutive model for partially saturated soils. Gotechnique 40(3): 405430. Cardoso, R. 2009. Hydro-mechanical behaviour of compacted marls. PhD Thesis. Instituto Superior Tcnico, Universidade Tcnica de Lisboa. Cardoso, R. & Alonso, E.E. 2009. Degradation of compacted marls: a microstructural investigation. Soils and Foundations 49(3): 315327. Della Vecchia, G. 2009. Coupled hydro-mechanical behaviour of compacted clayey soils. PhD Thesis, Politecnico di Milano. Kawai, K., Kato, S. & Karube, D. 2000. The model of water retention curve considering effects of void ratio. In H. Rahardjo, D. G. Toll & E. Leong (eds), Unsaturated Soils for Asia: 329334. Rotterdam: Balkema. Romero, E. & Vaunat, J. 2000. Retention curves of deformable clays. In A. Tarantino & C. Mancuso (eds), Experimental Evidence and Theoretical Approaches in Unsaturated Soils: 91106. Rotterdam: Balkema. Van Genuchten, M.T. 1980. A closed-form equation for predicting the hydraulic conductivity of unsaturated soils. Soil Sci. Soc. Am. J. 44: 892898. Vanapalli, S., Fredlund, D. & Pufahl, D. 1999. The influence of soil structure and stress history on the soilwater characteristics of a compacted till. Geotchnique 49(2):143159.
Many traditional models for water retention curve were developed more as fitting equations rather than built on a sound physical basis. Although they may provide good approximations to the data used in the calibration stage, in general they do not have good predictive capabilities with reference to coupled hydro-mechanical paths, especially for what concerns weathering induced by drying-wetting cycles. This makes difficult to incorporate them in constitutive models for unsaturated soils. When a more physically based approach is adopted, relying on the experimental evidence of the pore structure evolution, the predictive capabilities of the models are greatly enhanced at a reduced number of parameters. Here the approach was applied to model the retention behaviour of a natural marl from Portugal, subjected to dryingwetting cycles. An expression, which was previously proposed to model the retention domain of moderately active compacted clays, was adapted to account for the evolving pore size distribution of the marl in hydraulic cycles. The expression proposed considers the contribution of water in the volume changes of the matrix increasing the macroporosity of the material. The role played by microcracks in the low suction range will be the object of further examination.
1457
Effects of crude oil pollution on the shear strength behaviour of an unsaturated porous soil medium
M. Mohammadi, S.S. Yasrebi & A. Mirzaii
ABSTRACT: Crude oil leakage to the soils from facilities related to the oil industry can cause destructive environmental pollution effects. Additionally, oil contamination may also alter the mechanical characteristics of geomaterials such as shear strength behaviour which is requisite for assessment of numerous engineering problems. In the past, limited researches have been done to analyze the shear strength behaviour of oil contaminated soils while major of them were carried out in the saturated state. The objective of this research is to study the shear strength behaviour of polluted porous soil media both in saturated and unsaturated conditions. For this, a clayey-sand and one type of crude oil were selected, and a series of direct shear tests were carried out in saturated and unsaturated conditions. The contaminated samples were prepared by mixing the soil with 6% crude oil. Unsaturated direct shear tests were performed in the initial matrix suction of 30 kPa on specimens by using a novel unsaturated direct shear apparatus designed and built in Tarbiat Modares University, Iran. The results indicate an increase in the shear strength and cohesion and a slight reduction in internal friction angle of crude oil contaminated specimens while oil contamination content is increased. Keywords: Polluted unsaturated soils, oil contamination, shear strength behaviour, direct shear test. 1 INTRODUCTION and repulsive forces at particle level of clay-organic fluid systems. They showed the variation of the normalized shear strength of Kaolinite samples saturated with different pore fluids. It was also observed that the value of cohesion was increased with decreasing the dielectric constant of the pore fluid. Meegoda and Ratnaweera (1994) studied the compressible behaviour of contaminated finegrained soils through a series of consolidation tests on clays with low and high plasticity. They employed glycerol and propanol chemicals as contamination fluid. The results showed an increase in compressibility affected by mechanical and physicochemical factors. Bases on the results, they proposed correction factors to account the change in compression index caused by viscosity of the pore fluid. Alsanad et al. (1995) carried out a laboratory testing program to determine the influence of oil contamination on the geotechnical properties of Kuwaiti sand. In this study, they used artificially contaminated sand with different contamination contents to evaluate the effects of oil on the strength parameters, compressibility, permeability, and compaction characteristics of the sand. The influence of the type of oil, and relative density was also investigated by direct shear tests. The results indicated a small reduction in strength and permeability, thus an increase in compressibility due to the existence of the contamination.
Crude oil leakage to the soils from pipes, tanks, and refineries related to the oil industry can cause destructive environmental pollution effects. When oil spill occurs, soils might be contaminated by the leakage. In connection with the cleanup works and for any possible applications, knowledge of the geotechnical properties and behaviour of contaminated soils is required. This information is also required when oil leakage from storage tanks and processing plants cause oil contamination in the surrounding soils. In this case it is necessary to determine the effect of oil contamination on existing structures. Thus, few studies are carried out to assess geotechnical properties of contaminated soils. Moore and Mitchell (1974) proposed an analytical model taking into account the electromagnetic forces of interaction between particles. These forces of interactions are function of the dielectric constant of the pore fluid. They reported that a reduction in dielectric constant of pore fluid reduces the physicochemical interaction in the clay pore fluid electrolyte system, and hence should result a higher strength. Sridharan and Rao (1979) also presented an experimental study to determine the mechanisms controlling the shear strength of saturated clays. The authors introduced a new effective stress based concept by taking into account the electrical attractive
1459
Shin et al. (1999) studied the variation of shear strength and ultimate bearing capacity of shallow foundations above the sand contaminated by the crude oil. The bearing capacity tests were conducted with small-scale models in the laboratory. Based on the results, the appearance of oil contamination caused a decrease in the foundation bearing capacity. Shin and Das (2000) also studied the compaction, hydraulic conductivity, and shear strength of a sand contaminated with different oil pollutants. Based on the results, oil contaminations increased the compressibility of sand whereas it reduced the hydraulic conductivity and shear strength. It was also observed that the effect of the pollutants viscosity was the key affecting parameter on shear strength behaviour, and contaminations with higher kinematic viscosity caused more decrease in shear strength. Puri (2000) investigated the effect of oil contamination on the compressibility, shear strength, consolidation, and hydraulic conductivity of a sandy soil. Similar to preceding studies, the angle of internal friction of the sand tended to decrease with the presence of oil in the pore spaces. Ratnaweera and Meegoda (2006) conducted a series of unconfined compression tests on finegrained soils contaminated with varying amounts of chemicals. The results showed a decrease in shear strength and stress-strain behaviour due to the presence of contamination. These observations were attributed to changes in dielectric constant and pore fluid viscosity as a lubricant agent. Despite to the physicochemical effects caused by a reduction in dielectric constant, mechanical interactions at particle contacts, caused by enhanced lubrication of viscous pore fluids, caused a reduction in shear strength. Rehman et al. (2007) carried out a comprehensive laboratory testing program to compare the engineering properties of pure and contaminated clays. Crude oil was chosen as the contaminant fluid, and the clay samples were taken from the AlQatif area located in the east of Saudi Arabia. Test results showed that the contaminated clay behaved more like a cohesionless material, owing to the formation of agglomerates in the clayey soil. Khamehchiyan et al. (2007) studied the effects of crude oil contamination on geotechnical properties of clayey and sandy soils such as CL, SM and SP by conducting a series of different laboratory tests on clean and contaminated soil samples at the same densities. The results indicated a decrease in the strength, permeability, maximum dry density, optimum water content, and Atterberg limits. The major of previous studies were conducted in the polluted soil media in saturated condition. Thus in many geotechnical problems due to the presence of air in the void spaces, the application of saturated geotechnical properties will not represent diverse aspects of real soil behaviour. Nowadays,
most of contaminated geomaterials are found to be in gaseous areas and an appropriate study of their mechanical behaviour must be carried out within the principles of unsaturated porous media. The mechanical behaviour of unsaturated soils has been of interest during the past decades, and diverse engineering aspects of partially saturated soil media have been investigated by the researchers such as shear strength, volumetric behaviour, constitutive modeling, micro scale behaviour, and hydraulic and thermal couplings. As a new area, unsaturated behaviour of multi-phase geomaterials (e.g. oil, air, and water in analysis) has been of research interest in recent years. Shear strength behaviour of unsaturated soils has been studied in the works of many researchers to demonstrate a general shear strength equation for three phase media (air, water, and soil skeleton), thus the mechanical behaviour of unsaturated contaminated multi-phase geomaterials was not studied significantly. The objective of this research is to study shear strength behaviour of a polluted porous soil medium both in saturated and unsaturated conditions. A series of direct shear tests were conducted on saturated and unsaturated kaolin-sand mixtures to assess shear strength behaviour of understudying soil in pure and contaminated conditions for different saturation conditions. Saturated drained direct shear tests were carried out with a conventional Terzaghi type direct shear apparatus while constant water content unsaturated tests were performed with a novel direct shear apparatus designed and built in Tarbiat Modares University. 2 MATERIAL PROPERTIES
A soil mixture of 40% kaolin and 60% sand was used for this study, and crude oil was used as the fluid contaminant. The Sieve analysis of understudying soil was conducted according to ASTM D42263. The particle size distribution of the soil mixture is shown in Figure 1. Classification properties of understudying soil mixture are also shown in Table 1. The crude oil was used as the contaminant and was taken from Tehran Oil Refining Co, and its properties are represented in Table 2. Two sets of contaminated and uncontaminated compacted soil samples were prepared for saturated and unsaturated direct shear tests. Uncontaminated soil samples were statically remolded and compacted in the test mold and brought to desired initial conditions (Table 3). The amount of crude oil in contaminated soil samples was chosen to be 6% of total soil sample weight. Initially, contaminated multi-phase soil was
1460
Figure 1. soil. Table 1. soil.
Particle size distribution of understudying
Classification properties of understudying SC 9.5 2.66 40 9.58 19.71
samples. In these tests, soil samples were consolidated in vertical confining stresses of 50, 100, and 200 kPa and were horizontally loaded, and it was continued until soil samples reached to the failure. Unsaturated direct shear tests were carried out with a novel direct shear apparatus designed and built in Tarbiat Modares University. These tests include 6 constant water content direct shear tests on contaminated and uncontaminated soil samples. Initially, all samples were brought to the initial matrix suction of 30 kPa by using axis translation technique and were parallel consolidated in net normal stresses of 50, 100, and 200 kPa. At the end of equilibrium and consolidation stage, all samples were horizontally sheared in constant water content condition. 3.1 Saturated direct shear tests
Soil type: (unified system) Plasticity index % Specific gravity (Gs) Clay percent % Optimum water content % Maximum dry unit weight (kN/m3) Table 2. Crude oil properties.
Specific gravity at 15C (kN/m3) API gravity at 15C Viscosity (N s/m2) Salt content (lb/1000 bbl) Dielectric Constant Table 3. Initial properties of soil samples.
8.56 33.76 2.02 12 2.1
d (kN/m3) ww(%) Sr(%) e
17.17 9.6 50 0.51
Shear stress (kPa)
put in a close container for one week sufficiently for occurrence of any possible reaction between soil and crude oil. After one week, the moisture content of contaminated soil was increased to desired initial conditions (see Table 3) by adding distilled water to the soil mixture. Compacted polluted soil samples were then prepared by static compaction with regard to the desired condition in Table 3. 3 DIRECT SHEAR TEST PROCEDURES
Saturated direct shear tests (ASTM-D308072) were carried out to assess the effect of oil contamination on the shear strength behaviour understudying soil. These tests were performed in a cubical shear box with dimension of 6 6 2 cm with the shear speed of 0.5 mm/min in effective normal stresses of 50, 100, and 200 kPa. At the end of consolidation stage, horizontal shear loads were applied to the samples with the shear speed of 0.5 mm/min, and it was continued until soil samples reached to the failure. Figure 2 shows the variation of shear stress versus horizontal displacement for uncontaminated and contaminated saturated direct shear tests during the shear. Based on the results, it is observed that contaminated soil samples appeared to sustain more shear stress in compare to pure samples with similar testing conditions. Shear strength parameters are calculated for contaminated and uncontaminated soil samples in the saturated conditions with corresponding to the maximum shear stresses, and their values are shown in Table 4. Further investigation of the results emphasizes a slight decrease for internal angle of friction in contaminated samples while a
160
Confining Stress= 200 kPa
140 120 100 80 60 40 20 0

Confining Stress= 50 kPa Contaminated samples Uncontaminated samples Confining Stress= 100 kPa
A series 12 of direct shear tests were carried out on uncontaminated and contaminated compacted soil samples in different saturation conditions. 6 saturated conventional direct shear tests were performed on compacted pure and polluted soil
Figure 2. Plot of shear stress vs. horizontal displacement for saturated uncontaminated and contaminated samples during the shear stage.
1461
Table 4. Saturated Shear strength parameters from direct shear tests for pure and contaminated samples. Crude oil % 0% 6% c (kPa) 11.3 42.1 (Degree) 28.4 26
significant increase is observed for effective cohesion in polluted soil samples. As shown in Table 4, the value of effective cohesion is 11.3 kPa for uncontaminated samples while effective cohesion is 42.1 kPa when the oil content increases to 6%, and the value of effective cohesion in polluted saturated samples seems to be about 272% more than of its value in the pure condition. Thus, a slight decrease is observed in the value of internal angle of friction in polluted soil samples. The internal angle of friction is 28.4 for uncontaminated samples, and its value tends to decrease to 26 when the oil pollutant increases to 6%, and the value of internal angle of friction reduced 8% in contaminated soil samples. 3.2 Unsaturated direct shear tests A series of 6 unsaturated constant water content direct shear tests were carried out on compacted soil samples in pure and contaminated conditions. These tests were carried out with a novel unsaturated direct shear apparatus with using axis translation technique (Figure 3). At the end of equilibrium and consolidation stage, pure and polluted samples were horizontally sheared until they reached to the failure condition. Constant water content tests were carried out on contaminated and uncontaminated compacted samples with a novel direct shear apparatus designed and built in Tarbiat Modares University. Initially, the soil matrix suction brought to 30 kPa by application of axis translation technique. In the mean time, soil samples were consolidated in net confining normal stresses of 50, 100, and 200 kPa. Contaminated and uncontaminated samples were then horizontally sheared with the application of horizontal shear loads with the loading speed of 0.009 mm/min. During the shear stage, pore air was in drained condition while pore water was kept undrained and consequently caused a variation in the level of matrix suction. The shearing stage was continued until soil samples reached to the failure. The variation of shear stress versus horizontal displacement is plotted in Figure 4 for unsaturated direct shear tests for contaminated and uncontaminated samples in different net normal stresses. The variation of matrix suction is also plotted for unsaturated tests during the shear stage in Figure 5.
Figure 3. General layout of unsaturated direct shear apparatus.
Figure 4. Plot of shear stress vs. horizontal displacement for unsaturated uncontaminated and contaminated samples during the shear stage.
Figure 5. Plot of Matrix suction vs. horizontal displacement for unsaturated uncontaminated and contaminated samples during the shear stage.
Further investigation of Figure 4 emphasizes that contaminated samples tended to sustain higher values of shear stress than uncontaminated samples in similar testing conditions. Thus, contaminated samples tended to have less increment in the value of matrix suction than uncontaminated soil during the shear (Figure 5). Based on the results, shear strength parameters are calculated for understudying soil corresponding to the maximum shear strength condition, and they are represented in Table 5. These values are
1462
Table 5. Shear strength parameters attained from unsaturated shear tests on pure and contaminated samples. Contaminated Net Stress* 50 (Deg) c C ua uw b (Deg) 100 26.4 42.1 59.9 51.5 19 200 UnContaminated 50 100 29.9 11.3 28.5 63.4 15.2 200
of crude oil also increased the apparent cohesion in the unsaturated shear tests, similar to the increase of effective cohesion in saturated tests. The increment of apparent cohesion has also slightly increased the value of b in a nonlinear pattern. 4 INTERPRETATION OF THE RESULTS
45.6 21.3
45.3 58.1 21.4 16.6
57.1 16.8
evaluated in accordance to the extended Mohr Coulomb failure envelope for unsaturated soils suggested by Fredlund et al. (1978) as: c ( ua) tan (ua uw)tan b (1)
where is the shear stress, c is the effective cohesion, ( ua) is the net normal stress, is the effective angle of internal friction, (ua uw) is soil matrix suction (difference of pore air and water respectively), and b is the angle of internal friction with respect to matrix suction. It is also evident that the shear failure envelope suggested by equation (1) intersects shear stress versus matrix suction plane along a line of intercepts as: C c (ua uw)tan b (2)
where C is the total or apparent cohesion. The value of apparent cohesion is also calculated for understudying soil, and they are represented in Table 5. The value of was assumed to be constant by Fredlund et al. (1978) for all values of matrix suction and was taken to be equal to that measured in saturated conditions. However the use of a linear relationship between and ua uw and consequently a constant value for b was shown to be in error by Escario & Saez (1986). This nonlinearity was confirmed by Fredlund et al. (1987) who assumed b varied as a function of matrix suction. Thus, the values of b are calculated for each value of matrix suction and are represented in Table 5 for constant water content direct shear tests. In this calculation, effective cohesion is taken from conventional direct shear test results shown in Table 4. As shown in Table 5, the value of internal angle of friction is decreased in contaminated unsaturated soil samples. A similar behaviour was also seen in the saturated conditions. Thus in case of zero suctions (e.g. fully saturated condition), the existence of crude oil, as a lubricant agent in saturated pore spaces, slightly decreased the value of effective internal angle of friction. The appearance
In preceding sections it was seen that the appearance of crude oil caused an increase in the value of sustainable shear stress during the application of horizontal shear loads. This behaviour was observed both in saturated and unsaturated direct shear tests. In other words, the existence of crude oil caused an increase in the level of cohesion in contaminated soils in different saturation conditions while it reduced the value of friction angle in polluted samples. These phenomena can be interpreted in detail with consideration on viscosity of the crude oil, the dielectric constant, and capability of crude oil to act as a lubricant agent between soil particles. The observed increase in the value of effective and apparent cohesion of saturated and unsaturated contaminated soil samples can be explained with regard to the dynamic viscosity of the crude oil and its distribution in the soil pore spaces of saturated and unsaturated soil samples. According to Rehman et al. (2007) when contaminated clayey soils are saturated with distilled water, the water dissociates the oil-soil particles bonds and surrounds the soil particles. In this case, the contaminant fluid (e.g. crude oil) occupies remained volume of pore voids, particularly in fully saturated soils. This phenomenon is schematically illustrated in Figure 6. In this case, the oil acts as the main fluid in the pore spaces in occasion for saturated soils. Typically, crude oils are more viscose fluids than water, and for understudying case the dynamic viscosity of the crude oil is almost 20 times greater than the dynamic viscosity of water. The term viscosity is generally referred to the capacity of fluids to bear shear stress. A brief comparison of crude oil and water dynamic viscosity emphasizes significant capability of crude oil to resist against horizontal shear forces than the water. For the case of saturated condition, the crude oil contaminant has typically less contact with soil
Figure 6. Schematic layout of the dissociation of oilsolid soil particles bond with water.
1463
solid particles and can bear more shear forces as it is freely allocated in the pore spaces. Thus for the case of unsaturated condition, the distribution of crude oil in the pore spaces may create compound films of oil- soil particles and water-soil particles bonds and has a less effect on the value of sustainable cohesion during the application of shear forces. The reduction of the dielectric constant of the pore fluids also increased the value of effective and apparent cohesion in contaminated soil samples. Typically when clayey soils are contaminated, dielectric constant of the pore fluid reduces and consequently causes a decrease in the chemical interactions in the soil pore fluid electrolyte system (e.g. Moore and Mitchell 1974and Sridharan and Rao 1979) which causes an increase in the shear strength. For the case of understudying soil, 40% of the soil mixture is composed of kaolin which is light ionic clay and the presence of oil contamination has affected the repulsive potential and consequently increased the value of maximum sustainable shear forces. It was also observed that the internal angle of friction is also continuously decreased with the increment of contamination content. This behaviour is similar to the most of existing literature data for the behaviour of saturated crude oil contaminated grained geomaterials already cited in the literature. In fact, the existence of crude oil in the pore spaces acts as a lubricant agent and concurrently reduces the value of internal angle of friction and approves the compressibility of soil structure which has already been stated by Meegoda and Ratnaweera (1994) and Meegoda et al. (1998). 5 CONCLUSIONS
REFERENCES
Alsanad, H.A., Eid, W.K. & Ismael, N.F., 1995. Geotechnical properties of oil contaminated Kuwaiti sand. Journal of Geotechnical Engineering, ASCE 121 (5), 407412. Alsanad, H.A. & Ismael, N.F., 1997. Aging effect on oil contaminated Kuwaiti sand. Journal of Geotechnical and Geoenvironmental Engineering 290294. ASTM (American Society for Testing and Materials), 1999. Annual book of ASTM standards. Section 4, Construction. Soil and Rock, vol. 4.08. ASTM, Philadelphia, PA. Escario, V. & Saez, J. 1986. The shear strength of partly saturated soils. Some aspects of effective stress in saturated and partly saturated soil. Gotechnique 36(3): 453456. Fredlund, D.G., Morgenstern, N.R. & Widger, R.A. 1978. The shear strength of unsatuarted soils. Can. Gotech. J. 15(3): 177197. 313321. Fredlund, D.G., Rahardjo, H. & Gan, J.K.M. 1987. Nonlinerity of strength envelope for unsaturated soils. Proc. 6th Int. Conf. expansive soils. New Dehli, 4954. Khamehchiyan, M., Charkhabi A.H. & Tajik, M., 2007. Effects of crude oil contamination on geotechnical properties of clayey and sandy soils. Journal of Engineering Geology (89), 220229. Maaitah, O.N. & Tarawneh, S.A., 2003. Effect of treated waste water on the behavior of unsaturated soil. Pakistan Journal of Applied Sciences 3 (5): 360369. Meegoda, N.J., Chen, B., Gunasekera, S.D. & Pederson, P., 1998. Compaction characteristics of contaminated soils-reuse as a roadbase material. In: Vipulanandan, C., Elton, David J. (Eds.), Recycled Materials in Geotechnical Applications. GeotechnicalSpecial Publication, vol. 79. ASCE, pp. 165209. Meegoda, N.J. & Rajapakse, R.A., 1993. Long-term and short-term hydraulic conductivities of contaminated clays. ASCE Environmental Engineering Journal 119 (4), 725743. Meegoda, N.J. & Ratnaweera, P., 1994. Compressibility of contaminated fine grained soils. Geotechnical Testing Journal 17 (1), 101112. Moore, C.A. & Mitchell, J.K., 1974. Electromagnetic forces and soil strength. Geotechnique 24(4), pp. 627640. Puri, Vijay, K., 2000, Geotechnical aspects of oil-contaminated sands. Journal of Soil and Sediment Contamination, 9(4):359374. Ratnaweera, P. & Meegoda, J.N. 2006. Shear strength and stressstrain behavior of contaminated soils. ASTM Geotechnical Testing Journal 29 (2), 133140. Rehman, H.U, Abduljauwad, S.N. & Akram, T., 2007. Geotechnical behavior of oil-contaminated finegrained soils, Electronic Journal of Geotechnical Engineering Vol. 12, No. A, 2007, pp. 112. Shin E.C. & Das, B.M., 2000. Some physical properties of unsaturated oil- contaminated sand, Proceedings of sessions of Geo-Denver, Denver, Colorado, Geotechnical Special Publication No. 99, 142152. Shin, E.C., Lee, J.B. & Das, B.M., 1999. Bearing capacity of a model scale footing on crude oil-contaminated sand. Journal of Geotechnical and Geological Engineering, Kluwer Academic Publishers (17), 123132, Sridharan, A. & Rao, V.G., 1979. Shear strength behavior of saturated clays and the role of the effective stress concept. Geotechnique 2, 177193.
An extensive laboratory testing program was carried out to study the effect of crude oil contamination on the shear strength properties of a kaolin-sand mixture in saturated and unsaturated conditions, and a series of saturated and unsaturated direct shear tests were carried out on pure and polluted compacted soil samples. The crude oil was used as the contamination fluid. The results outlined the tendency of contaminated soils to sustain more shear stress than similar uncontaminated samples both in the absence and presence of air in the void spaces. The appearance of crude oil increased the value of effective and apparent cohesion in contaminated compacted samples while it slightly decreased the value of internal angle of friction. The increment of the apparent cohesion also increased the value of basic angle of friction in a nonlinear pattern. In the following, these phenomena were interpreted with consideration on the viscosity of the crude oil, dielectric constant, and capability of crude oil to act as a lubricant agent between soil particles.
1464
Effects of hydraulic cycling on the stiffness response of a rigid clay

J.A. Pineda, G. Mitaritonna, E. Romero & M. Arroyo
ABSTRACT: This paper presents the results of an experimental investigation aimed at evaluating the effects of hydraulic cycling on the small-strain stiffness of a rigid Jurassic clay (Opalinus clay, Switzerland). This clay was previously subjected to thermal loads during an in situ heating experiment (HE-D, Gens et al. 2007). After this in situ test, intact samples were retrieved and analysed at laboratory scale. In a first stage, the influence of thermal loads was studied to evaluate the initial state of the samples prior to hydraulic cycling. In a second stage, samples at different distances to the heater were subjected to wetting and drying paths using vapour equilibrium technique. The hydraulic effects were tracked at macroscopic scale using basic characterisation (water content and porosity) and non-destructive techniques (ultrasonic pulses and bender elements). Test results showed a noticeable change of stiffness with total suction. An expression to account for seismic velocity changes with total suction is proposed. 1 INTRODUCTION 2 OPALINUS CLAY AT MONT TERRI UNDERGROUND LABORATORY
It is a common observation in civil engineering works that the strength and stiffness of sedimentary fine-grained rocks, like shales, mudstones and claystones have a strong dependency on their hydraulic state. Under controlled laboratory conditions this dependency shows clear hysteretic patterns: cyclic desaturation and resaturation paths result in irreversible changes in the rock matrix. These changes can be observed at both microstructural and macroscopic specimen-scale. Due to their specific characteristics, such as low permeability and self-healing capacity when fractured, these formations have been considered as potential host geological media for underground repositories of high-level radioactive waste (Gens et al. 2007). Within this context, Opalinus clay has played a major role in understanding coupled thermo-hydro-mechanical processes (e.g., in situ heating test HE-D, Gens et al. 2007). This paper addresses the results of an experimental programme aimed at evaluating the influence of environmental actions (wetting-drying cycles) on the stiffness degradation of Opalinus clay. Undisturbed samples were retrieved after the HE-D experiment at different distances from the heater. After an initial characterisation study, samples were subjected to hydraulic cycles to study the evolution of the small-strain stiffness. Bender element and ultrasonic pulse tests were used to track the stiffness variations at different hydraulic and degradation states.
Table 1 summarises the main physical properties of Opalinus clay, a stiff overconsolidated clay of Middle Jurassic period, at the Mont Terri Underground Laboratory site (Switzerland). It contains between 40 to 80% clay minerals (illite, smectite, chlorite, kaolinite), besides carbonates (calcite, dolomite, ankerite and siderite) and quartz (Bossart et al. 2002). At Mont Terri, overburden varies between 250 and 350 m with closely spaced bedding planes at an angle of dip around 45 (Fig. 1a). Samples tested in this work were obtained from a horizontal borehole (90-mm in diameter), directed from a niche towards the heating test HE-D. A clear damaged zone, with lower stiffness and higher permeability, was identified by numerical back-analysis at this test area (Gens et al. 2007). Some parameters show differences depending on the orientation, reflecting the anisotropy caused by the intense bedding of the clay. Samples were obtained by dry-coring technique to reduce alteration on preparation of this water sensitive material, and cored along two main directions to analyse the influence of the bedding planes on the rock stiffness (Fig. 1b): a) perpendicular to the borehole axis, so that bedding planes showed their natural angle of dip ( 45); and b) inclined to the borehole axis with horizontal bedding planes at 90.
1465
Table 1. Property
Main physical properties of Opalinus clay.

Water content (%)
7,5 7 6,5 6 5,5 5 4,5 4 -6 -4 -2 0 2 4 6
Samples retrieved 2.382.42 2.252.27 5.86.5 2230 2.72 0.20 0.17 0.790.88 41 24
Average values 2.312.52 2.222.33 4.28.0 1016* 2.73* 0.160.22 0.140.18 0.830.93* -
heater
Density, (Mg/m3) Dry density (d (Mg/m3) Water content, w (%) Initial total suction, (MPa) Density of solids, s (Mg/m3) Void ratio, e Porosity, n Degree of saturation, Sr Liquid limit, wL (%) Plasticity index, PI (%)
Series1 =90 =45 lug-08
Horizontal distance from the heater borehole (m)
Figure 2. Water content distribution of Opalinus clay specimens.

0,4 Microporosity
Gens et al. (2007)
Muoz (2007)
Pore size density function, (-denw/d(logx)
0,35 0,3 0,25 0,2 0,15 0,1 0,05 0 1 10 100 1000 10000
-3.1m from heater -0.5m from heater 0.5m from heater 0.5m from heater
Ultramicroporosity
15nm
Mesoporosity
Macroporosity
Figure 1. (a) Opalinus clay bedding planes orientation under in situ conditions. (b) Specimen orientation for testing.
Figure 3. Pore size distributions of Opalinus clay specimens at different distances from the heater.
3 3.1
EFFECTS OF THE PREVIOUS HEATING TEST Water content distribution around the heater
Before studying the influence of hydraulic cycles on the mechanical-hydraulic properties of Opalinus clay, the effects of the previous heating process on samples located at different distances from the heater were evaluated. Figure 2 shows the variation of water content with the horizontal distance from the heater. A reduction of water content at distances closer to the heater was clearly detected after the thermal load. 3.2 Pore size distribution Mercury intrusion porosimetry tests were carried out on freeze-dried samples to characterise the pore network distribution using an AutoPore IV 9500 Micromeritics Instrument Corp. porosimeter. Figure 3 presents the pore size density functions. The graph shows a dominant pore mode at around 15 nm, as expected for the matrix of an argillaceous rock. No appreciable changes in the pore size density functions were observed for samples located at different distances from the
heater. This fact indicates that the previous heating test, which was associated with some drying of the material, did not induce important changes at microstructural scale, at least within the pore sizes detected by this technique. 3.3 Water retention properties
The water retention curve (hereafter WRC) of different specimens, expressed as total suction versus water content along a wetting-drying-wetting cycle under unconfined conditions, was obtained by a chilled-mirror dew-point psychrometer (WP4, Decagon Devices Inc, 2003). Vapour equilibrium technique using lithium chloride was used for reducing relative humidity down to 15%. Wetting paths were also applied by vapour equilibrium, using distilled water. Calibration and test procedures have been performed according to Cardoso et al. (2007). Figure 4 shows the drying and wetting branches of the WRCs for three different samples located at 3.1 m, 0.5 m and 0.5 m from the heater. Although no significant differences between curves are observed, there is some small increase in the water absorption capacity for the sample located at 0.5 m from the heater.
1466
1000 Initial states
100
Drying
10
Wetting
-3.1m from heater -0.5m from heater 0.5m from heater 0 2 4 6 8 10 12 14
0,1
dotted ellipse in Figure 2 encloses a common value of water content for two specimens with different bedding plane orientations, in which clearly 90 displayed a lower value (dotted ellipse in Fig. 5). Despite the water content decrease when moving closer to the heater, Figure 5 shows a consistent decrease in stiffness at points closer to the heater (particularly samples at 90) that is associated with some degradation induced on this heating episode. 4 EFFECTS OF HYDRAULIC CYCLING
Total suction (MPa)
Figure 4. Water retention curves for Opalinus clay specimens retrieved at different distances from the heater.
3000
4.1 Hydraulic tests performed

heater
2500 2000 1500 1000 500 0 -6 -4 -2 0
Vort =90 Vv =45
Horizontal distance from the heater borehole (m)
Figure 5. Variation of longitudinal velocity at different orientations and horizontal distances from the heater.
3.4
Longitudinal wave velocity measurements
The stiffness of the material under unloaded conditions was evaluated through the longitudinal wave velocity using an ultrasonic pulse device (V-Meter). This device is used to determine the travel time (tp) of a high frequency (54 kHz) and high voltage compression wave travelling alongside the sample axis. By knowing the travel length of the compression wave (l, sample height), the longitudinal wave velocity (VL) and the unconfined Young modulus (E) were determined as ( is the total density): VL l ; tp
2 E VL
(1)
Figure 5 show the variation of VL for samples with different bedding plane orientations () and located at different horizontal distances from the heater. The lower values of VL were obtained for samples with horizontal bedding plane orientation 90, which is consistent with the scheme in Figure 1b (lower stiffness direction matching up the axis). However, these results were also affected to some extent by the saturation of the sample. The
The influence of hydraulic cycling on the stiffness response of Opalinus clay was studied by applying two wetting-drying cycles using vapour equilibrium technique on two specimens, S-1 and S-2. Specimens S-1 and S-2 were trimmed perpendicular to the borehole direction ( 45) at distances 0.2 m and 0.5 m from the heater, respectively. Specimens with diameter 38 mm and height between 55 and 75 mm were equilibrated under controlled relative humidity (RH) in desiccators. During equalisation both porosity and water content were monitored using volume and mass measurements. Figure 6 (on the left) shows the time evolution of the water content along the hydraulic paths followed, whereas in Figure 6 (right) the combined measurements in terms of porosity and water content evolutions are presented. Lines with the same degree of saturation are also indicated in Figure 6 (right). Wetting paths were maintained for 33 and 50 days during the first and second cycles, respectively. Drying proceeded at a slightly faster rate, and was maintained for 30 and 45 days during the first and second cycles. The first wetting path led both specimens to similar final water contents, while for the second wetting path sample S-2 presented a slightly larger water storage capacity at saturation. The water contents at the end of the drying stages (first and second cycles) were around 2.93.4% and 3.53.9% for samples S-1 and S-2, respectively. These values corresponded to total suctions around 95110 MPa and 7590 MPa, for samples S-1 and S-2, as estimated from the WRCs (Fig. 4). As observed in Figure 6 (right), sample S-1 underwent a larger porosity change during the application of the two hydraulic cycles. It appears that the proximity to the heater has induced a larger damage on this sample. 4.2 Determination of elastic parameters The evolution of the stiffness along the hydraulic paths was tracked by two non-destructive methods
Longitudinal wave velocity, VL (m/seg)
1467
0.19 8
Sample S-1 Sample S-2
0.185 0.18
Water content (%)
7 6 5 Initial state 4 Initial state
0.175 0.17
0.165 0.16 0.155
3 0 15 30 45 60 75 90 105 120 135 150 3
Time (days)
Water content, (%)
Sr = 6 0%
Sr = 1 00%
Sr = 8 0%
Figure 6. cycles.
(Left) Time evolution of water content. (Right) Porosity and water content evolutions along the hydraulic
2400
(bender element transducers and ultrasonic pulse tests previously described). These two methods allow continuously tracking the unconfined Young modulus E (along the vertical direction) and shear modulus Gvh (vertical propagation direction and horizontal wave polarisation direction), through the measurements of the elastic wave velocities (longitudinal) VL and (shear) VS, respectively. The small strain shear stiffness is calculated as G(ij) tV2S(ij), where t is the total density, and VS(ij) li/tS. The wave travel distance (li) is taken as the distance between transducer tips (Viggiani & Atkinson 1995). Arrival time (tS) is determined by inspection of the received trace, looking for the first significant amplitude excursion. Bender elements, measuring along the vertical axis, have been previously applied in rocks by Pineda et al. (2008). Details about the application of bender elements on rocks including their implementation in a new high pressure triaxial cell can be found in Pineda (2010). The evolution of the shear wave velocity in vertical direction (VS(vh)) with water content is shown in Figure 7. The results of the ultrasonic pulse tests for determining the longitudinal wave velocity (VL) are also reported. Systematically lower values for both longitudinal and shear wave velocities were obtained for sample S-1, which was the sample placed closer to the heater. Again, it appears that sample S-1 has undergone a larger damage due to the proximity of the heater. As wetting preceded, both longitudinal and shear wave velocities consistently decreased. This decrease was smaller for sample S-1. During wetting and drying stages a quasi-linear tendency with water content was observed for the shear wave velocity. On the other
Seismic velocities, VL - Vs(vh) (m/s)
2200
Initial state for VL
2000
1800
1600
Sample S-1
1400
First cycle: filled symbols Second cycle: open symbols
1200
Initial state for Vs
1000
3400
Sample S-2
3200
First cycle: filled symbols Second cycle: open symbols
3000
2800
Initial state for VL
2600
2400
2200
2000
Initial state for Vs

1800 3 4 5 6 7 8
Water content, (%)
Figure 7. content.
Variation of the wave velocities with water
1468
Porosity, n (-)
3500
(a) (b)
30
Opalinus clay
G(vh) Sample S-1
3000
Initial state sample S-2 Initial state sample S-2 Initial state sample S-2 Initial state sample S-1 Initial state sample S-1 Initial state for sample S-1
25
E Sample S-2
2500
20
2000 Initial state sample S-1
15
1500
10 Initial state sample S-2
1000
500 1
Total suction, (MPa)
10
100
0.16
0.165
Porosity, n (-)
0.17
0.175
0.18
0.185
Figure 8.
(a) Variation of wave velocities with total suction. (b) Evolution of elastic moduli with porosity. Table 2. Values of obtained during drying paths. First drying VL (m/s) S-1 S-2 88 230 VS (m/s) 58 187 Second drying VL (m/s) 150 210 VS (m/s) 81 212
hand, the response of the longitudinal waves showed some small hysteretic behaviour, which is associated with its influence on the degree of saturation. For this reason, Vs measurements are more appropriate for studying structure changes on soils and rocks induced by hydraulic cycles. Water content data during the hydraulic cycles were converted into total suctions () using the different branches of the WRCs (Fig. 4). Figure 8a shows the variation of seismic velocities and elastic moduli with total suction and porosity. It can be seen that this variation of seismic velocities with total suction can be expressed as follows: VL, S ( ij ) ( ) VL, S ( ij ) ( 0 ) ln pref (2)
where VL,S(ij)( 0) is the value for longitudinal or shear wave velocities at zero total suction (Sr 1), a material parameter expressed in (m/s) and pref a reference pressure equal to 1 (using the same units of suction). Parameter can be associated with the sensitivity to total suction of the rock structure. Different values can be obtained along wetting and drying paths. Particularly, Table 2 summarises the variation of for both VL and Vs along first and second drying paths. Sample S-2, which was affected to a lesser extent by the thermal load (far from the heater), displayed a larger sensitivity of stiffness on total suction. This parameter increased along the second cycle for both samples, apparently linked to the degradation of the material that tended to a rock structure with a larger sensitivity to water content.
Figure 8b shows the evolution of the elastic moduli with porosity. By comparing the different values, it can be observed that the shear modulus for specimen S-1 (degraded sample closed to the heater) was almost four times lower than the corresponding value for specimen S-2 at the end of the drying. Again, sample S-2 showed more sensitivity to stiffness variations associated with porosity changes. Nearly reversible patterns were observed between rock stiffness and porosity changes for the two samples along the hydraulic cycles performed. This is not always the case, since an important irreversibility was observed by Pineda et al. (2008) on a claystone that underwent strong micro-fissuring on drying under unconfined conditions. 5 SUMMARY AND CONCLUSIONS
The research work presented in this paper investigated the influence of a previous heating test and subsequent hydraulic cycles on the stiffness response of Opalinus clay, a reference host geological formation for radioactive waste disposal. Hydraulic degradation was induced by applying
1469
Elastic moduli, E - G(vh) (GPa)
E Sample S-1 G(vh) Sample S-2
two wetting-drying cycles using relative humidity controlled conditions. Changes in elastic moduli induced by these hydraulic cycles were monitored by two non-destructive methods (ultrasonic pulses and bender elements). In addition, the hydraulic effects were tracked at macroscopic scale using volume and mass measurements (porosity and water content changes). Results showed that the previous heating test induced more degradation, in terms of loss in stiffness, in samples located near the heater. The samples placed closed to the heater were the ones undergoing a larger water content decrease. However, microstructural analyses by mercury intrusion porosimetry did not display major differences between samples located at different distances from the heater. It appeared that the thermal load induced rock degradation at macroscale level, i.e., macroscopic fissuring without changes in rock microstructure. Hydraulic test results showed a noticeable change of stiffness with total suction. An expression to account for seismic velocity changes with total suction was proposed. Hydraulic cycling induced an additional degradation of the specimens, which was detected by the increase in the sensitivity of stiffness on total suction.
REFERENCES
Bossart, P., Meier, P.M., Moeri, A., Trick, T. & Mayor, J.C. 2002. Geological and hydraulic characterisation of the excavation disturbed zone in Opalinus clay of the Mont Terri Rock Laboratory. Engineering Geology 66: 1938. Cardoso, R., Romero, E., Lima, A. & Ferrari, A. 2007. A comparative study of soil suction measurement using two different high-range psychrometers. Proc. Mechanics of Unsaturated Soils. T. Schanz editor, pp. 7994. Decagon Devices, Inc. 2003. WP4 Water Dewpoint Potentiometer. Operators Manual Version 2.2. Decagon Devices, Inc., Pullman, USA (www.decagon. com). Gens, A., Vaunat, J., Garitte, B. & Wileveau, Y. 2007. In situ behaviour of a stiff layered clay subject to thermal loading: observations and interpretation. Gotechnique 57(2): 207228. Jovicic, V., Coop, M.R. & Simic, M. 1996. Objetive criteria for determining Gmax from bender element tests. Gotechnique 46(2): 357362. Muoz, J.J. 2007. Thermo-hydro-mechanical analysis of soft rock: application to a large scale heating test and large scale ventilation test. PhD Thesis, Universitat Politcnica de Catalunya, Barcelona. Pineda, J.A., Arroyo, M., Romero, E. & Alonso, E.E. 2008. Dynamic tracking of hydraulically claystone degradation. Proc. 4th Int. Symp. Deformation Characteristics of Geomaterials. Burns, Mayne & Santamarina Eds., pp. 809816. Pineda, J.A. 2010. Swelling and degradation of argillaceous rocks. PhD Thesis, Universitat Politcnica de Catalunya, Barcelona. Viggiani, G. & Atkinson, J.H. 1995. Interpretation of bender element test. Gotechnique 45(1): 149154.
ACKNOWLEDGEMENTS The first author was supported by the EU ALBAN PROGRAMME grant number E04D027285CO.
1470
Effect of drying and wetting cycles on the shear strength of argillaceous rocks
R.D. Tovar
District Office for the Prevention and Attention of Emergencies, Bogot, Colombia Formerly at the National University of Colombia, Bogot D.C., Colombia
J.E. Colmenares
National University of Colombia, Bogot D.C., Colombia
ABSTRACT: Results of an investigation into the effects of drying and wetting cycles, on the shear strength of discontinuities, in natural argillaceous shales are presented. Shear strength in natural discontinuities was measured by triaxial and direct shear testing. Mohr-Coulomb shear strength envelopes were established for all wetting and drying cycles applied. After several cycles of drying and wetting the structure and geometry of the sample is lost, turning it into a particulate soil. This new material (particulate soil) was also tested to obtain a lower limit of resistance values. The results show that argillaceous rocks exhibit a big reduction in resistance when they are subjected to successive cycles of wetting and drying. Using a mineralogical analysis, it was found that the effect of degradation, caused by wetting and drying cycles, in the composition and proportions of the main chemical compounds of the rock, is not important. It revealed that the successive cycles of wetting and drying generates a physical degradation that takes place in very short periods of time and causes a considerable reduction of resistance that could explain the failure of natural deposits of these materials when they are exposed to atmospheric environmental changes. 1 INTRODUCTION strain energy from the diagenetic bonds. Low stored strain energy means low disintegration or degradation experienced by the rock. This concept can be explained considering that when the overload is removed, the material tries to find a balance with the new environment, releasing the excess of stored strain energy. The release of stored strain energy in conjunction with the exposition of the materials to environmental conditions, water and various chemical and biological agents lead to the alteration process (Alonso, E.E. & Pineda, J.A., 2006). The mechanisms of alteration of soils and rocks contain all the processes that produce hydromechanical degradation of geomaterials. Degradation is a destructive process that normally produces rocks of smaller sizes and different composition and shape in comparison with fresh rocks. Many authors (Taylor & Spears, 1981; Taylor & Smith, 1986; Taylor, 1988; Marques, 1992, among others) divide the degradation process in two types: physical and chemical degradation. Physical degradation is related to any process that produces fracturing without mineralogical changes, while chemical degradation is directly related to the presence of water which induces hydration and can produce more complex chemical reactions.
Recently, there has been an increased interest on the mechanical behaviour of soft rocks due to the large amount of geotechnical problems that such materials have caused on the infrastructure. Shales fall within the so-called soft rocks. This type of material in unaltered state exhibits good mechanical properties, however when exposed to atmospheric conditions those properties decrease considerably in a very short period of time, causing degradation and loss of strength. Soft rocks usually have an intermediate behaviour between rock and soil, and the fragility and complex behaviour of those rocks make it difficult to obtain samples with the necessary geometry to perform common laboratory tests to measure the shear strength. To understand the degradation process it is necessary to remind that the rock has been subjected to geological cycles that include depositation, geologic compaction, diagenetic changes, tectonic alteration, degradation and erosion. The combination of those processes defines the structure of the rock and controls its mechanical behavior. Bjerrum (1967) attributes the degradation of overconsolidated shales to the dissipation of stored
1471
According to Mitchell & Soga (2005), chemical processes include hydrolysis (reaction of minerals with hydrogen ions (H ) and hydroxide ions (OH) from the water), cation exchange, oxidation and carbonation. This paper is focused on studying the effect of wetting and drying cycles on shear strength of the shales, with the aim of establishing if cyclical changes in the moisture content of shale samples, resulted in the degradation of the rock. The results show that during the degradation process, significant changes in the structure of the samples are generated, leading to a significant decrease of resistance after a few wetting and drying cycles without significative mineralogical changes. 2 2.1 MATERIAL CHARACTERIZATION Basic properties
Table 2.
Values of point load index. Is (50) MPa 0.76 0.35
Test direction Same direction to bedding planes ( ) Perpendicular to bedding planes ()
point load index Is (50) and anisotropy index (Ia). The study materials are sedimentary rocks (shales) having well-defined bedding planes, therefore the point load resistance was conducted in the same direction of the bedding planes ( ) and perpendicular to these (). The results are showed in Table 2. The data showed in Table 2, allowed to obtain a value of anisotropy index (Ia) of 2.18. It confirms the anisotropic behavior of the material. The low values of Is(50) indicate that the material studied belongs to the group called soft rocks. 3 3.1 EXPERIMENTAL RESULTS Triaxial test for rocks
The samples used in this investigation were obtained from the basin of La Negra Creek, which is located northwest of Bogota DC, on the western side of the Eastern Mountain Range of Colombia in the Black River basin, belonging to the Magdalena River valley. The site is geographically located on the WGS84 system at 5 11 36.28 North latitude and 74 28 48.25 West longitude, between levels 2065 MASL and 497 MASL. From a geological point of view, the area consists of Cretaceous sedimentary rocks belonging to the Villeta group. This geological Formation is composed essentially by clayey rocks. The chemical and mineralogical characterization indicate that the rocks are composed of silica minerals (SiO2) and Aluminum oxide (Al2O3) represented in minerals such as Illite, Kaolinite, Gypsum and some secondary minerals such as Pyrite, Chlorite and Pyrophyllite. Table 1 shows some properties of the fine fraction, resulting from the decomposition of the rock such as natural moisture content (w), consistency limits (LL is the liquid limit, PI the plasticity index and PL is the plastic limit). Values of unit weight (t) and void ratio (eo) from fresh samples and slaking index (SI) according to the modified method of Woods & Deos (quoted by Sadisun et al, 2003) are also presented. 2.2 Point load test and anisotropy index
Three triaxial tests, were performed with samples on the least altered or intact conditions. Cell pressures of 9, 14 y 18 MPa were applied according to ASTM D 266404 standard and using a controlled stress cell - Hoek Cell (Hoek y Franklin, 1968). Samples were prepared from blocks using a rotary drill and diamond-tipped bit according to the procedure developed by Tovar (Tovar, 2008). The test results allowed to obtain a value of cohesion (C) of 493 kPa and a friction angle () of 38,4 grades. Figure 1 shows the results of stress strain behavior obtained from the test. Ductile behavior was observed when the sample is subjected to smaller cell pressures, while for greater cell pressures the behavior was fragile. This behavior is due to the fragile nature of the rock, which is deformed longitudinally slightly before reaching failure. At higher cell pressures, the rock has a fragile behavior but can deform more axially than radially before reaching the failure, confirming that the bedding planes are the weakest area of the rock. 3.2 Drying and wetting cycles
Point load test were performed, according to ASTM D 5731standard, in order to determine the
Table 1. Main properties of the material.
w (%) LL (%) PL (%) PI (%) t (kN/m3) SI (%) eo 3.6 28.31 18.13 10.17 25.5 410 0.212
Shale samples were protected with microporous tape in order to reduce energy gradients that could arise during the wetting stage. Wetting was applied by immersion of the samples in distilled water for five days (120 hours). After this period of time, samples reached a constant weight. The drying process took also 120 hours. It was performed using an oven at controlled
1472
70000 60000
1 - 3 (kN/m 2 )
50000 40000 30000 20000 10000 0
-0,0700
-0,0600
-0,0500
-0,0400
-0,0300
-0,0200
-0,0100
0,0000
0,0100
Axial strain (e =L/Lo)

s3 Mpa 3==99 MPa
3 = 14 MPa
3 18 = 18 MPa sc= MPa
Figure 1. Strain stress behavior from samples in triaxial tests.
Figure 2. studied.
Soil- water retention curves for the Shales
temperature: 48,5 1,5 degrees, trying to reproduce temperatures that samples can reach in their natural environment. The procedure is described in detail in Tovar (2008). Samples could expand freely throughout the process of wetting and drying and later they were placed in the equipment to be tested. Figure 2 shows the retention curve of a sample (diameter 6 cm., height 2 cm) subjected to several cycles of wetting and drying. Suction values were measured by the Filter Paper method as suggested by Ridley et al. (2003). In this research Wathman No. 42 filter paper was used. It was possible to apply four cycles of wetting and drying (N 4), before sample degradation. Each cycle included the process of wetting and drying (e.g. two cycles of W-D (N 2): W-D-W and D). 3.3 Direct shear test
Those cracks are responsible for a significant decrease of resistance. When the triaxial test was performed on samples after only 1 day of wetting, just the application of the confining pressure on the triaxial equipment caused the failure of the sample. In other cases the sample was so fractured, that it was impossible to place it in the Hoek cell (controlled stress) because it needs a uniform and continuous surface to place the strain gauges. These difficulties added to the problem of sample preparation did not make it practical to study the influence on the strength, of wetting and drying cycles from shale samples, using a Hoek cell or controlled stress cell. Therefore, taking into account the factors explained above, it was decided to make direct shear test trough natural discontinuities. Tovar (2008), developed a methodology for preparation of samples of shale to be tested with the direct shear equipment as described in ASTM D 3080. The assembly of the samples in direct shear equipment was made by placing layers of liquid sulphur in the ring to even lower and upper surfaces. Subsequently the samples were stuck firmly by using a glue which ensures that the sample will remain perfectly bonded to the equipment while the test is (see details in Tovar, 2008) Finally, the normal stress was applied until stabilization of the vertical strain gauge reading was reached. Normal stresses used during this research programme were 50, 100 and 200 kPa. The speed of the test was set to minimize the effect of pore pressure especially in wet samples (shear speed 0.0508 mm/min). Each test took approximately 4.5 hours to be completed. Mohr-Coulomb shear strength envelopes were measured for all wetting and drying cycles applied. Figure 3 shows Mohr Coulomb strength envelopes obtained from direct shear tests, which have been previously subject to least squares method adjustment. Tables 3 and 4 summarize parameters C and obtained to different wetting and drying cycles.
160 140 120 100
(kPa)
N=O N=1 N=3
(w) N=1 (H)

(w) N=3 (H)
N=4 (H) (w)
N=1 (Wet) N=0 N=1 N=3 N=2 (Wet) N=3 (Wet) N=4 (Wet)
(w) N=2 (H)

Particulated material - dry
80 60 40 20 0
Desestructurado Hmedo Particulated material - wet
100
150
200
During the wetting and drying process, the sample undergoes irreversible deformations, wich cause cracks, decreasing its resistance and generating blocks of smaller size than the original sample.
(kPa)
Figure 3. MohrCoulomb envelopes for wetting and drying cycles (N) apllied.
1473
250
50
After four wetting and drying cycles, samples were loosing their structure and geometry, making it impossible to handle them for the direct shear test. Tests were performed on de-structured samples in wet and dry conditions in order to measure the lower limit of strength envelopes. After four wetting and drying cycles, samples were loosing their structure and geometry, making it impossible to handle them for the direct shear test. Tests were performed on de-structured samples in wet and dry conditions in order to measure the lower limit of strength envelopes. They are shown in Tables 3 and 4 and Figure 3. After each wetting and drying cycle, the sample undergoes unrecoverable structural damage responsible for the material to behave hystereticly. This behavior may be explained by the loss of sample structure and decreasing of continuity as a result by the micro-cracks appearance. 3.4 Mineralogical analysis
In order to establish the possible influence of mineralogical changes that may occur as a result of degradation of the rock induced by the wetting and drying cycles applied, the results of fluorescence test carried out by Torres (2006) were analyzed. These tests were made for fresh samples and for completely degraded o altered samples. The samples were obtained from the same site
Table 3. Shear strength parameters for dry samples and several wetting and drying cycles. H-S Cycle (N) 1 2 3 Particulated* C (kPa) 0.0 6.0 9.5 3.6 () 35.38 30.14 24.46 22.88
(outcrop) of the samples studied here. The results of this test were used to evaluate the proportion of each of the chemicals present in the rock before and after the degradation process. Table 5 summarizes the test results and allows easy verification of the changes that took place in the chemical elements of rock, and their proportions. The results show that there were no significant changes in chemistry. Proportionally, some minor changes occurred: a greater proportion of Aluminum oxide (Al2O3), which is product of the decomposition from silicate oxide (SiO2). As already mentioned, it is considered that the chemical changes were minimal and could not explain the marked reduction in shear strength occurred after a few cycles of wetting and drying. Observations with the scanning electron microscope (SEM) were made, and it was established that during the degradation process a growth of crystals of calcium carbonate (CaCO3) occurred. They formed a white layer which was observed in the samples as the wetting and drying cycles were occurring. These crystals probably have some impact on the formation of cracks i.e., the growth of the crystals causes an increase in the
Table 5. Main chemical components in fresh and altered samples. Chemical compound Al2O3 CaO Fe2O3 K2O MgO MnO Na2O P2O5 SiO2 TiO2 Fresh sample value 7.56 16.98 1.43 0.75 0.26 0.0 0.1 0.14 54.67 0.74 Unit % % % % % % % % % % Altered sample value 17.82 12.29 5.07 1.59 0.59 0.02 0.40 0.31 41.87 0.61 Unit % % % % % % % % % %
* Particulated material obtained after several wetting and drying cycles (degradation process). Table 4. Shear strength parameters for wet samples and several wetting and drying cycles. Ciclo H-S (N) 1 2 3 4 Particulated* C (KPa) 12.1 8.4 27.2 32.2 5.9 () 24.47 20.82 8.43 5.21 15.27
* Particulated material obtained after several wetting and drying cycles (degradation process).
Figure 4. Development of calcium carbonate crystals. Magnifyed to 800X.
1474
opening of the micro cracks, generating discontinuity planes and degradation of the rock. Calcium carbonate is present in the rock and crystal growth is facilitated by the presence of water and the increase in temperature indicating that these crystals grow especially during the drying process (see Figure 4). 4 ANALYSIS OF RESULTS
120
Shear Strength (%)
100 80 60 40 20 0 0,000 0,500 1,000 1,500 2,000 2,500
n (kg/cm2)
N=2 (H)
N=0
N=1 (S)
N=1 (H)
N=3 (S)
N=3 (H)
N=4 (H)
The values of the angle of shear strength for the least altered samples o fresh samples are very similar in triaxial tests and direct shear tests. It can be said that, within limits, direct shear testing allows the measuring of reliable values of this parameter. Figure 3 shows that as we apply the wetting and drying cycles the values of shear strength angle decrease while the values of C (apparent) increase. Although the value of C is essential for building the envelope, the values obtained from direct shear test can not be taken as the value of real cohesion, because as mentioned before, the samples already had a discontinuity plane before they were tested, so strictly speaking this value should approach zero. Mitchell & Soga (2005) defined cohesion as the existence of shear strength in the absence of any compressive stress in mineral skeleton or in the fault plane. The authors identify three main sources of cohesion: cementation, electrostatic and electromagnetic forces, and finally primary cementing and adhesion; however, and since none of the above sources is what creates the cohesion observed during the tests, we can say that in the case of shale samples analyzed (with a plane of discontinuity) there is an apparent cohesion. The sources of apparent cohesion have also been identified by Mitchell & Soga (2005) and for the particular case of the shales studied in this research, it is due to mechanical forces (i.e., surface roughness). Figure 5 shows the reduction in maximum shear strength and allows to observe the effect of wetting and drying cycles. In this Figure we have calculated the percentage of resistance after each wetting and drying cycle compared to the maximum shear strength from the less altered samples or Fresh samples. The line at 100% represents the shear strength of the less altered or fresh samples, calculated from laboratory data. The results of direct shear tests show that drying and wetting cycles have a significant effect on the shear strength, decreasing the angle of friction and increasing the intercept of the envelope with the passage of W-D cycles. The shear strength reduction was significant (70%) when the samples were subjected to only four W-D cycles.
Figure 5. Shear strength variation for normal stresses and wetting and drying cycles (N) applied.
Figure 6. Void ratio change for normal stresses and wetting and drying cycles (N) applied.
Also, it was found that the observed reduction in resistance can be attributed to the changes of the material structure and the development of microcracks caused by expansion and contraction of clay minerals. This process generates an incre in void ratio as shown in Figure 6 and indicates that as the shale degradation is taking place, its structure becomes more open opposing less resistance during shear stress application. Because of this, as the cycles of wetting and drying were taking place, the material was able to accept more moisture, and this can be taken as an indication of the degradation experienced by the shales. 5 CONCLUSIONS
For the materials studied, the results of direct shear tests show that drying and wetting cycles have a significant effect on the shear strength, decreasing the angle of friction and increasing the intercept of the Mohr Coulomb envelope as the W-D cycles take place. Results of tests conducted during this investigation, show that successive wetting and drying
1475
cycles, generate alteration and destructuration of the sample, and an increased void ratio; there are the main causes of shear strength reduction. For the shale studied, mineralogical changes are negligible (during wetting and drying cycles) and the physical degradation is the main cause for the shear strength reduction. According to observations by using the scanning electron microscope (SEM), it is possible to argue that the expansion of clay minerals and the development of calcium carbonate crystals are the main factors triggering the degradation because they disturb the structure and destroy diagenetic bonds of the shales. After only four wetting and drying cycles, samples reach an advanced degradation state. It is consistent with the short period of time that it takes (under atmospheric conditions) in engineering projects for those materials to change its properties. Considering the tests carried out in this investigation, it was determined that as wetting and drying cycles are applied, the samples increase their ability to absorb moisture. Similarly, in the drying process, samples displace more moisture. It is possible to relate the value of the maximum moisture absorbed or expelled by the samples with the degradation of the rock. A methodology to assess and obtain an envelope of shear strength from the results of the effect of wetting and drying cycles in shales can be proposed. The suggested method uses a parameter named maximum moisture content ( max), which may be used to obtain values of cohesion (C) and friction angle () for three different normal stresses (50 kPa, 100 kPa y 200 kPa), allowing the estimation of Mohr-Coulomb shear strength envelopes. A relationship between shear strength () and the number of cycles of wetting and drying (N) (which may represent rain and drought cycles under real environments) can be proposed. The expressions allow the estimation of the shear strength and indicate how close is the material to physical degradation and loss of geometry, as observed in laboratory tested samples.
REFERENCES
Alonso, E.E. & Pineda, J.A. 2006. Weathering and degradation of shales: experimental observations and models of degradation. XI Congreso Colombiano de Geotecnia y VI Congreso Suramericano de Mecnica de Rocas. Cartagena: Sociedad Colombiana de Geotecnia. Bjerrum, L. 1967. Progressive failure in slopes of overconsolidated plastic clay and clay shales. J. Soil Mech. & Found. Division, ASCE, Vol. 93, pp. 349. Hoek, E. & Franklin, J. (1968). A simple triaxial cell for field and laboratory testing of rock. Trans. Instn Min. Metall. 77, A2226. Marques, E.A.G. 1992. Alteracao e alterabilidade de alguns folhelhos e siltitos da Bacia Sedimentar do Reconcavo, Bahia, Brasil. M.Sc.Thesis, Universidade Federal do Rio de Janeiro, Rio de Janeiro, 1992, pp. 163. Mitchell, J.K. & Soga, K. 2005. Fundamentals of Soil Behavior. 3rd edition. Jhon Wiley and Sons, Inc. New York. Ridley, A.M., Dineen, K., Burland, J.B. & Vaughan P.R. (2003). Soil matrix suction: some examples of its measurement and application in geotechnical engineering. Geotechnique 53, No. 2, 241253. Sadisun, I., Shimada, H., Ichinose, M. & Matsui, K. (2003). Study on the physical disintegration characteristics of Subang claystone subjected to a modified slaking index test. Geotechnical and Geological Engineering Vol. 23, Issue 3 pp. 199218. Taylor, R.K. & Spears, D.A. 1981. Laboratory investigation of mudrocks. Q. Journal of Engineering Geology. Vol. 14, pp. 291309. Taylor, R.K. & Smith, J. 1986. The engineering geology of clay mineral: swelling, shrinking and mudrock breakdown. Clay Minerals, Vol. 21, pp. 235260. Taylor, R.K. 1988. Coal measures mudrocks: composition, classification and weathering processes. Q. Journal of Eng. Geology. Vol. 21, pp. 8599. Torres, M. 2006. Influencia de procesos de descarga en la resistencia y la degradabilidad de rocas arcillosas. Propuesta de tesis doctoral. Universidad Nacional de Colombia. Bogot D.C. Tovar, R. 2008. Influencia de Ciclos de Humedecimiento y Secado en la Resistencia al Corte de Rocas Arcillosas. Tesis Maestra. Universidad Nacional de Colombia. Bogot D.C.
1476
Pore geometry induced suction and the slaking of shales

L.E. Vallejo & A. Stewart-Murphy
Department of Civil & Environmental Engineering, University of Pittsburgh, Pittsburgh, USA
ABSTRACT: Shales are used in the construction of embankments. Many shales slake when in contact with water. Here a micro-structural analysis of how pore geometry influences the degradation of shales is presented. The durability of sixty-eight shale samples from Appalachia was evaluated using a combination of jar slake, thin section, and X-ray tests, as well as fractal theory. Of the sixty-eight samples, fourteen degraded into mud, and fifty-four experienced no degradation. The samples that degraded were composed of kaolinite. Kaolinite shales slake as a result of pore-air induced compression. Pore-air compression is the result of water suction by the pores, and is favored by small pore radii. The samples that slaked had pore diameters equal to or smaller than 0.06 mm. However, there were samples that failed to follow this rule. Using fractal theory, this study presents an explanation for the non-slaking of these samples.
INTRODUCTION
Shale is the most common group of rock material found in the Earths crust. Because of this abundance, shales have been used in the construction of earth embankments for highways as well as earth dams. Shales also form part of the roof system in underground coal mines and of natural or engineered slopes. Many type of shales slake when in contact with water. Thus, shale degradation has been associated with many infrastructure problems such as settlement and instability of highway embankments, subsidence in areas with coal mines, and slope instability (Olivier, 1979; Cummings et al., 1983). The purpose of this study is to present a micro-structural analysis to investigate how pore geometry in shales influence their degradation when in contact with water. The pore geometry will be evaluated using fractal theory (Mandelbrot, 1977). 2 2.1 TESTING PROGRAM Shale samples
the photographs of the thin sections provided information about the pore geometry in the shale samples. 2.2 Durability and thin section analysis
In order to measure the durability of the shales when in contact with water, the jar slake test was used. In the jar slake test, oven dried shale samples are immersed in water for a period of 24 hours. After this period, changes in the shale samples are evaluated using a ranking system (the jar slake index, Ij) which is based on the appearance in the samples after soaking. The ranking system was developed by Lutton (1977) and is shown in Table 1. In the system the appearance of the sample is ranked between a value of 1 and a maximum value of 6 for the jar slake index, Ij.
Table 1. The slake ranking system by Lutton (1977). The jar slake ranking, Ij 1 2 3 4 5 6 Behavior Degrades to a pile of flakes or mud Breaks rapidly and/or forms many chips Breaks slowly and/or forms few chips Breaks rapidly and/or develops several fractures Breaks slowly and/or develops few fractures No change
In order to understand the influence of pore wall roughness on the slaking of shales, sixty eight shale samples were collected from near blasted high walls at surface mines in Kentucky, Tennessee, Virginia, and West Virginia. The samples were subjected in the laboratory to a combination of slake durability tests and thin section analysis. Image analysis of
1477
The Lutton (1977) ranking system was applied to sixty-eight shale samples forming part of this study. Irregular-shaped dry samples weighing 100 grams were used for the jar slake test. Of the sixty-eight shale samples tested, fourteen degraded into mud (Ij 1), and fifty- four experienced little or no change (Ij 5 or 6). Figure 1 shows a sample from Tennessee (TN-5) with Ij 1, and Figure 2 shows a sample from Tennessee (TN-9) with an Ij 6. In addition to the jar slake test, the sixtyeight samples were subjected to X-ray diffraction analysis. The X-ray analysis indicated that the most prevalent clay mineral present in the shales was kaolinite. The X-ray analysis also indicated that no expansive clay minerals were present in the samples. As a complement to the X-ray diffraction analysis, a thin section analysis of the hales was also performed. This later analysis involved the use of thin sections (30 m in thickness) that were first examined optically with a polarizing microscope and then the thin sections were photographed (Vallejo et al., 1993).
Figure 2. Tennessee TN-9 shale samples before and after the jar slake test.
SLAKING MECHANISM
There are different mechanisms discussed in the geotechnical literature which explain the slaking of shales when immersed in water (Surendra et al., 1991; Vallejo, et al., 1993). One slaking phenomenon is attributed to the compression of entrapped air in the pores of the shales when water enters them as a result of capillary suction (Moriwaki, 1974). This entrapped air in the pores exerts tension on the solid skeleton, causing the material to fail in tension. According to Moriwakii (1974), pore-air compression is the predominant slaking mechanism in shales composed primarily of kaolinite. Since the sixty-eight shale samples tested in the jar slake test have kaolinite as the primary clay mineral in their structure, pore-air compression is the primary mechanisms causing their failure. 3.1 Pore pressure
Figure 1. Tennessee TN-5 shale samples before and after the jar slake test.
The mechanism for pore-air compression that breaks the shales was first presented by Vallejo et al.,
1478
(1993) and is shown in Figure 3. Figure 3 shows a shale sample with a system of cylindrical macropores that run continuously through it [Fig. 3(A)]. These macropores which are assumed to be not connected, resemble small cylindrical tubes inside the shale. When the sample is immersed in water, water will be pulled into the individual macropores as a result of capillary forces, and the air that originally filled the macropores will be subjected to compression [Fig 3(B)]. The system of forces acting at the interface between the air and the water in a macropore are presented in Figure 3(C). At equilibrium conditions the following relationship applies: dTs p(d2)/4 u(d2)/4 0 (1)
those shales containing macropores with small diameters. In addition, small diameter macropores more readily confine the air pressure developed during the suction process. That is, the diffusion of the air pressure will decrease with a decrease in surface area (which is a function of the diameter of the macropore) of the pore that is in contact with the air. Thus, the diameter of the macropores in shales has a marked influence on their slaking with water. 3.2 Pore diameter of macropores
where, d is the diameter of the cylindrical macropore, Ts is the surface tension of water acting on the meniscus, p is the air pressure, and u is the pore water pressure at the level of the macropore. From, Eq. (1), the following relationship can be obtained: p u (4Ts)/d (2)
An analysis of Eq. (2) indicates that the pore air pressure, p, in the portion of the macropore filled with air [Fig. 1(B)] increases as the diameter, d, of the cylindrical macropore decreases. Thus, the smaller the diameter of the macropore, the larger is the pore air pressure, p. Since pore-air compression is favored by small pore radii, slaking of the shales by air compression will be more pronounced in
The pore diameter of the sixty-eight shale samples was obtained from thin sections of the shales. Photographs of the thin sections were made using a polarizing microscope. From the photographs, the cross sectional areas and shapes of the perimeters of the pores in the shales were obtained. Using standard digitizing procedures, the profiles of the pores were obtained from the photographs of the thin sections. Figure 4 shows the typical results of the process for the case of shales TN-5 and TN-9 (Figs. 1 and 2). The average diameters of the pores of the TN-5 sample was equal to 0.053 mm. The average diameter of the pores of TN-9 sample was equal to 0.107 mm. The jar slake index, Ij, for TN-5 was equal to 1. The jar slake index, Ij, for TN-9 was equal to 6. The fourteen samples that degraded
Figure 3. Forces acting in a macropore in a shale sample.
Figure 4. Pore geometry of samples TN-5 and TN-9.
1479
into soils had average diameters that were equal or smaller that 0.06 mm. Fifty four samples did not slake. Most of these samples had pore diameters that were greater than 0.06 mm. However, eight of the fifty four samples had pore diameters that were smaller than 0.06 mm. An explanation for why these eight samples failed to follow the smaller than 0.06 mm pore diameter rule follows next. This explanation uses fractal theory. 4 ANALYSIS AND INFLUENCE OF PORE ROUGHNESS
The roughness of the pore boundaries was evaluated using fractal theory. Fractal theory makes use of a number called the fractal dimension, D, to evaluate the degree of irregularity of objects in nature (Mandelbrot, 1977). In this study the fractal dimension was used to measure to measure the degree of roughness of the pore boundaries of the 68 shale samples. The fractal dimension of the pore boundaries were calculated using the area-perimeter method (Korvin, 1992; Hyslip and Vallejo, 1997). The fractal dimension, D, is obtained from the slope, m, of the best fit line that connects the values of the area and perimeter for each of the pores in a shale sample (Fig. 4). Figure 5 shows this plot for the samples TN-5 and TN-9. Once the slope, m, is obtained, the dimension id calculated from the ratio between 2 and m (that is D 2/m). The fractal dimension, D, measures the average roughness of the pore boundaries in the shale samples. The higher the value of D, the rougher are the pore boundaries in the shales. The fractal dimension, D, for the pore boundaries of sample TN-5 (Fig. 4) was equal to 1.1989 (Fig. 5). D for sample TN-9 was equal to 1.4587 (Fig. 5) 4.1 Pore diameter and degree of slaking
Figure 5. Area-perimeter method to obtain fractal dimension values for samples TN-5 and TN-9.
As previously mentioned, from the jar slake tests on the 68 shale samples, fourteen slaked (Ij 1). These slaked samples had average pore diameters equal or smaller than 0.06 mm. Of the 54 samples that did not slake (Ij 5 or 6), 8 samples had diameters smaller than 0.06 mm. Thus, the size of the pore alone does not seem to explain completely why the shales slake as a result of pore air-compression as stated by Eq. (2). 4.2 Pore roughness and slaking of shales
Even though the size of the pores in shales is a good indicator for the slaking susceptibility, it was not a parameter that indicated without a doubt the slaking behavior of shales. Table 2 shows the results
of the jar slake tests and the fractal analysis on the 8 samples that did not slake even though the average size of their pores was less than 0.06 mm. The fractal analysis for these shales indicated that their pore walls were very rough (Fig. 6). These high degree of roughness of the pore walls is reflected on the high values of the fractal dimension for the pore perimeters. An explanation for the lack of slaking of the shales of Table 2 that have small but rough pores seems to rest on the degree of roughness of the pore boundaries (Fig. 6). According to Ransohoff and Radke (1988) when a capillary tube with either square or irregular cross section as shown in Figure 7 is immersed in water, the water does not advance in the capillary tube following the whole cross sectional area, but advances in the tube following the corners and crannies of the tube. This partial filling of the tube cross sectional area will prevent the development of air pressure that is necessary to cause the slaking of
1480
Figure 6. Pore geometry of non-slaking TN-17 shale sample (d 0.056 mm and D 1.4762, see Table 2).
Table 2. Properties of the 8 shale samples that did not slake. Diameter (mm) 0.056 0.045 0.056 0.052 0.051 0.055 0.052 0.051 Jar slake index, Ij 6 6 6 6 6 5 5 5 Fractal dimension D 1.7767 1.4246 1.4762 1.4907 1.4900 1.6020 1.5473 1.4406 Figure 8. Advancement of water in a rough macropore. No pore air compression develops.
Sample TN-7 TN-16 TN-17 KY-19 KY-22 KY-32 WV -8 WV-14
the pore walls will prevent the development of the air pressure that is required to cause the breakage of the shale [Fig. 1(B)]. Thus, the roughness of the boundaries of the pores (measured by the fractal dimension) has a significant influence on the slaking of the shales and needs to be considered in any evaluation of the durability of shales in water. 5 CONCLUSIONS
The slaking of shale samples in contact with water was studied using a combination of slake durability tests, thin section photographs and fractal theory. Form these analyses, the following conclusions can be made: 1. The slaking of the shales was the result of pore air compression. 2. The slaking by pore air compression was directly related to the average pore diameter and the roughness of the pore boundaries. The smaller the diameter and the smoother the boundaries of the pores, the more pronounced was the slaking of the shales by air compression. 3. The roughness of the pore boundaries in the shales was determined using the fractal dimension concept from fractal theory. The larger the fractal dimension, the rougher were the boundaries of the pores. 4. Regardless of the diameter of the pores, shales with very rough pores did not slaked as a result of pore air compression. These rough pores moved the capillary water through their
Figure 7. Water advance in capillary tubes with irregular or rough boundaries.
the shales Figure 8 shows how water will advance in a rough pore such as that shown in Figures 6 and 7. The water at the extreme portions of the macropore will cover its whole cross sectional area, however, after a short distance from the ends of the macropore, the water will follow the corners and crannies of the pore walls (Fig. 7). This movement of the water through the corners and crannies of
1481
corners and crannies. Thus, the water did not move through the whole cross sectional area of the rough pores. This type of water movement prevented the development of pore air compression that is needed for the breaking of the shales.
REFERENCES
Cummings, R.A., Singh, M.M. & Moebs, N.N. 1983. Effect of atmospheric moisture on the deterioration of coal mine roof shales. Mining Engineering, 35 (3):243245. Korvin, G. 1992. Fractal Models in the Earth Sciences. Elsevier, Amsterdam. Hyslip, J.P. & Vallejo, L.E. 1997. Fractal analysis of the roughness and size distribution of granular materials. Engineering Geology, 48 (3, 4): 231244. Lutton, R.J. 1977. Design and Construction of Compacted Shale Embankments: Slaking Indices for Design. Report FHWA-RD-1, Federal Highway Administration, Washington, D.C.
Mandelbrot, B.B. 1977. Fractals: Form, Chance and Dimension. W.H. Freeman, San Francisco. Moriwaki, Y. 1974. Causes of Slaking of Argillaceous Materials. Ph.D. Dissertation, Dept. of Civil Engineering, University of California at Berkeley. Olivier, H.J. 1979. A new engineering-geological rock durability classification. Engineering Geology, 14: 255279. Ransoff, T.C. & Radke, C.J. 1988. Laminar flow of a wetting liquid along corners of predominantly gas occupied non circular pores. Journal of Colloidal and Interface Sciences, 121 (2): 391401. Surendra, M., Lovell, C.W. & Wood, L.E. 1981. Laboratory studies of the stabilization of non-durable shales. Transportation Research Record, 79: 3340. Vallejo, L.E., Welsh, R.A., Lovell, C.W. & Robinson, M.K. 1983. The influence of fabric and composition on the durability of shales . In: Rock for Erosion Control, C,H. McElroy, & D. Lienhart, Editors, ASTM STP 1117:: 1528.
1482
Influence of suction on the mechanical behaviour of rock joints under shear loading
Maria Teresa Zandarin, Eduardo Alonso & Sebasti Olivella
ABSTRACT: A direct shear equipment was modified for testing rock joints with suction control. Modification includes a vapour forced convection circuit connected to the air chamber of shear cell for controlling relative humidity during testing. The main features of the device are described and test program procedures are presented. In addition the results of tests conducted under different suction, net normal stress and asperity roughness angle conditions are presented and analyzed. Finally, new expressions of resistance parameters of hyperbolic failure envelope are proposed incorporating the effect of suction and asperity roughness angle. 1 INTRODUCTION originally developed by researchers of Department of Geotechnical Engineering and Geosciences of Technical University of Catalonia. Further refinements of the equipment were done to carry out test with suction control. A vapor circuit was connected to the shear cell; and the data acquisition system was improved incorporating an analogue data acquisition device which allows to record automatically shear forces, vertical displacements, temperature and relative humidity. 2.2 Shear cell The basic device consists of a shear box to fix the sample and two mechanisms for applying horizontal displacements and vertical loads. A push rod moved by stepper motor applies the horizontal displacement controlling the rate on the shear cell. The vertical load is applied by air pressure. The pressure is transmitted by a Belofram diaphragm on the piston which moves with the centering bushing. The amount of air pressure is controlled by a valve and a manometer. A cross section of the cell is shown in Figure 1. 2.3 Vapor force convection circuit system The system consists of an air pump that impulses the vapor produced by a saline solution contained in a closed recipient. The system is closed and continuously impulses the vapor contained in the recipient towards the air pressure chamber of the shear cell and extracts the air from it. This allows keeping constant the relative humidity within the air chamber during the test.
Many researchers have studied rock joints experimentally and theoretically. Most of the previous laboratory experiments for the mechanical properties have been focused on determining the peak shear strength and the stress-displacements relation under unidirectional shear loading (Barton et al, 1987). Also, the behaviour of rock joints under cyclic shear loading and the asperity degradation due to this load conditions have been reported by Lee et al. (2001). On the other hand, several empirical strength criteria, mostly in the form of power equations (Hoek & Bray (1981), Hassani & Scoble (1985) and Nilsen (1985)) or logarithmic equations (Barton (1973) Barton & Choubey (1977)) have been reported in the literature. Gens et al. (1985) and Carol et al. (1987) have proposed a hyperbolic function for describing non-linear failure surface for rock joints. However, a lack in the literature is the treatment the mechanical behaviour of rock joints with different moisture content. Therefore, an experimental program has been developed and it is described in this paper to test rock joints with suction control. Furthermore, from the analysis of test results a new model to calculate the strength parameters for a hyperbolic failure envelope is proposed considering the effect of suction. 2 DIRECT SHEAR EQUIPMENT
2.1 General description The tests were performed using a Direct Shear Cell with Suction Control. The equipment was
1483
Samples were prepared from two specimens of 50 mm of diameter and 12 mm of height drilled from a borehole core Lilla claystone. The joint surfaces were carved with a diamond drill in each specimen. The joints obtains had different asperity roughness angle a 0, 5, 15, 30, and 45 degrees respectively (Fig. 2). 3.2 Equilibration with RH environment
Figure 1.
Cross section view of the shear cell.
2.4
Data acquisition
The shear loads are measured by a load cell with a nominal capacity of 500 kg. The vertical displacements are measured by a Linear Variable Displacement Transducer (LVDT) with a range of measure of 5 mm. A humidity and temperature transmitter measure the relative humidity and temperature of air within shear cell during tests. The measurements are recorded automatically on PC windows through a multifunctional analogue to digital, digital to analogue input/output board using a USB6009 data acquisition device. A digital program was designed by using LabVIEW programming language for building data acquisition. 3 EXPERIMENTAL PROCEDURE
The equalization with a relative humidity of the samples was performed by placing them in a desiccator with a solution with known concentration at a constant temperature of 20C (Romero, 2001). The thermodynamic relationship between rock total suction and relative humidity of the reference is given by the Kelvins law (Fredlund and Rahardjo, 1993). Some of the samples were dried up placing them with pure lithium chloride. The pure lithium chloride reaches a RH 20% (approximately a total suction 200 MPa). Other samples were wetted using distillate water. A quasi-saturation condition was obtained with a RH86% and a 20 MPa. And other samples were exposed to the laboratory room environment (RH50%, 100 MPa). The equalization was considered complete when there was no appreciable change in the weight of samples (no changes in water content). The total suction measured on a small sample of rock using a dew-point psychrometer (WP4, Decagon Device) was used as a standard average value for samples with joints. 3.3 Direct shear testing Once samples reached the equalization they were placed in the shear cell. The joint was oriented normally to shear displacements direction. The relative humidity within the shear cell began to be controlled by the vapor system. Then, net normal stress was applied; tests were performed for 30, 60 and 150 KPa respectively. After the normal displacements reached a constant value the shear displacement was applied with a rate of 0.05 mm/min.
The testing procedure begins with the preparation of the samples by carving joints with different geometric angles; equilibration of the samples with a prescribed relative humidity applying the diffusion vapor equilibrium technique; and the application of the direct shear load. 3.1 Samples preparation
The material used for samples is Lilla claystones. This rock has two main components; the host argilliceus matrix (composed by illite, paligorskite, dolomite and quartz); and the sulphated crystalline fraction (composed mainly of anhydrite and gypsum (Berdugo, 2007). The density of the rock varies from 2.56 to 2.58 g/cm3. The clay matrix has a low plasticity (IP 5%), with high in situ moisture (w 3.24.1%). The porosity varies from 0.09 to 1.1. The Young Modulus varies from 26.5 to 28.5 GPa and the shear stiffness varies from 11 to 12.5 GPa (Pineda, 2008).
Figure 2. and 45.
Rock joints samples with a 0, 5, 15, 30
1484
120
Normal displacements [mm]

0 2 4 6
t [KPa]
The shear load and normal displacements during the test were automatically recorded by the data acquisition system each second. The shear test finishes when the residual shear stress is reached. 4 EXPERIMENTAL RESULTS AND DISCUSSION
a) Net Normal Stress 30 KPa

160 0.3 0.2 0.1 0 -0.1 0 2 4 6
80 40 0
Shear displacements [mm]
The results of the tests are plotted in Figures 37 for the different asperity roughness angle of the joints (a 0, 5, 15, 30 and 150) and for the different normal stresses a) 30, b) 60 and c) 150 KPa. The shear stress-shear displacement curves and the normal displacements-shear displacement curves are displayed. Three curves belonging to the different values of suction used in the tests are represented for comparison. 4.1 Shear strength
b) Net Normal Stress 60 KPa

200
Normal displacements [mm]

0 2 4 6
250
0.4
t [KPa]
150 100 50 0
0.2

c) Net Normal Stress 150 KPa Normal displacements [mm]
400 300 0.2
The measured shear strength of the joints depends on three variables; normal stress; suction; and joint roughness angle. The effect of the normal stress is well known; at larger normal stress the shear strength is also larger.
a) Net Normal Stress 30 KPa Normal displacements [mm]
80 60 0.05 0 -0.05 -0.1
0.1 0
Suction [MPa]
200 100 20
t [KPa]
200 100 0 0 2 4 6 8
-0.1 0 2 4
Figure 4.
Shear test results for the joint with a 5.
t [KPa]
a) Net normal stress 30 KPa Normal displacements [mm]

0 1 2 3 4 400 300 0.8 0.6 0.4 0.2 0 0 1 2 3 4
40 20 0 0 1 2 3
t [KPa]
-0.15 -0.2 0 1 2 3
200 100 0

b) Net Normal Stress 60 KPa Normal displacements [mm]
120 0.05 0 -0.05
Shear displacements [mm] b) Net normal stress 60 KPa Normal displacements [mm]
0 2 4 6 8 400 300 1 0.8 0.6 0.4 0.2 0 0
t [KPa]
80
40
t [KPa]
0 1 2 3
-0.1 -0.15 -0.2
200 100
0 0 1 2 3
c) Net Normal Stress 150 KPa Normal displacements [mm]

200 160 0.05 0 -0.05 -0.1 -0.15 -0.2 0
Suction [MPa]
200 100 20
Shear displacements [mm] c) Net normal stress 150 KPa Normal displacemnts [mm]
600 1.6 1.2 0.8 0.4 0 0

Suction [MPa]
200 100 20
t [KPa]
80 40 0 0
t [KPa]
1 2 3 4
120
400
200
0 0 2 4 6 8
Figure 3.
Figure 5.
1485

400 300 0.8 0.6 0.4 0.2 0 0 1 2 3 4
200 100 0 0 1 2 3 4

b) Net normal stress 60 KPa Normal displacements [mm]
400 300 1 0.8 0.6 0.4 0.2 0 0
200 100 0 0 1 2 3 4 5
Shear displacements [mm] c) Net normal stress 150 KPa Normal displacemnts [mm]
600 1.2

Suction [MPa]
200 100 20
t [KPa]
400
0.8
200
0.4
0 0 1 2 3 4 5
0 0 1 2 3 4 5
The value of suction imposed also affects the peak and residual shear strength. The greater the suction is the greater the values of peak shear strength are obtained. However, the effect of suction on the residual strength is not seen as clearly. Because residual strength not only depends on the suction, but it also depends on the degradation of the asperities. And degradation of asperities is influenced not only by suction but also by the irregular matedness due defects of joint construction and by the heterogeneity of the rock. This implies the existence of contact areas with higher or lower strength. Moreover, it is also shown that larger asperity roughness is associated with larger strength. Furthermore the roughness also affects the strength softening of the joint. In the joints with higher roughness the residual strength is reached for smaller displacements. For example, for joints with roughness of 45 the residual strength is reached for a displacement of approximately 2.5 mm, while for a joint with roughness of 15 the residual strength occurs for a displacement of 6 mm. And the flat joint shows a ductile behaviour; 4.2 Dilatancy
t [KPa]
t [KPa]
Figure 6.

300 1 0.8 0.6 0.4 0.2 0 0
t [KPa]
200
100
Shear displacements [mm] b) Net normal stress 60 KPa

500 400
Normal displacemnts [mm]
0.8 0.6 0.4 0.2 0 -0.2 0 1 2 3 4
t [KPa]
300 200 100 0 0 1 2 3
Shear displacements [mm] c) Net normal stress 150 KPa Normal displacements [mm]
800 600 0.8 0.6 0.4 0.2 0 -0.2 0

Suction [MPa]
200 100 20
The dilatancy of the samples is irregular. Not always for a higher normal stresses it was registered a lower dilatancy (see Fig. 5 and Fig. 6 a, b). It is believed that this anomalous behaviour is due to the irregularity of matdness; the heterogeneity of the rock; the degradation of asperities; the quantity and shape of the gouge material. In the plots it is shown that rougher asperities (45) leads to lower dilatancy. This is because these asperities present more degradation during the test. And the flat joint presents a smooth contractive behaviour. From the charts the influence of suction on dilatancy can also be observed. The joints with less suction ( 20 MPa, RH 86%) undergo lower dilatancy, whereas if the suction increases the dilatancy also increases. Suction increases strength of the rock matrix and therefore; the sliding of the joints walls one over another up occurs without breakage. However, if breakage occurs it is more probable that the gouge material with higher suction roll-on over the joint surface. In contrast, this material with lower suction is easier to crush without rolling. 5 INFLUENCE OF SUCTION AND JOINT ROUGHNESS ON RESISTANCE PARAMETERS
t [KPa]
400 200 0 0 1 2 3
Figure 7.
Shear test results for the joint with a 45
The shear strength can be expressed in terms of a two parameter hyperbolic failure surface (Gens, et al., (1985) and Carol, et al., (1997)):
1486
F 2 (c0 - - tan 0)2
(1)
where the shear stress; c0 the initial effective cohesion; the net normal stress; and tan0 the tangent of initial effective angle of internal friction. Both parameters c0 and tan0 define an envelope curve. The envelope curve for each test was obtained by plotting the maximum shear stress measured against net normal stress (Fig. 8). Then, c0 is the distance between the origin and the intercept of the asymptote with axis in the - plane. And tan0is the slope of the asymptote. After c0 and tan0 were drawn against asperity roughness angle (a) and suction (Figs. 9a, b and 10a, b respectevily). In Figure 9a minimum values of cohesion are observed for a 0, but then sharply increase for a 15. And it almost maintains constant for other values of a. In Figure 9b it is shown a linear increment of c0 with suction. Next, the mathematical expression proposed to c0 taking account the influence of suction and asperity roughness angle is: c0 ( , a ) (c 0 c1 ) ( b0 b1 )(1 e
Asperity angle 0
800 800
where c0(,a) effective initial cohesion; the total suction; c0 cohesion for 0 and a 0; c1 slope of c0 vs. suction line for a 0; b0 average value of c0 for a 1545; b1 parameter of the model that control the increment of cohesion with for a 1545; and b2 parameter of the model. In Figures 10a, b it is observed that tan 0 increases with a and suction. The increment
400 300
a)
Suction [MPa]
200 (Experimental) 200 (Model) 100 (Experimental) 100 (Model) 20 (Experimental) 20 (Model)
c'0 [KPa]
200 100 0 0 400 300 10 20 30 40 50
aa
b) Asperity angle
0 (Experimental) 0 (Model) 5 (Experimental) 5 (Model) 15 (Experimental) 15 (Model) 30 (Experimental) 30 (Model) 45 (Experimental) 45 (Model)
b2 tan a
(2)
c'0 [KPa]
200 100 0 0 50 100 150 200 250
Asperity angle 5 Shear Stress [KPa]
Shear Stress [KPa]
600 400 200 0 0 40 80 120 160
600 400 200 0 0 40 80 120 160
Suction [MPa]
Net Normal Stress [KPa]

Parameters of failure surfaces c'0=58.69KPa tgF'0=0.69 F'0=34.61 c'0=28.42KPa tgF'0=0.84 F'0=40.03 c'0= 9.22KPa tgF'0=0.86 F'0=40.70

Parameters of failure surfaces c'0=128.69KPa tg F'0=1.30 F'0=52.43 c'0= 75.00KPa tg F'0=1.02 F'0=45.56 c'0= 56.00KPa tg F'0=0.90 F'0=41.98
Figure 9.
3 2.5
a. c0 vs. aa and b. c0 vs. suction.
Asperity angle 15
800 800
Asperity angle 30 Shear Stress [KPa]
a)
Suction [MPa]
200 (Experimental) 200 (Model) 100 (Experimental) 100 (Model) 20 (Experiemntal) 20 (Model)
Shear Stress [KPa]
600 400 200 0 0 40 80 120 160
600 400 200 0 0 40 80 120 160
tg F '0
2 1.5 1 0.5 0 3 10 20

Parameters of failure surfaces c'0=307.60KPa tg F'0=1.81 F'0=61.07 c'0=225.00KPa tg F'0=1.10 F'0=47.72 c'0=184.00KPa tg F'0=0.90 F'0=41.98
Parameters of yield surfaces c'0=292.85KPa tg F'0=1.79 F'0=60.81 c'0=237.27KPa tg F'0=1.27 F'0=51.78 c'0=159.28KPa tg F'0=0.91 F'0=42.30
aa
30
40
50 Asperity angle
Asperity angle 45
800
Shear Stress [KPa]
600 400 200 0 0 40 80 120 160
tg F'0
Suction [MPa] 200 (Experimental) 200 (Fit) 100 (Experiemtal) 100 (Fit) 20 (Experimental) 20 (Fit)
2.5 2 1.5 1 0.5 0
b)

Parameters of failure surfaces c'0=296.00KPa tg F'0=2.40 F'0=67.38 c'0=244.38KPa tg F'0=1.39 F'0=54.26 c'0=168.00KPa tg F'0=0.90 F'0=41.98
0 (Experimental) 0 (Model) 5 (Experimental) 5 (Model) 15 (Experiemental) 15 (Model) 30 (Experimental) 30 (Model) 45 (Experimental) 45 (Model)
50
100
150
200
250
Suction [MPa]
Figure 8. Resistance parameters of hyperbolic failure surface.
Figure 10.
a. tan0 vs a and b. tan0 vs suction.
1487
Table 1. Parameters used for adjustment of experimental values c(, a) and tan(, a). Parameter c0 c1 b0 b1 b2 t0 t1 d0 d1 Value 2.8 KPa 0.3 170.0 KPa 0.3 5.0 0.7 0.001 0.2 0.008
rougher joint implies a more brittle behaviour. This brittle behaviour causes higher damage on the joints surfaces and lower dilatancy. The mathematical expressions proposed in this study for the initial effective cohesion and initial effective tangent of internal friction angle taking account the influence of suction and asperity roughness angle closely replicate the experimental results. REFERENCES
Barton, N. (1973). Review of anew shear strength criterion for rock joints. Engng Geology, 7, 287332. Barton, N. (1977). The shear strength of rock and rock joints. Inr. J. RockMech. Min. Sci. 13, 255279. Barton, N. & Choubey, V. (1977). The shear strength of rock joints in theory and practice. Rock Mech. 10, I 54. Berdugo, I.R. (2007). Tunnelling in sulphate-bearing rocks expansive phenomena. PhD Thesis. Department of Geotechnical Engineering and Geosciences, UPC. Carol, I., Prat, P. & Lpez, C.M. (1997). A normal/ shear cracking model. application to discrete crack analysis. ASCE Journal of Engineering Mechanics, 123(8):765773. Fredlund, D.G. & Rahardjo, H. Soils mechanics for unsaturated soils. New Wiley & Sons. Gens, I. Carol, & Alonso, E.E. (1990). A constitutive model for rock joints; formulation and numerical implementation. Computers and Geotechnics, 9:320. Hassani, F.P. & Scoble, M.J. (1985). Frictional mechanism and properties of rock discontinuities. In Proc. Inf. Symp. Fundamentals Rock Joints. Bjorkhden, Centek, pp. 185196. Hoek, E. & Bray, J.W. (1981). Rock Slope Engineering, 3rd edn. The Institution of Mining and Metallurgy, London. Lee, H.S., Park, Y.J., Cho, T.F. & You, K.H. (2001). Influence of asperity degradation on the mechanical behaviour of rough rock joints under cyclic shear loading. Int. Journal of Rock Mechanics & Mining Science, 38 967980 Nilsen, B. (1985). Shear strength of rock joints at low normal stresses-a key parameter for evaluating rock slope stability. Fundamentals of rock joints. In Proc. Ini. Symp. &ndamenials Rock Joints, Biorkliden, Centek. DD. 487494. Pineda, J.A., De Gracia, M. & Romero, E. (2010). Degradation of partially saturated argillaceous rocks: influence on the stability of geotechnical structures. 4th Asia-Pacific Conference on Unsaturated Soils, Newcastle, Australia. Unsaturated soils-Buzzi,Fityus & Sheng (eds.). Taylor & Francis Group. Romero, E.E. (2001). Controlled suction techniques. Proc 4 Simposio Brasileiro de Sols Nao Saturados. Gehling and Schnaid Edits. Porto Alegre, Brasil, pp. 535542. Tarrag, D. (2005). Degradacin mecnica de argilitas sulfatadas y su efecto sobre la expansividad. BSc dissertation. Universitat Politcnica de Catalunya, Barcelona.
respect a is considered dependent on the tana; and the increment with respect to suction is considered linear. The equation proposed for tan 0 is: tan 0 ( , a ) (t 0 t1 ) (d 0 d1 )tan a (3)
where tan 0(, a) the tangent of the effective initial angle of internal friction; the total suction; t0 the value of tan 0 for 0 and a 0; t1 slope of tan 0 vs. suction line for a 0; d0 and d1 model parameters which control the increment of tan 0 with suction for a 15-45; and tana is the geometric tangent of the asperity roughness angle. In Figures 9a, b and 8a, b it is also shown the fitting of the experimental values of c0(, a) and tan 0(, a) with the equations previously proposed and the parameters are listed in Table 1. 6 CONCLUSIONS
A direct shear device was successfully modified to test rock joints under control of relative humidity of the specimens. Modifications included the addition of a vapour circulation system and incorporation of an analogue data acquisition device. The vapour circulation system allows studying the influence of suction on the mechanical behaviour of the joints. The carving process adopted to construct different asperity roughness angle allowed to explore its effects on the shear strength and dilatancy of joints. The shear test results obtained not only show the well known dependency of the shear strength with the net normal stress but also show a marked dependency on suction and asperity roughness angles. It was observed a decrease of the shear strength and of dilatancy when suction decreases. Also, it was observed that greater asperity roughness angle implies greater shear strength. However,
1488

Unsaturated PDF

Uploaded by

Copyright:

Available Formats

Unsaturated PDF

Uploaded by

Document Information

Original Title

Copyright

Available Formats

Share this document

Share or Embed Document

Sharing Options

Did you find this document useful?

Is this content inappropriate?

Copyright:

Available Formats

Unsaturated PDF

Uploaded by

Copyright:

Available Formats

UNSATURATED SOILS

Eduardo Alonso & Antonio Gens

Technical University of Catalonia, Barcelona, Spain

VOLUME 1 Keynote lectures

Unsaturated soil behaviour

157 161 167

Stress-strain and strength characteristics

205 211 217 223

253 259 265 271

277 283 289

309 317 323 329

335 341 347 353 359 365 371

375 381 387 391 397

Water retention and hydraulic behaviour

429 433 441

447 453 461 467

473 479 485 491 495

Thermal and chemical effects

573 579 585

Testing techniques: Laboratory, in situ and centrifuge

951 957 965 973 979 987 993

Analytical and numerical modelling

1067 1075 1081 1087 1093

Engineering applications and case histories

1161 1169 1175

1183 1189 1195 1201 1209

1273 1279 1285

1407 1413 1419 1425 1431 1437

1459 1465 1471 1477 1483

Design of stiffened slabs-on-grade on shrink-swell soils

Elevated Structural Slab on Piers

Thin Post Tensioned Slab

Figure 1. Some foundations used for light buildings on shrink-swell soils.

Figure 2. Geometry and boundary conditions for an edge lift case.

Figure 3. Geometry and boundary conditions for an edge drop case.

0.09 0.08 0.07

Soil mound and foundation elevations

0.06 0.05 0.04 0.03 0.02 0.01 0.00 -0.01 0 1 2 3 4 5 6 7 8

Initial Mound Elev. Final Mound Elev. Final Found Elev

Shearing Force, V Bending Moment, M

100 80 60 40 20 0 -20 -40 -60 0 1 2 3 4 5

Figure 5. Bending moments and shearing forces results.

Influence of soil shrink-swell potential on Leqv

Normalized Leqv (Leqv/Leqv (reference case))

1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5

Edge drop case Edge lift case

Normalized Lss (Lss/Lss (reference case))

Influence of depth of active moisture zone on Leqv

1.2 1 0.8 0.6 0.4 0.2 0 2 0

Influence of soil surface suction change on Leqv

Normalized Leqv (Leqv/Leqv (reference case))

1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5

Edge drop case Edge lift case

Edge drop case Edge lift case

Normalized H (H /H (reference case))

Normalized U0 (U0 /U0 (reference case))

Normalized Leqv (Leqv/Leqv (reference case))

1.2 1 0.8 0.6 0.4 0.2 0 0 0.5 1 1.5 2

Edge drop case Edge lift case